Personal Networks - Telenor

Personal Networks

1 Guest Editorial;Ramjee Prasad

4 Wireless Personal Area Networks – The PACWOMAN Vision;Yaoda Liu

12 Personal Networks as Business Strategy for the Wireless CommunicationFuture; Knud Erik Skouby, Karsten Vandrup

17 PN Business Models and Strategies – The Operator’s Perspective;Su-En Tan, Rune Roswall

26 Interconnection and Billing Policies for Personal Networks;Rajeev R Prasad, Vasileios S Kaldanis

34 Extending Private Personal Area Networks to Personal Network Federations inHeterogeneous Ad Hoc Scenarios; Luis Sanchez, Jorge Lanza, Luis Muñoz

45 Personal Networks – An Architecture for 4G Mobile Communications Networks;Anthony Lo, Weidong Lu, Martin Jacobsson, Venkatesha Prasad,Ignas Niemegeers

59 Wide-Area Publish/Subscribe Service Discovery – Application to PersonalNetworks; Wassef Louati, Djamal Zeghlache

70 Challenges and Solutions in Achieving Personalisation Through ContextAdaptation; Rasmus L Olsen

85 Personal Network Directory Service; Nikko Alutoin, Sami Lehtonen, Kimmo Ahola, Jori Paananen

93 Risk Analysis in an ‘Insecure Wireless World’;Sofoklis Kyriazakos, Neeli Prasad

101 Coexistence Concept for the Implementation of LDR/HDR WPAN MultimodeDevices; Mauro De Sanctis, John Gerrits, Julian Pérez Vila

113 The Unpredictable Future – Personalized Services and ApplicationsArchitecture; Mary Ann Ingram, Ramjee Prasad, Kim Skaue

Status

125 ITU Plenipotentiary Conference 2006 – PP-06, Antalya, 6-24 November 2006– An Overview of Main Results of the Conference;Anne Lise Lillebø

Contents

Telektronikk

Volume 103 No. 1 – 2007

ISSN 0085-7130

Editor:

Per Hjalmar Lehne

(+47) 916 94 909

[email protected]

Editorial assistant:

Gunhild Luke

(+47) 415 14 125

[email protected]

Editorial office:

Telenor R&I

NO-1331 Fornebu

Norway

(+47) 810 77 000

[email protected]

www.telektronikk.com

Editorial board:

Berit Svendsen, VP Telenor Nordic

Ole P. Håkonsen, Professor NTNU

Oddvar Hesjedal, VP Project Director

Bjørn Løken, Director Telenor Nordic

Graphic design:

Design Consult AS (Odd Andersen), Oslo

Layout and illustrations:

Gunhild Luke and Åse Aardal,

Telenor R&I

Prepress and printing:

Rolf Ottesen Grafisk Produksjon, Oslo

Circulation:

3,700

Networks on networks

Connecting entities through networks – in

technological, societal and personal terms –

enables telecommunication. Networks occur on

different levels, form parts of larger networks,

and exist in numerous varieties. The artist Odd

Andersen visualises the networks on networks

by drawing interconnected lines with different

widths. Curved connections disturb the order

and show that networks are not regular but are

adapted to the communication needs.

Per H. Lehne, Editor in Chief

1

Danish King Harald Blåtand – for whom, a thousandyears later, Bluetooth for wireless personal area net-works was named – is known for uniting parts ofSweden, Denmark and Norway. Uniting computers,mobile phones, and personal devices is the goal ofWireless Personal Area Networks (WPANs), whichare meant to become a major part of future mobilecommunication networks and the future generation(FG). This introduction provides an abstract view ofwhat a WPAN is, or should look like.

The Personal Area Network (PAN) is a network foryou, for you and me, and for you and the outer world.It is based on a layered architecture where differentlayers cover the specific types of connectivity (seeFigures 1 – 3).

Guest Editorial – Personal Networks

R A M J E E P R A S A D

Ramjee Prasad

is Director of

Center for Tele-

infrastruktur

(CTIF) at

Aalborg Univer-

sity, Denmark

Telektronikk 1.2007

Figure 2 PAN is for you and me. When people and appliances meet, PAN becomes a dynamic distributedapplication platform where gatekeepers are needed

This connectivity is enabled through the incorpora-tion of different networking functionalities into thedifferent devices. So, for the stand-alone PAN, theperson is able to address the devices within his per-sonal space independently of the surrounding net-works. For direct communication of two persons(i.e. their PANs), the bridging functionality is incor-porated into each PAN. For communication throughexternal networks, a PAN implements routing and/orgateway functionalities.

Layer-oriented scalable architecture supports the func-tionalities and protocols of the first three layers andprovides the capability to communicate with the exter-nal world through higher layer connectivity. It pro-vides the appropriate middleware structures and con-sists of a well-defined protocol stack, with identifiedinformation transfer through appropriate interfaces.

The PAN can use various access technologies, callingfor reconfiguration. Moreover, according to the appli-cations, PAN systems provide automatic service andresource discovery, provide QoS (e.g. for multimedia

applications), and are scalable in terms of network size.

PAN invisibility is essential to the user, and so, PAN devices are able to adapt themselves automatically to the environment and can, for

Sensor oractuator

Computationengine

Radio

SP

SP

SPSP

SP

SP

SP

SPSP

SP

SP

SP

SPSP

SP

SA

Figure 1 PAN is for you. A PAN constructs apersonal sphere of smart peripherals

ISSN 0085-7130 ©Telenor ASA 2007

2 Telektronikk 1.2007

example, download the appropriate applications andaccess techniques automatically.

Frequency planning and coexistence with the existingsystems is important for designing novel PANs.PAN-oriented applications mostly use the unlicensedfrequency bands. For the higher data rates, the 5 GHzfrequency band, and possibly the 60 GHz, can beused (Figure 4).

The concept of the personal network (PN) goesbeyond the commonly accepted concept of a PAN.The latter refers to a space of small coverage aroundthe person where ad hoc communication occurs.This is also referred to as a personal operating space(POS). PNs extend the local scope of PANs by

addressing virtual personal environments that spana variety of infrastructures (as well as ad hoc net-works). Even though we have described the PANview as addressing the problem of the communicationbetween you and the outer world, PN extends thePAN concept even further, as the POS can be dis-tributed all over the world. Figure 5 illustrates theconcept of personal networks. An important new ele-ment suggested by the figure is that the composition,organisation, and topology of a PN are determinedby its context. By this we mean that the geographicallocation of a person, the time of day, the electronicenvironment, and the explicit or implicit wishes touse particular services determine which devices andnetwork elements will be incorporated in a PN.

Figure 4 Frequency bands (MBS: Mobile Broadband System, UNII: Unlicensed National InformationInfrastructure)

Figure 3 PAN is for you, me, and the outer world. Extending your reach requires a multimedia gatewayas well as a distributed resource control with Quality of Service (Qos). (GSM: Global System for Mobilecommunication, UMTS: Universal Mobile Telecommunications System)

SP

SP

SPSP

SP

SP

SP

SPSP

SP

SA

SA

GSM

UMTS

WLAN

Frequency [GHz]

Europe

Japan

USA UNII*

ISM

UNII*

WLANMBS

WLANMBS

WLANMBS

2.402.4835

5.165,25

5.35.47

5.7255.825

10 16 17,1 17,3 59 64


3Telektronikk 1.2007

The present issue is composed of 12 contributionsthat cover the results of PACWOMAN (Power AwareCommunications for Wireless OptiMised personalArea Networks, URL: http://www.imec.be/pac-woman/), MAGNET (My personal Adaptive GlobalNET, URL: http://www.ist-magnet.org), andMAGNET Beyond (My personal Adaptive GlobalNET Beyond, URL: http://www.ist-magnet.org).

The personal network is the future for the wirelessand the mobile communications. In the view of the

author, future generation (FG) can be defined by thefollowing equation:

B3G + PN =∆ FG

where B3G stands for beyond third generation, whichis defined as the integration of existing systems tointerwork with each other and the new interface.

Cluster of localpersonal devices

Office cluster Car cluster

Home cluster

Gateway

Secure tunnel

PN

Interconnection structure(Internet, WLAN, GSM,UMTS, PSTN, ad hoc …)

Ramjee Prasad is a distinguished educator and researcher in the field of wireless information and multi-

media communications. Since June 1999, Dr. Prasad has been with Aalborg University, where currently he

is Director of Center for Teleinfrastruktur (CTIF), and holds the chair of wireless information and multimedia

communications. He is coordinator of European Commission Sixth Framework Integrated Project MAGNET

(My personal Adaptive Global NET) Beyond. He was involved in the European ACTS project FRAMES (Future

Radio Wideband Multiple Access Systems) as a Delft University of Technology project leader. He is a project

leader of several international, industrially funded projects. He has published over 500 technical papers,

contributed to several books, and has authored, co-authored and edited 20 books. He has served as a

member of the advisory and program committees of several IEEE international conferences. In addition,

Dr. Prasad is the coordinating editor and editor-in-chief of the Springer International Journal on Wireless

Personal Communications and a member of the editorial board of other international journals. Dr. Prasad

is also the founding chairman of the European Center of Excellence in Telecommunications, known as

HERMES and is now the Honorary Chair.

Dr. Prasad has received several international awards; the latest being the Telenor Nordic 2005 Research

Prize. He is a fellow of IEE, a fellow of IETE, a senior member of IEEE, a member of The Netherlands

Electronics and Radio Society (NERG), and a member of IDA (Engineering Society in Denmark).

Dr. Prasad is advisor to several multinational companies.

email: [email protected]

Figure 5 Conceptual illustration of Personal Networks



1 Introduction

During the last decades, we have seen the explosivedevelopment of wireless communication technolo-gies. Many technologies have been brought to ourdaily life and have been proven to be successful,e.g. GSM and WLAN. And many more technologies,such as UMTS and WIMAX are on the way to com-mercialization world-wide and hopefully to a success.At the same time, the networking technology forwireless communication is paving the way for a newparadigm, i.e. from the model of fixed–mobile to themodel of mobile–mobile. With such a model, manynetworking technologies are being developed, e.g.wireless mesh networks, mobile ad hoc networks,wireless sensor networks. With all these technologies,a wireless terminal is enabled to communicate withother wireless terminals directly without sending traf-fic through an intermediate node connected to thewired network.

From the success of GSM and WLAN, it is not diffi-cult to conclude that the service enabled by the tech-nologies for the end user is a key factor for the suc-cess of a technology. Following this logic, there hasbeen a strong consensus on the requirement for newtechnologies:

• Person Centered: Technologies should be centeredon the user, improving quality of life and adaptingto the individual. While the traditional communica-tion paradigm aims to establish the communicationlink between devices, the focus now shifts to thecommunication among the persons and services.

• Pervasive Service: The communication and com-puting technology will tend towards “invisibility”and “calmness” [1]. The offered services tend to bepervasive, causing minimum distraction to the userwith respect to their configuration and usage. Thecomputing environment is becoming smarter and

more responsive, with devices being able to estab-lish disposable, seamless connection to the requiredresource.

The concept of wireless personal area communicationis developed as an implementation of the personalcentered communication paradigm. IST-PAC-WOMAN1) (Power Aware Communications forWireless OptiMised personal Area Networks) is aresearch project in the fifth framework program ofthe European Commission that has been devoted tothis topic as a pioneer step to the future personalcommunication paradigm.

In this paper, we introduce the PACWOMAN vision2)

on the future WPAN. We start with a discussion ofthe emergence of the person centered concept and theWPAN paradigm, followed by the design objectiveand technical challenges foreseen by the PAC-WOMAN consortium. We then discuss the role ofnetwork and service provider in the WPAN para-digm, and end the paper with some concludingremarks.

2 Emergence of Wireless Personal

Area Network

Besides the explosive development of communicationand networking technologies in the last decade, theperson centered communication concept has beenevolving. With the person centered concept, thefuture communication paradigm is believed to movefrom communication between devices to communica-tion between people. With such a concept, the under-lying communication and networking technologiestend to be invisible and transparent to the user sothat the requirement on the user’s technological back-ground and the distraction to the user can be mini-mized.

Wireless Personal Area Networks

– The PACWOMAN Vision

Y A O D A L I U

Yaoda Liu is a

PhD candidate

at Aalborg Uni-

versity, Denmark

Together with the advance of wireless communication technology, the person centered network

concept has been evolving to the concept of Wireless Personal Area Network (WPAN) in the last

decade. WPAN is foreseen to bring new services to the user and improve our daily life. For researcher

and network operator, a promising area has been opened up. In this paper, we introduce our vision

on future Wireless Personal Area Network developed in a pioneer project in this field.

1) The PACWOMAN consortium consists of IMEC (Belgium), CPK/AAU (Denmark), CSEM (Switzerland), Lund University (Sweden),MILTECH (Greece), MOTOROLA (UK), ICCS/NTUA (Greece), Universtiy of Cantabria (Spain).

2) Most of the material in this paper is derived from the PACWOMAN project.



The addresses of sources/destinations in communica-tion links are determined either by the person thatowns the device, the service they are capable to offer,or the resource’s contents. This causes radicalchanges in the design, for example, in addressing(content-based or capability based), security etc. Asa consequence, new research topics are emerging,addressing different aspects of this problem. Someexamples of new exciting research fields are dis-cussed in the following, although we do not aim atproviding an exhaustive list of them. The first exam-ples are service portability and virtual home environ-ments [2], concepts aiming at providing users withthe same service experience independently of the userinterface, terminal capabilities, access network tech-nologies, network providers, and service providers.Another important and related emerging area is per-vasive computing targeting environments where net-worked computing devices are ubiquitous and evenintegrated with the human user [3].

Due to the increasing demand of connected anywhere,the wireless communication technology has beenplaying a more and more important role in the personcentered communication paradigm. The paradigmshift mentioned above implies different approachesto the development of wireless communications.As concluded by the Wireless World Research Forum(WWRF) [4], a purely technical vision for the wire-less development is not enough. In other words, theinvestigation of, for example, new network technolo-gies or radio interfaces will not be sufficient to cometo grips with the future. Rather, such a technical viewmust be broadened or complemented by:

• Person-centered approach, looking at new waysusers will interact with the wireless systems;

• New services and applications that become possi-ble with the new technologies;

• New business models that may prevail in thefuture, overcoming the by now traditional user,server, provider hierarchy.

There is an essential difference in thinking about the4th generation (4G) wireless systems compared to theway 3G and other present wireless standards are pro-duced. While the latter standards have been put in atechnology-driven development process, early 4Gphilosophy is being approached from an applicationviewpoint, with an implied assumption that technol-ogy will follow to enable the realization of the appli-cation vision [5]. The essence is to provide a ubiqui-tous networking capability in which questions of data

speeds are rendered irrelevant by the universal avail-ability of more bandwidth than the vast majority ofusers would ever need.

The 4G wireless communications will tend towardspersonal [6]. The user will no longer be “owned” byany operator: the users, or their trusted agents, willselect at each instant the best system available that iscapable of providing the required service and perfor-mance. The selection will be made according to theuser’s profile, the type of data stream and the trafficload on the available networks.

WPAN comes into play as an implementation of thepersonal centered communication paradigm. Techni-cally a WPAN is a networked collection of devices inthe geographic vicinity of a person. This collection ofdevices forms a wireless “bubble” around the person,referred to as Personal Operating Space (POS).Besides the connection among the personal deviceswithin a WPAN, the WPAN should also provide theuser and the devices with a seamless, ad hoc connec-tion to the world out of the POS. The organization ofWPAN is expected to be transparent to the user, butprovide the user with much better experience of ser-vice. One example3) scenario could be a user at homehaving a video conversation with the customer; whenthe user moves to a room with a big screen (e.g. lap-top, or LCD TV), the big screen can join the user’sPOS, and according to the user preference specifiedbeforehand and current situation of the room (e.g.anyone else in the room), the video conversation maybe moved to the big screen from the phone screen.

The present notion of WPAN came about as an accre-tion of several developments and tendencies. Someof them were strongly interrelated from the verybeginning; nevertheless, all tendencies now tend tobe merged into a unique conception. These factors ledto the emergence of the PAN, which traced its inde-pendent evolutionary line afterward, defining ownapplication scenarios and motivating the appearanceof new applications and services.

• Bridging different wireless standards. Today, weare surrounded by a diverse set of wireless accesstechnologies applied in wide-area cellular networks(GSM, IS-95, IMT-2000), personal communicationsystems, and wireless local area networks (802.11,HIPERLAN). Most of these systems, however, arestill tailored towards a narrow and specific applica-tion scenario. Hence, there is a need for a singleuniversal wireless communication system thatoffers a user-friendly and efficient way to accessinformation with a variety of devices such as

3) The example is inspired by a demonstration of NICT, Japan, in the CTIF-Kyoto joint workshop, Aalborg, Sept, 2006.



mobile PCs, mobile phones, PDAs, pagers, anddigital cameras. Such wireless solutions wouldbring together all these technologies applied indifferent sectors and at the same time provide auniversal and ubiquitous connectivity solutionbetween computing and communication devices.

• Very high wireless data rates. The user’s needfor bandwidth is increasing continuously. In fact,the need for higher data speeds had driven the evo-lution of 2G wireless systems to the 3G UMTS.Further increasing demand of data rates beyondUMTS requires usage of pico-cells. The low-power, picocellular nature of WPANs implies highspatial capacity, i.e. it enables a more efficient spa-tial reuse of the radio spectrum. The short-rangewireless networks, such as WPANs and WLANscan support significantly higher data rates than theones offered by the 3G wireless systems. Figure 0.1depicts the mobility vs. data rate graph for theexisting and future wireless technologies.

• Cable replacement. Here we refer to the initiativesfor developing a cable replacement technology or“last meters” technology instantiated through thespecifications of IrDA, HomeRF and Bluetoothworking groups. Each of these technologies sur-passed their initial targets, offering far more flexi-bility to the electronic devices than the mere cablereplacement.

• Ergonomic settlement of personal electronicdevices. This is in close relation with the cablereplacement. The possibility of wireless intercon-nection of proximal devices motivates investigationof new computing structures, directed towards thecalm technology [8]. For example, the PDA’s key-board can be a control interface to all other per-sonal devices.

• Ubiquity of Internet access. The number of accesspoints to the wired Internet has grown significantly.People have a need for Internet access everywhere:at homes, enterprises, public spaces. The WPANwill equip the individual with a “wearable” Internetaccess.

• Cheaper hardware. The shrinking semiconductorcost, as well as the lower power consumption forsignal processing, make it feasible to build/upgradepersonal computing devices with wireless commu-nication capability.

3 Design Objective and Technical

Challenges

In this section we present the PACWOMAN vision onthe development of future WPAN communicationsystems. The main design objectives of WPAN tech-nology foreseen by the PACWOMAN consortium are:

• Low power consumption: The low power consump-tion is a critical issue since the rate at which batteryperformance has been improved is fairly slow com-pared to the explosive overall growth in wirelesscommunications. Therefore, the wireless protocolitself should employ economic usage of the batteryenergy.

• Operation in the unlicensed spectrum: The WPANsystems use license-free wireless links, because itis the only way to achieve ubiquitous connectivitywithout adverse impact to an existing wirelessinfrastructure.

• Low cost and small package size: The low cost,small size single-chip solution is the economic andergonomic conditions for widespread use of theWPAN technology.

Figure 1 The settlement of the existing and future wireless technologies

20 155

Indoor

Pedestrian

High speedvehicular

rural

Mobility & range

Personal area

Vehicularurban

IEEE 802.11a/b(WLAN),

Hiperlan2,MMAC

0.5 2

UMTS

DECT

Fixed urban

Total data rate per cell

10

BRAN

broadband WPANWPANBluetooth

400 Mb/s

GSM



• User friendly operation: For widespread use of theWPAN technology, user friendly operation is an-other ergonomic condition. From a technologicalpoint of view, the solution should provide seamlessconnectivity and services to the user in an auto-configured manner.

• Context awareness and adaptability: To providethe user with seamless connectivity and services,the understanding of surrounding environments(context) the capability of adaptation of the under-lying technologies utilizing the awareness to thecontext are very important.

A user-centric network architecture has been sug-gested by the PACWOMAN consortium as illustratedin Figure 2, which contains a 3-level hierarchy,namely PAN, CAN and MAN. A user is surroundedby various devices moving together with the user ortemporally around, potentially with different tech-nologies and capabilities. Despite heterogeneity intechnology and capability, all these devices are con-nected and form the WPAN in the following manner.Basic terminals (BTs) with low capability (computa-tional, battery, communication) are attached to someadvanced terminal (AT), forming a virtual device(VD). Multiple users can form a Community AreaNetwork (CAN), either with ad hoc connectivity orwith infrastructural connectivity.

Based on the above network architecture, the maincharacteristics and challenges of future WPAN com-munication systems have been derived as follows.

Heterogeneity in devices

WPAN devices can be categorized taking into con-sideration the applications for which they will betargeted. Roughly, we can distinguish between LowData Rate (LDR) devices, in which binary transmis-sion speeds are usually below tens of kilobits per sec-ond and Medium/High Data Rate (M/HDR) devices,characterized by capacities of up to tens of megabitsper second. The former group will basically comprisesensors and actuators, whereas high capable devices,generally known as Advanced Terminals (AT) withinthe PACWOMAN nomenclature, such as PDAs andlaptops are illustrative examples of the second group.Interoperability between devices belonging to the twodifferent groups is a key issue, as not many solutionshave been proposed to overcome this problem. Tradi-tionally, IP has served as a global interconnectiontechnology, but it is more likely that the LDRdevices, due to their inherent characteristics, will notbe IP capable, so a different approach must be taken.Within the PACWOMAN project, a hierarchicalapproach has been followed. The proposed schemeassumes that a single person might be wearing a num-ber of LDR devices, known as BT within the PAC-WOMAN architecture, which will be able to commu-

a T

a TV D

a T

a T

a T

G WM

b T

b T

V D

bT = basic Termnal

aT = advanced Terminal

M = Master terminal

GW = Gateway

VD = Virtual Device

PAN = Personal Area Network

CAN = Community Area Network

Backbone

VD

PAN PAN

CAN

Figure 2 PACWOMAN network architecture



nicate by means of a proprietary protocol with anM/HDR terminal, being characterized by having adual protocol stack as shown in Figure 3. This termi-nal will act as a Master for the BTs and will be actingas a manager of communication between all BTsbelonging to the same user, establishing a traditionalstar topology and forming what has been called theVirtual Device within the PACWOMAN project. TheMaster node will also act as a “gateway” for the com-munication between all BTs and entities out of thevirtual device.

In this way, the PACWOMAN architecture bringsabout a novel concept where Layer 2 mechanisms areused to cluster low-capability, low-power and low-cost devices, while at Layer 3 traditional routing tech-niques are used, apart from their legacy role, to allowthe rest of the devices within the PACWOMAN archi-tecture to access the information provided by the BTs.

Heterogeneity in terms of bit rates and capabilitiesof the devices that will be part of the architecture isa fundamental feature to be tackled. The coexistenceand interoperability of heterogeneous technologiesare mandatory steps towards the achievement of thenext wireless communication user centered paradigmand, more specifically, of the WPAN concept.

Ad hoc routing support

At the moment, commercial wireless communicationsystems rely on an adjacent infrastructure, and mostof the time a communication comprises just one wire-less hop; that is, from the mobile terminal itself to thefirst point of attachment to the network (a base stationin the case of cellular communications or an accesspoint with WLAN architectures). However, it is fore-seen that the terminal to terminal communication willplay a key role in the future wireless communicationsystem. The major advantage of terminal to terminalcommunication is the potential of co-operation

between terminals, which may enhance the spectrumefficiency as well as the network coverage. The mainrequirement for such architectures can be summa-rized as follows:

• Willingness of intermediate nodes to relay informa-tion for other nodes as the source and the destina-tion may not be on each other’s physical vicinity;

• Capability of self-organizing and self-configuringin a distributed manner due to the lack of a centralmanagement entity.

IETF MANET working group is devoted to the pro-vision of ad hoc multi-hop connectivity. In the lastdecade, tremendous efforts from both academia andindustry have been spent in this filed.

Compensation for wireless link impairments

Wireless links are exposed to constraints such as highbit error rate and limited throughput. These character-istics are due to the intrinsic limitation of the radiochannel. Although to some extent those are compen-sated for by the link layer techniques that are in-cluded within the different technologies (channelcoding, medium access control and error control), thebehavior exhibited whenever IP traffic is layered overthese wireless technologies differs from being accept-able, with the obvious result of performance degrada-tion (decreasing throughput and/or increasing latency).Hence, complementary machinery is needed to com-pensate for wireless link impairments; this comple-mentary machinery has been approached from twodifferentiated points of view; in terms of modifyingthe proper higher protocols so as to adapt them to thecharacteristics of wireless channels, or by proposingintermediate layers that hide the wireless impairmentsto the upper layers, which do not need to be furthermodified.

Technology independency

The PACWOMAN architecture design has been doneto be platform-independent. A large number of wire-less access technologies are envisaged to co-exist infuture wireless communication spaces, so the neces-sary methods for them to inter-work seamlessly haveto be deployed. In this sense, the corresponding wire-less network driver(s) and link layer protocol(s)should be accessed from upper layer protocols andapplications for control purposes, in a generic man-ner, independent of the type of technology that isbeing used (in the same way upper layer protocolsand applications access the underlying protocol stackthrough the socket interface for data purposes).

Thus, a common interface is required for both wire-less drivers and lower layer protocols to be uniformly

Figure 3 Basic terminal and Master protocol stacks

MAC/PHY

LDR

Application

MAC/PHY

LDR

Application

MAC/PHY

LDR

Application

MAC/PHY

M/HDR

MAC/PHY

LDR

PCOL

IPv6

TCP/UDP

Application

MAC/PHY

M/HDR

MAC/PHY

LDR

PCOL

IPv6

TCP/UDP

Application

MAC/PHY

M/HDR

MAC/PHY

LDR

PCOL

IPv6

TCP/UDP

Application

bT = Low DataRate (LDR)

M = Dual Protocol Stock



accessed by upper layer protocols. Such a commoninterface, that resembles the traditional API, shouldsupport the necessary service primitives for (1) con-figuration of wireless drivers and link layer protocols,(2) retrieval of statistics, and (3) event handling.

Service and applications deployment

support

Unlike in the fixed Internet, services and applicationsin the Mobile Internet will have to take into accountthe specific characteristics of different mobile envi-ronments. This can be achieved by constructing a ser-vice/application framework from a set of generic ser-vice elements in the middleware. In recent years,terms like pervasive and ubiquitous computing aregaining a lot of relevance and the middleware con-cept, based on a distributed software infrastructure,appears as a good choice to fulfill their requirements.Future services and applications will need to be para-meter/context aware, which concerns the user profile,user location, network context, underlying technol-ogy, and so on, so as to adapt their behavior accord-ingly.

Power consumption

WPAN devices such as laptops, palmtops and PDAsexhibit an upper bound on the operation time due topower (battery) constraints. Power consumption isdirectly related to the processing tasks that a deviceruns, as well as the communication tasks. From thenetworking perspective, the communication task andrelated processing tasks are of most interest. With thecommercialization of WPAN, the communicationtasks are expected to be grow explosively, which con-sequently leads to faster battery exhaustion. Mitiga-tion techniques should be enforced to reduce powerconsumption as much as possible without sacrificingthe overall networking performance. This optimiza-tion task involves almost all layers in the OSI proto-col stack. Potential optimization techniques withrespect to power consumptions are

• Power control at physical layer: By tuning thetransmission power to a lower but yet high enoughlevel, power consumption is expected to be mini-mized. As a side effect, interference is alsoexpected to be lower.

• Error control at link layer: Adding error correctioncodes adaptively according to channel conditioncould potentially minimize the overall transmitteddata volume, consequently minimizing the powerconsumption.

• Medium Access Control protocols at link layer:It plays an important role in reducing the powerconsumption due to collisions.

• Routing protocols at network layer:- Routing protocols taking into account other met-

rics than hop count (e.g. channel condition),together with the above techniques should beable to achieve lower power consumption.

- Efficient algorithms for routing information collec-tion (e.g. link state updates) are foreseen to reducethe power consumption for routing purposes.

4 PAN and CAN Optimization Layer

In this section, we present some technical resultsbrought by the networking package of PACWOMAN.We start with the introduction of the PAN and CANOptimisation Layer (PCOL), followed by some adap-tation schemes built on top of the PCOL layer.

PCOL is an enabling technology designed specifi-cally for WPAN communication systems as shown inFigure 4. In this sense, data packets will traverse a setof protocol boosters that aim at optimizing the net-work and communication performance managed bythe PCOL, while the control plane will embrace anumber of different components that will cope withsome management task.

PCOL data plane

PCOL will accommodate a diverse set of protocolboosters, i.e. link layer targeted to packet processingduring transmission/reception to/from the wirelessdriver, so as to enhance performance of particulartypes of traffic.

Each of the different boosters will be applied todownstream traffic (meaning traffic going out ofa particular host), according to both its particularrequirements and the varying conditions (channelquality, remaining energy level, etc.). In this sense,the PCOL will adapt its operation in order to opti-mize communication performances while maintaininga power aware activity.

PCOL control plane

PCOL control plane contains all the necessary ma-chineries to fulfill the PACWOMAN requirements.

Layer 2 service discovery

Ease of use and auto-configuration, as well as servicediscovery are mandatory topics to be tackled withinthe WPAN communication system. The Layer 2 Ser-vice Discovery module will be in charge of the auto-matic selection of a Master, from all the possible can-didates, within a single PAN. In addition, it will per-form the corresponding actions so as to allow the restof PACWOMAN devices to reach the informationprovided by the BTs.



Power manager

Battery awareness is one of the major requirements ofa WPAN communication system. The power managermodule will be in charge of optimizing energy con-sumption within an M/HDR device through utilizingthe awareness towards the environmental conditions(i.e. link conditions, remaining battery level, etc).We have seen that an energy-aware solution is morelikely to be achieved targeting other communicationlayers, specially MAC and physical levels, but takingadvantage from the control possibilities that are pro-vided by subjacent technologies, some reductions canbe achieved.

PAN&CAN membership

PAN&CAN membership module is the key compo-nents for security purposes. In WPAN, security mustbe assured on all communication links, and it is there-fore a mandatory requirement that the membershipneed to be implemented.

Ad hoc routing enhancement

The main goal of this module is to take the advantageof the information provided by the PCOL to adaptad hoc routing protocols. Apart from being efficientfrom a performance point of view, it might helpleveraging a power aware behavior using batteryinformation and reducing the number of retransmis-sions. Figure 5 shows an illustrative example of howa route towards a destination could be selected usingthe SNR as a new metric, achieving a better commu-nication quality (both in terms of performance im-provement and error decreasing) while saving batteryon the sender, which would need to perform a consid-erably smaller number of MAC retransmissions.

Uniform Wireless Interface

The UWI provides a uniform set of manipulationfunctions that hide wireless driver singularities to thePCOL.

Figure 4 PCOL high-level protocol architecture

Data plane Control plane

Applications

TCP / UDP

Wireless driver

IPv6

Protocolboosters

Uniform wireless interface

PCOL Powermanager

PAN & CANmembership

Ad hocenhancement

L2 servicediscovery

Middleware



5 Conclusion

Wireless Personal Area Network, as an implementa-tion of the person centered networking concept, willplay an important role in the next generation of wire-less communication systems. The services brought bythis WPAN concept are bringing new opportunities,but at the same time challenges.

In this paper, we have shown that with the currenttechnology WPAN can already be realized as wehave done in the PACWOMAN project. However,for the real commercialization, there are still a num-ber of issues to address; service and applications,naming and addressing, throughput enhancement,efficient energy utilization, and security. To copewith those challenges, the PACWOMAN consortiumhas developed an enabling framework, i.e. PAN andCAN Optimization Layer.

Besides the technical challenges, business model andnew services are two important issues. In the businessmodel, the question of how the service and networkproviders are involved in the WPAN communicationsystem needs to be answered. And the WPAN para-digm can never be commercialized and successfulwithout services of interest to the end user at anaffordable cost.

Acknowledgement

The author would like to thank the PACWOMANconsortium, especially the networking package(WP5) members for a fruitful project.

Reference

1 Weiser, M, Seely Brown, J. Designing CalmTechnology. PowerGrid Journal, 1.01, July 1996.(http://powergrid.electriciti.com)

2 Daoud, F, Mohan, S. Service Portability and Vir-tual Home Environments. Guest editorial. IEEECommunications, 40 (1), 76–77, 2002.

3 Gupta, S K, Lee, W-C, Purakayastha, A, Srimani,P K. An Overview of Pervasive Computing.Guest editorial. IEEE Personal Communications,8 (4), 8–9, 2001.

4 Wireless World Research Forum. 2006, October27 [online] – URL: http://www.wireless-world-research.org.

5 Richardson, P. Personal to Global: WirelessTechnologies, 2005 – 2010. Gartner Group Inc.,Research Brief, February 23, 2001.

6 Pereira, J M. Fourth Generation: now, it is Per-sonal. Proc. PIMRC 2000, September 2000,1009–1016.

7 Rabaey, J. PicoRadio Networks: An Overview.Berkley Wireless Research Center Focus 2000Session, July 2000.

8 Weiser, M, Seely Brown, J. Designing CalmTechnology. PowerGrid Journal, 1.01, 1996.(http://powergrid.electriciti.com)

Yaoda Liu received his B.Eng from Shanghai Jiaotong University, China, in 2000, and his M.Eng from

National University of Singapore in 2003 with focus on Mobile Ad-hoc Networks. In October 2003 he joined

Aalborg University, Denmark as a PhD candidate. Since then he has been working in the IST PACWOMAN,

MAGNET, and HIDENETS projects. His research interests include algorithm and protocol design for wireless

multi-hop networks.


Figure 5 Example of a multi-parametric route election

SNR?

SNR?

SNR?Legacy Route

Modified route(using SNR information)



Assessment of the 3G Success

Until the beginning of 2006 very few outside South-East Asia would defend the idea of a success for 3G– and even here the profitability of the new servicewas questionable and the use dominated by ‘exotic’services such as download of ring tones. This gloomypicture was supplemented by the continued success of2G, especially GSM. By mid-2006 there were 2 bil-lion GSM users serviced by 784 networks in 209countries/territories [2]. Compared to this, the 3G sta-tus does not seem very impressive. WCDMA – whichhas now emerged as the dominating 3G standard –has 70 mill. subscribers served by 122 networks in55 countries. The development during the first sixmonths of 2006 has however opened for some opti-mism as the figures are the result of a 45 % growth insubscribers during this period. This again is arguablythe result of a combined development in technology/devices and services. A number of HSDPA (the firstevolution of WCDMA) products has been launched;many also supporting GSM/EDGE, thus ensuring ser-vice continuity. This has enabled especially enhancedInternet/data communication and mobile TV. Thisagain has resulted in the first profitable 3G opera-tions, but real take-off approaching, e.g. GSM num-bers, requires devices and services demanded by areally large number of users.

Introduction to PN Applications and

Services

Personal Networks, PNs, is an essential concept fordeveloping mobile applications, services and net-works in the future. The idea behind the concept isto bind all available networks together giving globalconnection and supporting users’ technology usageboth in their close vicinity (referred to the PersonalArea Network, PAN) and in the more remote or dis-tributed network islands containing work environ-ment, location of friends, family members and otherpersonal contacts.

A PN can be considered as an ordinary PAN, butwithout geographical limitations, see Figure 1. In aPAN all the devices are within a certain distance –say up to 10 metres. This gives certain suggestions ofconnection technologies used in the PAN (Bluetooth,IR, etc.). In a PN, devices can be separated by hun-dreds of kilometres and still belong to the same vir-tual PN. It means that in a PN, connection technolo-gies are not only short range like Bluetooth orWLAN, but also medium range and national or con-tinental range, like GPRS or UMTS. This calls formore service requirements for the PN as comparedto the PAN. Those are especially different when wetalk about technical requirements, because someissues in a PN are much more complex to achievethan in PAN. Seamless service, mobility, single sign-on, context discovery, self-organization, roaming,handover, context transfer and session continuity arethe requirements that will be the main technologicalproblems for PN. If the issues in PN could be handledwith the same quality as in a PAN, then there wouldnot be much difference from the service point of viewbetween the PN and PAN concepts. The physical net-work structure would be different but the logicalstructure still remains the same [3].

A key feature of the PN is that the PN emphasizes thetrust relationship between the user and the devices.The concept of PN can thus bring a solution fortrusted communication between the many local andremote personal devices in view of the support of avariety of personalized and context-aware services.A Personal Network is a protected secure person cen-tric network that connects all the nodes of a personover ad hoc as well as infrastructure networks andthat provides context-aware services and applications.As such, it is a dynamic collection of interconnectedheterogeneous active personal devices, not only thelocal devices centered around the person, but alsopersonal devices on remote locations such as devicesin the home network, the office network and the carnetwork.

Personal Networks as Business Strategy for the Wireless

Communication Future

K N U D E R I K S K O U B Y , K A R S T E N V A N D R U P

Professor Knud

Erik Skouby is

Director of CICT

at the Danish

Technical Uni-

versity, Lyngby

Karsten Vandrup

is Senior

Research

Manager in

Nokia Technol-

ogy Platforms,

Denmark

In the wake of the ongoing 3G rollout in Europe – as well as in other parts of the world – the research

on what will be the next G in mobile communication has taken off even more rapidly than the research

on a 3rd generation system did start. However – many top researchers see difficulties in just letting

the technology race set the standards for the future; user and market issues must be taken into

account already in the early research. This paper takes its point of departure in the quote “3G +

Personal Networks = 4G” [1], and assesses the theories of Personal Networks in the context of future

business strategies.



User Requirements – the Soft

Values are Becoming Harder

It is a challenge to identify and explicitly formulateneeds and user requirements. Users themselves havedifficulties in explicitly expressing their preferencesand needs, in particular when referring to situationsincluding future, not yet developed technologies. Andif user requirements are expressed, these may changein time with the trends, new technological develop-ments, traditions and situations in which users areactive.

One tradition has found the solution in associatingthe notion of user requirements with methods tryingto identify ‘pure user needs for technology’, i.e.attempts to identify what the user really needs set upas a contrast to what the market ‘forces’ the user tobuy. This has not been very helpful in the process ofchoosing between different developments of technol-ogy as it tends to focus only on the user and not onthe specific requirements.

To discuss choices of technology a shift is needed inthe focus to analyses of how specific technologiescan serve different users in specific social settings.

What is needed is a methodology expressing individ-ual user perceptions of specific needs as requirementspresent in given social settings served by, e.g. thetechnology and components of a future PN architec-ture.

Based on well-established methodologies, e.g. partic-ipatory design, a template for discussing user require-ments in relation to future technologies has beendeveloped in the MAGNET project involving sce-nario construction. Central for the MAGNET sce-nario approach is that it mixes the futuristic scenarioconstruction with participatory design principles.This makes the template relevant in a discussion ofinnovative user-centred situations, needs and require-ments, which will challenge existing networks andtechnologies and call for the new solutions.

Further, User requirements in this setting have beenspecified as the result of combining user needs, socio-economic trends and business models into realized– or in this context rather expected realized demand.Business models and socio-economic trends formrequirements and analyses at a different level, sincethe prospects of making an ICT business work in

Figure 1 A Personal Network and its PANs, interconnection structure, etc. MAGNET Beyond 2006

PANPersonal PAN

Local

foreign

devices

Smart building

Home cluster

Remote personal devices

Remote foreign devices


Corporate cluster


Vehicular clusterInterconnecting structure

Internet, UMTS, WLAN

Ad Hoc, etc



relation to various operators and suppliers are ana-lysed. The business model activities, on the otherhand, are built on the user requirements and results ofthe user case studies and user involvements. This sumof two rather complicated analyses is conceptuallyuseful in opening for analyses of the elements in tech-nical solutions that are potentially important for com-mercial success [3], [4].

Several very promising technologies has over theyears failed on the market, due to the lack of useracceptance, and by disqualifying themselves lackingthe simplicity and readiness that are required if largeramounts of customers with various backgrounds andexperiences are the main target group.

The dominant approach to user requirements in thetraditional telecom industry is that services and appli-cations are shaped by the combined influence fromterminals and networks developed according to thecurrent technological possibilities, i.e., user require-ments are not taken into account during initial con-ception. Service development then involves the PANand a combination of networks; PSTN, cellular net-works, digital broadcasting networks, as well as Blue-tooth, WLAN, the Internet, etc., and the combinationof these heterogeneous networks.

In MAGNET, the approach to user requirements isdifferent. The methodology to describe and developan understanding for implementation of an efficientPN-solution in a heterogeneous, multi-modal environ-ment involves ‘technology’, ‘user needs’ and ‘eco-nomics’. A key element of ‘user needs’ is perceivedto be quality of service – or quality of experience –associated with given private and/or business activi-ties and its relation to the underlying technologies.

The introduction of PN services along with the asso-ciated technologies will constitute a major paradigmshift. There are currently no business models or sce-narios in place for PNs; however, an enhanced under-standing and knowledge of possible business modelsolutions as well as market and socio-economicaspects are necessary in order to achieve the fullbenefits of a heterogeneous communication modelas proposed in the PN concept [3], [4].

Business Strategies for SMEs Based

on PNs

In retrospect, providing systems to the current andpast generations of mobile communication systemshas been a game of the world’s large enterprises,willing to invest billions of dollars, euros and yen inproduct development in order to be present on themarket during the first hype. Later, when the tech-

nologies become commodities, the business toengage in the system and device side is significantlyless attractive. Obviously, as the mobile communica-tion business on the application side is moving intothe area of more or less open operating systems, somenew businesses have been created providing servicesand applications to the end users. In the large picture,the revenue of these businesses still only counts for afraction of the entire business.

Moving into systems like “Personal Networks”, theinfrastructure is a combination of various existingcommunication technologies, complementary work-ing together to provide the best possible serviceand/or price for the end user, and combine it withstate-of-the-art PAN solutions. This opens for a vari-ety of opportunities in the technology area, in the sys-tem area and not least in the business area.

As the businesses and services over the years havebecome a set of more and more complex arrange-ments with more technologies, systems and servicesoffering businesses that interact closely, the tradi-tional mobile and wireless value chain is fairly oftendescribed as a value complex.

The value complex concept as an approach to charac-terize “Personal Networks” can be considered as acomposition of the traditional one-dimensional valuechains, adding up a number of smaller value chains toconstruct a system. On the business and technology-provisioning side, this potentially opens for a newand strategic role for smaller and medium sized enter-prises. In a personal network future, SMEs can focustheir activities on one or more of these smaller valuechains; build R&D, production and marketing tosupport this limited aspect of the value complex.The agility of the SMEs is the key argument for theircentral role in the new PN-based communicationstructure.

Why SMEs? Several surveys and reports publishedin the EU stress the point that the success of Europein a globalized world very much depends on the suc-cess of its SMEs [5]. Especially in the northern Euro-peans states – the Nordic and Baltic countries – thebusiness environment has a large percentage of SMEscompared with other regions of the world. Andindeed successful SMEs, competing on global termsin global markets.

Some of the general challenges the traditional Euro-pean SME is up against are:

• Lack of awareness of new market opportunities;



• Lack of knowledge, expertise and financialresources to carry out in-depth research in order toappropriately assess the current and potential mar-ket situation with regard to products and servicesin international markets;

• Lack of knowledge, technical skills and financingto effectively carry out R&D, marketing and pro-motional campaigns;

• Lack of specialized technical skills, financialresources and contacts with buyers/technologyholders in foreign markets needed for strengtheningthe supply base at an accelerated pace [6], [7].

Assessments on each of these points will show thatwithin the segment of mobile and wireless communi-cation, SMEs will have better opportunities with adecentralized PN structure than with the traditionaltelecom set-up:

• By gathering together in PN developer’s commu-nities, the access to new markets and partnershipswill be more visible.

• In-depth research has been carried out – e.g. bythe European Commission IST research projectsMAGNET and MAGNET Beyond, and as a resultlarge parts of the technologies, market analyses,user requirements are documented thoroughly.

• The theories of Personal Networks are developedin global projects and communities, engagingresearcher from Japan, China, India, USA andEurope, making PN competences present in mostparts of the world.

However, the fact that the personal network sub-sys-tems, applications and services can be approachedindividually, each with a complete value chain andnew open markets in the context, is probably thestrongest incentive for SME’s to enter the worldof PNs. In this type of decentralized systems, therequired and focused R&D effort is manageable forSMEs – and the big players may regard them as non-threatening, but useful partners.

Conclusion

As mobile communication systems beyond 3G willbe a more diverse setup, consisting of a variety ofheterogeneous systems, working together in more orless transparent ways, there will be room for smallerand medium size players in the various markets ofdevices, applications and services.

To summarise, stepping up to the next generation ofpersonal communication systems – e.g. by introduc-ing Personal Networks – huge changes will take placein the business and market places globally.

Acknowledgement

MAGNET Beyond is a continuation of the MAGNETproject (www.ist-magnet.org). MAGNET Beyond isa worldwide R&D project within Mobile and Wire-less Systems and Platforms Beyond 3G. MAGNETBeyond will introduce new technologies, systems,and applications that are at the same time user-centricand secure. MAGNET Beyond will develop user-cen-tric business model concepts for secure Personal Net-works in multi-network, multi-device, and multi-userenvironments. MAGNET Beyond has 30 partnersfrom 15 countries, among these highly influentialIndustrial Partners, Universities, Research Centres,and SMEs.

References

1 Prasad, R, Deneire, L. From WPAN to PersonalNetworks. Artech House, 2006.

2 GSA – The Global mobile Suppliers Association.2006, October 27 [online] – URL:www.gsacom.com

3 Prasad, R, Skouby, K E. Personal Network (PN)Applications. In: Wireless Personal Communica-tion, 33 (3-4), 227–242, 2005.

4 Prasad, R, Farseruto, J, Vandrup, K. MAGNETpaving a path towards the future wireless commu-nication. ECWT, Paris, 2005.

5 European Commission. Innovation and Researchin Small and Medium Enterprises. Brussels, 1996.European Report on Science and Technology.

6 Hibbert, E. The Globalisation of markets – howcan SME’s compete? Middlesex University Busi-ness School, 2000.

7 International Trade Centre UNCTAD/GATT.Product and Market Development for Export.Geneva, 1995.



Knud Erik Skouby is professor and founding Director of the Center for Information and Communication

Technologies (CICT) – a center providing a focal point for multi-disciplinary research and training in

applications of ICT at the Danish Technical University. His main area of research interests includes the

techno-economics and regulation of the telecom sector and of new telecom applications and services.

He has participated as project manager and partner in a number of international, European and Danish

research projects. He has served on a number of public committees within the areas of telecom, IT and

broadcasting; as a member of boards of professional societies; as a member of organizing boards,

evaluation committees and as invited speaker on international conferences; published a number of Danish

and international articles, books and conference proceedings in the areas of telecommunications regulation,

technology assessment (information technology and telecommunications), demand forecasting and political

economy.


Karsten Vandrup is Senior Research Manager in Nokia Technology Platforms, and serves as Technical

Manager and Deputy Coordinator of the EU FP6 Integrated Project MAGNET Beyond. Prior to this position,

Vandrup has had several positions within the Nokia Corporation in Copenhagen, Denmark, and in Espoo,

Finland. He holds a degree in Telecommunications and Electronics from the Technical University of

Denmark, supplemented by studies at INSEAD, UCLA and the Swedish School of Economics and Business

Administration in Helsinki, Finland.

Besides his work in Nokia, Vandrup is chairman of the Technology Foresight Council – Mobile and Wireless

Technologies under the Danish Ministry of Science, Technology and Innovation, and holds seats on a

number of boards, advisory boards and steering committees, both national and international.



1 Introduction

The main emphasis of this report is on business mod-els and business strategies of the present day mobileoperator in the Personal Network (PN) environment.The main objectives are the identification of theimpacts of PNs over existing infrastructure-based net-works and the development of new business modelsand business strategies within the context of MAG-NET (My personal Adaptive Global NET) leading toprofitable endeavours for the mobile operator. Theaspects considered are related to usage situations,economics, markets and sociological factors1) [1].

The work primarily builds on theoretical approachesto business modelling as well as input from techno-logical research in the PN area. The potential usabil-ity of the work will be for the architectural work onPNs and, consequently, for the work on standardiza-tion.

We conceptualize PNs in the following manner: ‘PNsare configured with functionality supporting securetunnels, as the opportunity and demand arise to sup-port personal applications. PNs consist of communi-cating clusters of personal and general digital devicesshared with others and connected through varioussuitable interconnection agreements2) [1]. PNs thuscomprise potentially of ‘all of a person’s devicescapable of network connection whether in his or herwireless vicinity, at home or in the office3) [1]. ThePN concept is closely related to, for instance, the Vir-tual Home Environment concept promoted in 3GPPand other similar concepts related to the use of het-erogeneous networks for delivering personalized ser-

vices to end-users4). However, the specificity of theMAGNET project on PNs is the focus on PersonalArea Networks (PANs), P-PANs (Private PANS) andpeer-to-peer organized networks. The implication isthat there is an emphasis on the self-organizedaspects of networks and applications.

Standardisation bodies like ETSI often refer to con-vergence as the convergence of fixed and mobile net-works. From this reference point ETSI could foreseeconvergence in other telecommunication domains,informatics, broadcasting and entertainment. Con-verged mobile services are predicted to change thetelecoms world.

Many mobile and fixed telecommunication operatorslike Telenor, TeliaSonera and Tele Danmark are fac-ing declining voice revenues. Today they show animmense interest in Internet protocol (IP) servicessuch as voice-over-IP (VoIP) and rich multimediaservices.

Convergence allows Operators to deliver services ontheir own networks as well as to subscribers utilizingroaming on other networks and they will have achance to compete beyond price.

The convergence of different mobile and wirelesstechnologies has resulted in a need for operators toreassess their earlier business models and to developnew ones to address this phenomenon. In PNs, differ-ent types of access technologies will work hand inhand to deliver communication and services to users.

PN Business Models and Strategies

– The Operator’s Perspective

S U - E N T A N , R U N E R O S W A L L

Su-En Tan is a

Post-Doc at

CICT, Denmark

Rune Roswall is

Senior Business

Manager at

TeliaSonera,

Sweden

The EU project MAGNET-Beyond has examined the impacts of Personal Networks (PN) on business

models for existing infrastructure-based network vendors and operators. In the current communi-

cation markets, companies deliver services to customers in cooperation with other market players

building on different business models. The aim has been to examine the implications of PNs on existing

business model and business strategies of the mobile operator. This work consists of two main

elements: a) A presentation of the business model concept and its implications for PNs and mobile

operators, b) examinations of the PN business model as well as possible business models and

strategies that the operator could develop for the PN.

1) MAGNET: Annex 1 – ’Description of Work’, 2003, page 40.2) MAGNET: Op.cit., page 6.3) Ibid. page 6.4) See, e.g., Jani Suomalainen: ‘Service provisioning in the Virtual Home Environment’, Helsinki University of Technology Telecommu-

nications Software and Multimedia Laboratory, htth://www.tml.hut.fi/T110.551/2002/papers/May/Jani.suomalainen.pdf, and UMTSWorld, Virtual Home Environment, http://www.umtsworld.com/technology/vhe.htm

17Telektronikk 1.2007 ISSN 0085-7130 ©Telenor ASA 2007


This has important consequences for business mod-elling. Generally, business modelling is a supply sideexercise. User needs, targeted market segments andvalue propositions must be part of the modellingexercise. Basically, however, business modellingdeals with the relationships between the players onthe supply side in order to determine how they canservice the needs on the demand side. In the case ofPNs, the demand side has to be directly involved inthe creation of business models. The reason is thatuser groups can set up parts of the network infrastruc-ture and construct and deliver the services them-selves, and will often only need to interconnect andwork together with commercial network providersfor parts of the network and service delivery assign-ments. PNs will therefore often consist of combina-tions of service delivery relations (i.e. from a busi-ness enterprise/operator to an end-user) and self-organized networks and applications.

PN business modelling may also contribute to stan-dardisation work. Standards Development Organisa-tions (SDOs) look to standardise technologies withinindustries. Business modelling shows how differentrelationships exist between different actors in a valuechain. This could in turn assist SDOs when lookingat which technologies to standardise. Standards worklooks at the entirety of the technology in questionand business modelling shows the different actorsinvolved in the technology in question. Thus, whenstandardisation work is taking place, different actorsand their contribution to the technology may be takeninto account.

2 Business Models

The business model concept is used to analyse notonly the service aspect of PNs but also the organiza-tional, technology and financial aspects of PNs. Forour ongoing work on PN business modelling we areusing works by [4], [2], [5], [6] and [7]. Making useof the paper by Faber et al. (2003), we have analysedthe four interrelated design domains, which areshown in Figure 1. It should be noted that the Fabermodel bears much similarity to the Osterwaldermodel, and that other related ontologies and modelsexist.

There are four basic constituent elements in thisbusiness model ontology: Service design, technologydesign, organisation design, and finance design.Using this business model ontology, it is possible toinclude all relevant elements of business modellingand, in the context of the present paper to examinethe implications of the development of the PN con-cept for existing mobile and wireless operators.

Because the finance domain is somewhat related tothe charging model of a PN, it is possible to make useof the finance design of the business model to analyseand describe what the PN charging model will looklike. The four domains are interrelated domains suchthat decisions made in one or the other would in someway affect what happens in another domain. Briefly,the four domains are described here [2]:

Service Design: Description of the service (valueservice) that this network of companies will offer toa target group of users.

Organisation Design: Description of the network ofdifferent actors that is required to deliver the valueservices to the end users. Also the roles played byeach actor in the network.

Technology Design: Description of the fundamentalorganisation of the technical system and technicalarchitecture that is needed to deliver the value service.

Finance Design: Description of revenue that isintended to be obtained or earned from the value ser-vice. It includes risks, investments and revenue divi-sion amongst the different actors.

The present day mobile operator’s business modelmay be represented by Figure 2. In this model, themobile network operator is expected to be the maincustomer facing unit of the value chain. Other mem-bers of this model are the portal/content aggregator,the service providers, the ISP, the mobile devicemanufacturer, other types of network providers(fixed, Wi-Fi and other mobile network operators)and the network equipment vendor. These are themain players in this particular business model butthere are obviously other smaller members that havenot been indicated here. The relationships betweenthe different players are indicated by the arrowsdrawn, and the different services and revenueexchanges are also indicated. What is important tonote here is that this business model represents themobile network operator as the most significant role

Figure 1 The four inter-related design domains [2]

Servicedesign

Technologydesign

Organisationdesign

Financedesign



of the value chain. Although different actors all arelooking at increasing revenues, the mobile networkoperator seems to be in the best position to createnew business different from its traditional core busi-ness in mobile service provisioning.

As the mobile industry moved towards 3G and dataservices, the mobile operator showed interest intaking over some of the functions that were previ-ously performed by other actors. This is the naturalway to act for the mobile operator as their operationalcosts must be reduced while new service concepts aredeveloped in order to increase their efficiency andARPU (Average Revenue per User). Several tasksthat are performed by other actors may soon fallunder the domain of the mobile network operator.Examples include Mobile Virtual Network Operators(MVNOs), portal and content aggregation, paymentservices and also service provisioning.

The model shown in Figure 2 is valid for today’smobile industry but it may also be valid in a PN envi-ronment though the number of actors and their rolesmay change. It is, however, just as likely that newbusiness models will be presented in the future PNenvironment.

The business model shown in Figure 2 is an evolutionfrom the 2G or GSM business model. Earlier businessmodels were simpler because there were few or nodata services at all offered for the users. However, astechnology has evolved and data services have beenmore popular, the business model can be representedas in Figure 2. Newer services and applications willmake changes to partnership relations and the busi-ness model of the mobile operator.

In this model it would be difficult for other roles toemulate in a short time span because of the manyadvantages held by the mobile network operator. In

Figure 2 Business model of today’s mobile industry

Mobile network

operatorMobile device

manufacturer

Network

equipment

vendor

Other

operators

Revenue

from sales

Service

providers

Functionality

Functionality

Business

oppottunies

Revenue

Access and

interconnectivity

Subscription

and revenue

Internet

access

Customers

and revenues

Portals and

applications

ISPPortals/content

application

Revenues from

sales and

maintenance RevenueCustomers

and services

Volume sales and

distribution channel

Co-marketing

strategies

Functionality



a PN environment, however, the role of the mobilenetwork operator may change.

The mobile operator’s role today is developed fromthe 2nd generation operator role, with the additionof mobile data services. The market for present daypacket based data network services is still consideredvery young and will continue to grow. Voice stillplays a big role in the operator’s revenue stream, butas the volume increases, the revenues from data ser-vices are likely to increase even more rapidly due toadded value and efficiency. The operator’s serviceofferings are changing with the new technologies thatare being introduced. This also results in organisa-tional changes as well as in the network operators’financial performances.

In a converged arena, the business model adopted byoperators would also change. Convergence results innew services and new ways of combining services.It also results in new business opportunities. Earlier,the industry saw the convergence between fixed andmobile services. Today, we still see this, but with theaddition of further convergence between wireless andmobile technologies. With wireless technologiesbecoming ever more popular, some operators havetaken wireless technologies as a complementaryproduct to their mobile services and offered them inpackages to users. Convergence has therefore led tooperators redefining their business model and makingchanges to the way old services are sold.

The future mobile industry will likely be personalisedand quite industry specific in terms of its offerings.

General services like voice and simple data serviceswould probably still be offered but value added ser-vices and applications over and above these generalofferings will be more important for the actors of thevalue network.

The business model design domains as developed byFaber et al. and used here are not static models inpractise. Variable and relation changes result in adynamic model that is being adjusted constantly. Thebalancing of all these factors and variables is whatcreates a dynamic model. Changes in one domain willlikely affect one or more of the other domains. Theeffect may not always be direct but could sometimesbe observed as an indirect consequence.

3 PN Business Model

MAGNET is all about Personal Networks or PNs. APN is the total network made up from P-PANs and aninterconnecting infrastructure. Communication withina PN can be anything from short range communica-tion between personal devices (e.g. between a laptopand a mobile phone) to wide coverage communica-tion such as UMTS [8]. In a personal network, theuser is responsible for choosing between the availablechoices. The business model in a PAN or PN envi-ronment has been categorised into a self-organisedand a service-oriented business model, or into a com-bination of both. Figure 3 shows the differencesbetween each of these definitions.

The self-organised model is a model where no finan-cial exchange takes place, for example there could betwo users connecting to each other’s devices usingBluetooth. It is also possible to refer to the modelwhen a user connects to a Wi-Fi network withoutpaying for this service (it may already be paid for byhis company or it belongs to a friend). In a technicalcontext, a self-organised network is based on its owncapabilities and preferences in contrast to externallyforced actions. By combining the financial definitionof the self-organised model and the technical defini-tion of a self-organised network, the self-organisedbusiness model is the one that is formed based on itsown actions independent of any external chargeableresources (i.e. no financial transaction in the immedi-ate sphere).

The service-oriented model is a model where a finan-cial transaction takes place. In this case there will bea payment by the user to the Wi-Fi Service providerin exchange for connecting to the Internet.

The combination model would encompass both ear-lier models where a self-organised and a service-ori-ented model exist. This will probably be the most

Figure 3 Examples of self-organised, service-oriented and combinationmodels

User 1 User 2Bluetooth

connection

Self-organised

model

Wi-Fi

connection

Internet

Service

oriented

model

Combination

model



common case in a PN environment, where there willbe different types of communication, either through anetwork operator’s connection or through a personalpeer-to-peer connection.

Ad hoc networks may exist in any of the combina-tions. Since ad hoc networks are defined as beingwireless self-organising systems formed by the co-operating nodes within the communication system,they form temporary networks5) with a dynamic anddecentralised topology. Self-organised networks cantherefore be built upon PN agent networks in theMAGNET scenario.

The Magnet Beyond project participates actively inthe IETF and other Standardisation groups.

Routing in personal networks poses two special chal-lenges. First of all traditional IP and Mobile networksolutions are reasonably stable. In a personal net-work, the network topology is constantly changing.

Second, in traditional routing the solutions rely ondistributed routing databases, maintained in either thenetwork nodes or specialized management nodes. Inmobile ad hoc networks, nodes cannot be assumed tohave persistent data storage, and they cannot alwaysbe trusted.

The Internet Engineering Task Force (IETF) and itsmobile ad hoc networking groups MANET, Nemoand autoconf have studied solutions that haveaddressed those challenges.

4 Operator’s PN Business Model

The PN bring a lot of new opportunities for thedifferent members of the industry, and particularlyso for the mobile operator. The mobile operator hasalways enjoyed a dominant position in the mobileindustry and in a PN environment, this could continue.

In the PN, there will be new functions that a mobileoperator will wish to take up. One example is the PNagent. The PN agent is a management entity locatedin the interconnecting structure (most likely to be theInternet) that keeps track of each Personal Node andall the clusters within a PN. Because there are manyways to implement a PN agent, it is viewed more asa concept than a physical node. The PN agent maybe centralised and under the control of an operator orservice provider or distributed over several operators.It can also be hosted by the user’s terminal [3].

P-PANs and PNs are likely to play a big role in themobile operator’s future service offering. For themobile operator of today, one of the challenges is theconvergence of different networks: fixed, mobile andwireless. Fixed network operators have been offeringa wide range of services from networking to applica-tions and software to clients in order to provide amore comprehensive suite of services and a one-stop-shop option. Vertical integration of services hasplayed a part in the growth of the fixed network oper-ator for some time now. With mobile operators mov-ing into the same service offering, there are severalthings first to consider6):

• Ability to integrate different network types• Availability of devices for use• Customised applications/software for each market• Simple Charging and Billing• Differentiated Quality of Service• Personalisation of Services, Security and Privacy

The mobile operator seems to be moving onwardsfrom their present 2G or 3G business models intonew waters. The 3G and beyond business models aredifferent for different operators. Depending on thestrategy that the company has adopted, the differentoperators could choose different aspects of 3G topursue.

One important new capability of the PN that couldaffect the mobile operator is that of a self-configuringnetwork. A self-configuring network is a network thatis able to simplify the life of the user by configuringitself to the user’s needs and requirements based ona set of pre-defined and configured rules or policies.This is in theory the P-PAN network concept itself.The P-PAN is a network that is able to automaticallyconfigure itself, based on earlier set requirements andinformation in order to give the best trusted connec-tivity available. Its goal is to make things easier forthe user.

For the future there will be new business modelsrelated to the mobile operator role. Mobile Operatorscould offer Edge routers/PN-Agents with PPAN con-nectivity services in their networks. They could alsooffer managed PN services enabling easy authentica-tion, authorisation, and billing for users with multi-access support. New business models would berequired for such services.

5) http://fismat.umich.mx/adhocnow/ – cited 280705.6) There are probably other factors to consider but those listed here are of great importance to the mobile operator and their service

offering in a PN concept.



5 The Future Mobile Operator

– General Aspects

The Mobile Operator faces a future full of newopportunities as well as threats, depending on the wayit sees it. Although the Mobile Operator’s core com-petencies are to build and manage mobile networks,this will not be enough in the future industry whereconverged services and technologies will be inte-grated. In the mobile value chain, it is likely that theMobile Operator will continue to play a significantrole. The Mobile Operator has already taken up sec-ondary roles in service and content provision andtransaction management.

Most often the Mobile Operator has a strong financialposition and some of them could be active playerswhen it comes to finding good positions in the futurevalue networks and Business Models. With thisfinancial strength, several important attributes that theMobile Operator needs to further develop could be:

• Brand• Service differentiation• Simplicity• Trust and security• Customer service.

Whichever area the Mobile Operator chooses to moveinto or to integrate into its present business, these fea-tures will continue to be needed. A good brand namenot only catches the attention of potential users, italso invokes a feeling of trust when users recognisethe brand name.

Trust and security are particularly important featureswhen it comes to financial transactions. This alsoapplies when personal data is exchanged. MobileOperators will have to work closely with financialinstitutions to develop processes where these twofeatures are of the utmost importance. A workingrelationship with security firms will also be neededin order to provide secure transactions to users.

Efficient customer service and help services gainedfrom past experiences should become an integral partof the Mobile Operator’s business processes. Thisservice can also be seen as a means of channelling thecustomer’s requirements, needs and preferences tothe Mobile Operator which the operator can then useto improve services.

Service and product differentiation are needed toaddress different corners of the market. Some usersfind certain services more important and useful toothers. Also, to differentiate their service from othermobile operators, it is also possible that special or

unique services will play a distinguishing role. Withso much competition amongst Mobile Operators,price is one way to differentiate and with schemessuch as pre-pay or flat rates with data services,mobile operators can then differentiate themselvesfrom others. Of course, other service differentiationmethods exist for example after sales service andcustomer care or bundling of services.

Simplicity and ease of use are something that userswould want, even as the number of services andapplications increases. This is one area where theMobile Operator has an advantage over the otheractors. The Mobile Operator is in the best position tointegrate different services and to offer them as onesimple package to users. The customer expects to beable to communicate and connect to the services overdistances, at any time and everywhere. There shouldbe no technical borders as to service availability androaming. Life should be made simpler and not morecomplicated with new services. Services should beintuitive to use and technology should be invisible.

In order to survive, the Mobile Operator beyond 2010will have to face key issues such as:

1 Implications of a converged mobility – broadbandenvironment

2 Business refocus3 Network sharing4 Finding new and profitable business models.

The traditional strengths of the operators in the valuechain network have been their network assets, butnow there is a tangible shift towards brands, organi-sation and market channels.

There is a consensus among network operators thatlong-term convergence of voice and non-voice net-works must be more than an exercise in cutting costs.The main drivers of change will be to find valuesaround mobile and broadband. The focus of theindustry has to shift towards services and there willbe an increased competition from players from otherimportant industry sectors.

The future operator will mainly sell access and arange of applications, content, devices and services.Operators have to formulate new business strategiesthat in co-operation with its customers could derivevalues from intelligent edge applications and devices.In the long term, operators will have to evolve frombeing network service providers to being communica-tions enablers. The new opportunities could includetools for proper interoperability management ofhome, public, business and other private networks.



6 Operator’s PN Strategies

As PNs and other new concepts in technology as wellas new technologies are introduced to the mobileindustry, they provide a multitude of new areas whichold players can get involved in. For the mobile opera-tor, being the player with the greatest market influ-ence at this point in time, it is likely that related ser-vices would fall under their service offering. Networkprovisioning will not be sufficient to carry the mobileoperator into the future but together with other ser-vices, the mobile operator will be able to provide asignificant number of value-added services that arenot part of their business today. In a PN environment,some areas (old and new) which could lead to newrevenue streams for today’s mobile operators could be:

• Customer aggregator• Value network integrator• Content provider or content aggregator• Clearinghouse for Billing, DRM and Security• Financial service provider• Mobile, Internet and• Ad Hoc Network Service Provider (providing

Hosting services, Edge routers, PN agents etc.)

The above list7) mentions some of the possible rolesthat the mobile operator could go into. Some of theseroles are already in place but others are new andcould become a requirement in a PN scenario. As theindustry actor with the most customers today, themobile operator has a large potential base with whichto work with. For the mobile operator, two key func-tions to concentrate on could be converged servicesand the personalisation of PN services for its cus-tomers.

One aspect to note in the PN scenario is that the net-work operator’s role in the PN will probably dependvery much on the type of industry it is addressing. Aswe move from a general market to one that is spe-cialised and personalised in many ways, the businessmodels will become more industry specific to addressthe needs of each particular industry.

Industry specificity may result in the mobile operatorand other players coming up with differentiated ser-vices for each case and for groups of users withineach case. Of course, it all depends on the businessstrategy that the players choose to adopt and the busi-ness models developed. Within the general PN con-cept, the business model will still be used to describethe different operators’ relationship with each other,their relationships with content aggregators, serviceand application providers, device manufacturers, plat-form and equipment manufacturers or providers, and

also other peripheral players in this mode. However,for the specific cases, different business models maybe in place and therefore relationships between actorscould differ for each of the cases.

The Regulator is responsible for the allocation ofspectrum, definition of usage policies and rules butalso for handing out penalties for violation of respon-sibilities. Players in this area include the government,regulation authorities, event associations and stan-dardization groups.

PNs are expected to operate in two different modes,High Data Rate (HDR) and Low Data Rate (LDR).PN technology could be temporarily deployed at localsites like a sports arena or an accident area. Spectrumissues might have time, cost, interoperation and loca-tion based constraints.

Publishing rights (DRM) might be a possible conflictarea. The 2001 European directive on copyrightforces member states of the European Union toimplement legal protections for DRM. Digital RightsManagement however has an uncertain legal statusin many countries since the content rights of usersand content publishers are ill-defined.

Customer aggregator

PNs and WLAN have the advantage over UMTS orGPRS of providing high bandwidth at low cost. TheMobile Operator could provide the end-user withdetailed position information about their reachableWLAN interconnection points.

Locations where PN Networking will be commonneed interconnection to IP networks. Many mobileoperators have already got UMTS and also WLAN-networks. While WLANs today address areas whereprofessional users have a special interest to connect itmight also be useful for the operator to consider loca-tions such as shopping malls, city centres and variouspublic arenas for their future WLAN roll out.

UMTS deployment is a common pan-European initia-tive with special licence requirements on coverage,capacity and roll out. WLAN and WiMAX technolo-gies are global initiatives which are developing inde-pendently and quickly in many environments. Due tothe different scopes of the technologies there mightbe delays in their convergence.

The service could be combined with informationabout the closest Internet connection and other usefulinformation such as price, network load and other

7) The authors acknowledge that other new roles could exist for the mobile operator that have not been listed here.



services offered at the interconnection points such asprinting or other non-technical services.

Value network integrator

The Mobile Operator could offer a complete conceptfor P-PAN clusters and support them with customercare, billing, sales and marketing. Overall technicalsupport for PN technology could be added (installa-tion, operation and maintenance etc). The technologyshould be so simple that anybody could install andoperate it.

In a future with maybe thousands of PPAN clustersthe cost for the roaming, DRM and security processeswill be much too large for a small PN operator. AMobile Operator can use its knowledge and experi-ence to set up necessary agreements and processes tocost effectively deal with a large number of businesspartnership agreements. The costs could be financedby small transaction fees as a percentage of the valueand/or a fixed fee per transaction.

The billing alternatives for a PN operator are to:

• Invoke their own billing system• Use the billing system of an ASP• Use the billing systems of the mobile operators.

Billing systems are expensive and set up will causetime delays for the service launch plan. To buy abilling system is therefore not a realistic option fora small and independent PN operator.

The competitive advantage for a Mobile Operator vs.an ASP is the customer base and the existing billingprocedures for mobile services. There are alsonational accounting laws that could complicate thismatter further. In principle it is also possible to useprepaid cards issued by the Mobile Operator to payfor the network access.

Service and content distribution

PNs will enable new ways of broadcasting and/ormulticasting. The mobile network can be used to dis-tribute services or content to a limited number of dis-tribution points. The advantage of using the mobilenetwork instead of the fixed network is that the distri-bution points may be mobile or be located in loca-tions beyond the reach of the fixed network.

The content could be forwarded from the distributionpoints to the end-users via WLANs or P-PANs. Thisprovides an opportunity for a limited number of indi-viduals who are within the communication range of thedistribution point to consume content of common inter-est. They could also share the cost of providing contentto the distribution point via the mobile network.

mCommerce

A PAN gateway related to a special individual and noone else should be able to communicate with Point ofSales (PoS) terminals enabling mCommerce. The PN-device could then be used as an electronic wallet. ThePoS terminal will have to verify the credit worthinessof the individual customer requesting the service andcheck the user’s unique rights to use the dedicatedterminal. Security is absolutely essential for possiblemCommerce applications.

The central position that mobile operators hold todaymeans that they will probably have a significant shareof the PN and are able to hold a strong position in thePN market. However, these operators will have toadapt to a new type of business where they are notjust selling access or bandwidth. Service provisioningand other value-added services will be important inthe PN market and the operator is in a good positionto provide all these. The ability to develop new ser-vices and to change focus is likely to be key ingredi-ents to the mobile operator’s role in the PN.

Operators have good abilities to compete with ICThouses by offering total support for PN Services. Theyhave a unique ability to make combined service pack-ages from PN Network and other services. Anotherrole could be to provide global PN roaming support.

7 Conclusion

In the current communication markets, companiesdeliver services to customers in cooperation withother market players using different business models.The aim of this work has been to examine the impli-cations of PNs on existing business models, whichmeans development of business models for PNs.

The business model concept has been examined andmade operational by adopting the business modelontology developed by Faber et al. There are fourbasic constituent elements in this business modelontology: Service design, technology design, organi-sation design, and finance design. From the businessmodels sketched out it is possible to see a multitudeof opportunities for the actors in the value networklike the mobile operator and the device manufacturerwith respect to PNs. The implementation of the busi-ness model will be dependent on existing market con-ditions and actor capabilities.

One of the MAGNET project’s goals is to utilize thePN technologies to make the public life much easier.The public life covers the individual’s interaction withthe public system. This includes the welfare system,tax system, libraries, transportation, schools, nursinghomes, and health care system just to mention some.



PNs will consist of a self-organised model, a serviceoriented model and a combination model. The opera-tor’s role in the PN is still undefined but its dominantrole in today’s mobile industry ensures that it has agood chance of succeeding in the many opportunitiesthat will arise from PN services.

New business strategies and business models willhave to be developed to cater to the operator’s newworking environment. PNs present both threats andopportunities to the mobile operator. Some new andthrilling roles that could be further developed by theMobile Operators are Customer Aggregation, ValueNetwork Integration, Service and Content Distribu-tion and mCommerce.

In the long term a split-up of the Value Networkcould be foreseen and Partnership Relation Manage-ment will be increasingly important to boost themarket for PN Services.

Acknowledgement

This paper describes work performed under thesupervision of the Magnet-Beyond project which ispart of the EU’s IST program. MAGNET-Beyond isa continuation of the MAGNET project (www.ist-magnet.org). MAGNET Beyond is a worldwide R&Dproject within Mobile and Wireless Systems and Plat-forms Beyond 3G. MAGNET-Beyond will introducenew technologies, systems, and applications that areat the same time user-centric and secure. MAGNETBeyond will develop user-centric business modelconcepts for secure Personal Networks in multinet-work, multi-device, and multi-user environments.MAGNET Beyond has 30 partners from 15 countries,among these highly influential Industrial Partners,Universities, Research Centres, and SMEs. The viewsand conclusions contained herein are those of theauthors and should not be interpreted as necessarilyrepresenting the Magnet-Beyond Project. The authors

would like to thank all members of the Magnet-Beyond Project, who have contributed to the develop-ment of the concept presented in this paper.

References

1 MAGNET. Description of Work. Annex 1, 2003.

2 Faber, E et al. Designing Business Models forMobile ICT Services. 16th Bled Electronic Com-merce Conference, Bled, Slovenia, June 2003.

3 IST-MAGNET. Refined Architectures and Proto-cols for PN Ad-Hoc Self-Configuration, Inter-working, Routing and Mobility Management.October 2005. (IST-MAGNET WP2, Task 4,D.2.4.3 Draft)

4 Osterwalder, A, Pigneur, Y. An e-Business ModelOntology for Modelling e-Business. 15th BledElectronic Commerce Conference, Bled, Slove-nia, June 2002.

5 Afuah, A, Tucci, C L. Internet business models.New York, McGraw-Hill/Irwin, 2001.

6 Methlie, L, Pedersen, P. Designing business mod-els for customer value in heterogeneous wirelessnetworks. Bergen/Grimstad, May 2005.

7 Porter, M. Competitive Advantage. New York,Free Press, 1985.

8 Niemegeers, I G, Heemstra de Groot, S M. FromPersonal Area Networks to Personal Networks:A User Oriented Approach. Personal WirelessCommunications, Kluiwer Journal, May 2002.

9 IST-MAGNET. Socio-Economic Impac and Busi-ness Models for PNs. December 2005. (IST-MAGNET WP1, Task 4, D1.4.1b)

Su-En Tan is a post-doc researcher with the Center for Information and Communication Technologies

(CICT) at the Technical University of Denmark. She recently completed her PhD in Electronics and Commu-

nications from the same university. She is currently working on the IST MAGNET-Beyond project.


Rune Roswall, MSc BA, graduated from Linköping Technical University in 1975 and holds an MSc Degree

in Technical Physics / Applied Mathematics and a B.A. Degree in Economics. Rune Roswall is primarily respon-

sible for technical and business development within TeliaSonera Sweden, focusing on Multi-Access, Wi-Fi

technology, IPv6 and Peer-to-Peer networking. At present Mr. Roswall is working with Business Models, User

Requirements and User Profiles for Personal Networks for the European 6 FP Magnet Beyond project.




1 Introduction

My Personal Adaptive Global Network (MAGNET)will develop user-centric business model conceptsfor secure Personal Networks (PN) in multi-network,multi-device, and multi-user environments. By meansof adding personal network into the definition ofarchitecture the focus of the legacy business modelshifts more towards users’ needs and requirements.

It is assumed throughout the MAGNET project thatthe user requires greater access to a variety of data.There would be a multitude of sources from transmit-ting and/or accessing information. Since the marketentrance of GSM the value added service offered isincreasing and with that the business model is evolv-ing, as we notice in the case i-mode and 3G. Around1994 state control over the incumbents throughoutthe European Union started to dilute and policieswere put in place to facilitate greater competitionamong network providers. With the evolution ofthe mobile telecommunication the personalisationtowards the user has also been enhanced, thus thepersonal area network technologies have started pen-etrating the market, such as Bluetooth and Zigbee.

Wireless Personal Area Network (WPAN) hasenabled new applications under different user scenar-ios where the WPAN expands the coverage area.

Certainly, network providers retaining their centralrole in the business model would seem ideal sincethey have the economies of scope and scale and thebilling mechanism to facilitate cost and traffic con-trol. Wireless Local Area Network (WLAN) has wit-nessed a diverging business model where wirelessinternet service providers (WISP) and independentWLAN providers are driving the value chain, withbilling and network managed by them.

Network operators have also entered this market andsome by acquiring WISPs. The business model willbe the key determinant in devising the policies forinterconnection and billing.

Table 1 presents a list of scenarios generated pertheme as provided in [1]. These cases share themeswith one another; therefore a common platform canbe used by the content provider to provide service.

Interconnection and Billing Policies for Personal Networks

R A J E E V R P R A S A D , V A S I L E I O S S K A L D A N I S

Rajeev R. Prasad

is Director in

Sabita Holding,

Denmark

Vasileios S.

Kaldanis is

Research

Engineer in

National Techni-

cal University of

Athens

MAGNET proposes a dynamic shift in the business model from supply centric to demand centric

models. Policies for access and interconnections would need to be addressed in order to maximize

the social welfare and competition. Billing would be a complex issue that would need readdressing with

the changes in the paradigm. Requirements of internet’s stakeholders and the telecommunications’

stakeholders would have to be integrated in a single billing model.

Diabetes Smart Student Mobile Distributed Digital Whole Virtual Emergency

case shopping case gaming work living person home case

truck

Transportation X X X X

Entertainment X X X X X X

Society/Citizen X X X

Health care X X

Emergency X X

Surveillance X X X X X

Collaborative work X X X X

Community X X X X X X X X

Travelling X X X

Education X

Shopping X X

Table 1 MAGNET users’ case and themes



Such as in the case of Health Care, the hospital couldbe providing service over its personal network andthen the question would arise of how the billing andinterconnection could be managed if the businessmodel were to be service/content provider centric?

The question that arises from implementation of aMAGNET concept such as MAGNET.Care1) is howis the service managed? There is a multitude of inter-connections and a multitude of formats of data. Wepropose a possible concept for billing and look intoInternet Engineering Task Force’s (IETF) and3GPP’s efforts in understanding the requirementsof users.

The expansion of PAN towards a PN will stress theexisting radio spectrum due to the increasing demandfor data anytime and anywhere. Therefore, the trafficneeds to be optimally managed, for which we proposeconverged billing. Telecom operators who must sharerevenues and profits coming from multiple services,content or application providers whose customersexpect to receive a single bill for all the services theysubscribe to. Adoption of the right billing strategyamong telecom operators will lead to better serviceexploitation, optimization and integration or inter-operation of future solutions with existing network.

This paper will present the evolving business modelsbased on network provider centric and service/con-tent provider centric in Section 2. Section 3 will

determine the problems relating to access pricing andinterconnections in the PN environment and the sub-sequent section, Section 4 will address how billingwill be managed for the users/subscribers and theaccess costs. Section 5 will be the conclusions of thestudy.

2 Business Model

The supply chain has become a major focus in thetelecom industry, many papers have been publishedproposing a shift in the business paradigm [2–4].The value chain will be less supply driven and moredemand driven than before. Addressing the dynamicsin the value chain is essential since it opens the gateto the uncertainties in regulating the interconnectionsand billing mechanisms for PNs.

Currently, the mobile industry is mobile networkprovider (MNO) centric as illustrated in Figure 1. Inthis situation the user/subscriber has a contract withthe MNO who also manages the billing system.

In case of 3G the practice among the operators isaptly illustrated also in Figure 1, it is via the MNOthat the user can access the services and contents. Therevenue generated is shared among the providers, forwhich the billing is managed by the MNO. Since theliberalisation in early 90s and common EU policieswere brought into place in order to catalyse competi-tion, operators expanded into new markets.

Figure 1 MNO Centric Business Model

1) http://www.ist-magnet.org/pr

User roaming in

Revenue sharing btw MNO & CP & ISP

Billing & charging agreement between user & MNO

Payment agreement btw users & content providers & users & 3rd party ISPs

User

1

User roaming in

foreign network

Contentprovider

ISP

Mobilenetwork operator

User

Roaming partnerMNO

User

1: Access Contract for Network sharing charges & User roaming revenue sharing

Payment



The regulations stipulated incumbents to share theirnetwork at a fair price with the new operators. Overthe years this has made the market more efficient andhas diluted the power of the incumbents, the businessmodel evolved into new MNOs leasing the networkfrom the incumbents and/or the carriers by placingthemselves as mobile virtual network operators.Nonetheless, this shift did not change the businessmodel fundamentally, except added another stakeholderin the value chain. However, this led to diffusion ofpower from the supplier as the competition grew andthe market forces were more strongly determined bythe users’ needs and requirements. Internet representsa close to equilibrium business model, the users of theInternet participate in providing contents and servicestoo for both financial and non financial purposes.

The user makes an independent contract with theinternet service provider (ISP) and service/contentprovider. The internet allows the user/subscriber toco-create in the virtual domain and also allows theuser to build peer-to-peer network. Figure 2 illus-trates the architecture for a peer-to-peer network.

This is facilitated by a service provider such asKazaa, eDonkey, iMesh etc. Similarly, PN is exhibit-ing development of use cases where peer-to-peer willbe possible, hence the business model would be ser-vice/content provider specific. In user cases such as

mobile gaming the user will be able to play by estab-lishing peer-to-peer network, the PN would furtherenable users to also multicast similar to InternetRelay Chat (IRC) protocol. IRC is provided by enti-ties like MSN, the server forms the backbone of IRC,providing a point to which clients may connect to talkto each other, and a point for other servers to connectto, forming an IRC network.

In MAGNET.Care the scenario developed suggestsintegration between various networks and Internet.

Figure 3 represents users moving and communicatinghaphazardly. There is intra-PN communication, com-munication to and from other networks etc. The databeing exchanged is from voice to high data rate beingtransmitted, e.g. from an ambulance and/or useraccessing information from an Internet site. In thisscenario we are assuming that a service provider isnot the same as an MNO, thus the business model isservice/content provider centric.

In this scenario the content provider is controllingthe charging and billing, moreover presence of themobile operator in the model MNO can also be sub-stituted in several use case scenarios. A shoppingmall (service and/or content provider) can deploy itsown network and a user visiting the shopping mallcan conduct trade over the mall’s network. This value

Connection manager

No peers, connection manager

is waiting for peers to identify

themselves.

Another peer identifies itself, they learn

about each other from the connection

manager and open up a connection to

each other.

Peer connects

Peer connects

Connection manager

Peer 1, port 2234

Peer 1

One peer identifies itself

Connection manager

Peer 1, port 2234

Peer 2, port 2235

Peer 3, port 2236

A third peer joins the net and all three

peers find each other from the

connection manager.

Connection manager

Peer 1, port 2234

Peer 2, port 2235

Connection manager

Peer 1, port 2234

Peer 2, port 2235

The second peer disconnects, Peer 1

and Peer 3 stay connected and get

an updated peer list from the

connection manager.

Peer connects

Peer disconnects

Peer 1 Peer 2

Peer 1 Peer 3

Peer 1 Peer 2

Peer 3

Figure 2 Connection management Peer-Peer [5]



chain can also get more dynamic if the mall allowsremote access to its network, and then the MNO willalso be a participant in the value chain. Finally, thedynamism of the model is more enhanced with thepossibility of deeper convergence between telecom-munications and other industries. There will be aneed for a coherent and equilibrium inter-industrialbusiness model, in order to attain value flow.

Now we have described two different business mod-els, the latter is comparable to the business activitiesconducted on the Internet. However, in the telecom-munication industry network interconnection is a crit-ical issue and also limitation of the bandwidth. Inex-pensive technologies such as WLAN are allowingmuch smaller players to enter the network market,where they are able to provide network coverage toa city centre. Furthermore, with the advent of mobilenodes that can roam between cellular network andWLAN the interconnection issues thus become amore dominant aspect.

3 Interconnection Policies and

Personal Networks

Attempting to model an appropriate billing schemefor personal networks several approaches could befollowed based on existing solutions as applied byMNOs today. As described before the two generalapproaches (MNO centric and service/contentprovider centric) may able to provide a possiblehybrid billing solution in the case of PAN/PNs wherenumerous interconnections occur dynamically amongdifferent actors.

In an MNO centric model approach where the under-lying network infrastructure is owned by the mobileoperator itself, the expected interconnection policieswill not be much different than what it is right now.However, in the case where the service/contentproviders maintain their own network infrastructuresas a channel to distribute their digital products (e.g. aWLAN in the city centre) these providers may canni-balize into the high return market area of the MNOs.For example, voice over IP (VoIP) is quite a cheapersolution than cellular voice calls; therefore, demandfor this service would be greater. Consequently, aroaming PN user may choose to perform a long dis-tance voice call exploiting the VoIP capability of hisPDA rather than making a more expensive cellularcall, using a service provider’s application (e.g.skype) via its own enabling private network. In thisway, the PN user could effectively benefit from peer-to-peer technologies in order to avoid the imposingroaming costs of cellular networks.

Without MAGNET technologies, supporting a roam-ing user equipped with a PDA capable of accessingall wireless network technologies requires a directnegotiation regarding access fees while also provid-ing efficient billing management handling among theinfrastructure owners end-to-end. The user mayseamlessly roam from a 3G network to a WLAN onewith no service break, e.g. mobile gaming online inan IRC comparable architecture. The user’s require-ment will be certainly not to receive different bills forevery time he walks in and out of a cellular networkto WLAN. Nonetheless, a user who is either conduct-ing high data rate transfer and/or has high elasticity toprice will require being able to switch to a high band-width and/or inexpensive network.

The access pricing is essential to be established inorder to avoid ineffectiveness in ex-post applicationof Competition Law. Since the first steps towards lib-eralisation until 2002 incumbents still have consider-able strength, most notably in local loop, not muchcompetition is observed and the unbundling is pro-gressing slowly [6]. Several papers [7] have analysedone-way and two-way access scenarios to mitigateanticompetitive behaviour and to also encourageother ex-ante objectives.

In the PN case where we assume the co-existence ofsmall access network owners such as a shoppingmall, unbundling the local loop (ULL) is not essentialfor the shopping mall; however calls originating froma UMTS network and terminating in the shoppingmall creates a one-way access. Manifesting the powerwith the MNO to determine the access price with the

Figure 3 Service/Content Provider Centric BM

MNO 2

Shopping mall

MNO 1

Hospital

Access & roaming contract among all networks

1. Ambulance accessing patient’s record

2. Doctor accessing patient’s record

3. Multicasting

4. Exchanging data

5. Sending alert to family

4

2

5

31

1



shopping mall also exposes the mall to predatorypricing, excessive pricing, price squeeze, etc. by theMNO, for instance.

ULL becomes a lesser issue for MNO who does nothave infrastructure covering the last mile and whoowns the WLAN network in the shopping mall undera revenue sharing context, for instance. The MNOcan circumvent the incumbents of last mile by pricingtechniques and also by innovative technology, e.g.WIMAX (see Figure 4).

As seen in Figure 4, the concept of pricing PNs usingthe last mile access (network and services) via thehome or a roaming MNO is illustrated. Several ser-vice providers (SPs) can be aggregated by an Aggre-gated Service Provider (AgSP) and make their ser-vices available to an MNO via specific interfaces.Furthermore, the MNO itself may establish ServiceLevel Agreements (SLAs) with local third partynetwork infrastructure providers (e.g. WLAN, Wi-Fi,short range WiMax, etc) in order to simplify thebilling process for its client who aims at using aplethora of cost effective PN services. Each localnetwork/service access provider (as an SLA partnerof the MNO) may individually charge the client forthe relevant service costs consumed directly at hismonthly account bill. In this way the client can havea direct access to web services by paying low costconnectivity for his PN (see [16]).

In a PN scenario the interconnection access chargedoes not vary from how it is today regarding local

loop and therefore does not warrant a new theory.The regulatory bodies will have to consider the inter-connection under the one-way and two-way accesscontext. In our scenario of a shopping mall complica-tion of a two-way access is foreseeable if the ser-vice/content providers have their own infrastructure.The regulators will face following issues in determin-ing a fair access price [7]:

The first issue concerns optimal access pricing andwhether it should be reciprocal; the access priceshould be offset against the social welfare maximisa-tion where the total industry profit is sufficient tocover the total fixed costs. Regarding reciprocal pric-ing, agreement can easily be reached between opera-tors originating similar size traffic or similar marginalcosts. However, an operator starting from scratch willfind it more difficult to enter a two-way access agree-ment.

The second issue in the PN with the smart shoppingscenario we have to regard is whether the mall’s net-work is competing with the MNO’s network or arethey complementary? Nonetheless, the size of thetraffic over a mall’s network would be a combinationof WLAN and WPAN techniques, so potentially thesize of a single user’s data could be larger if originat-ing from the mall’s network and terminating in anMNO’s or carrier’s network as opposed to vice versa.Should the price be determined by the regulators orthe market? If determined by the market informationsymmetry becomes a critical issue, where perfectcompetition will have to be assumed [6].

Finally, according to Laffont (2001) [8], all intercon-nection costs will be set at competitive level if themissing price is recovered. He is referring to thecalling party pays principle; there is a missing price,namely price for receiving costs. This pricing modelwas resorted to in the Indian market, but is now abol-ished although it is a lucrative marketing strategy foroperators. Nonetheless, the case is not attractive forusers.

The main item to address is the billing management.This is a major issue to tackle when we are expectingpeer-to-peer network (see Figure 2) and multicastingthat is comparable to IRC protocol services in PN. Asillustrated in Figure 3 the service/content provider ismanaging the billing, however how would the roam-ing cost be managed?

4 Billing Management and Roaming

Mobile business models of the present are MNO cen-tric and will continue to be so as long as billing isperformed based on the typical roaming process of

Figure 4 A PN billing scheme via MNO

MNO

SP 1

AgSP

…

WiMAX

Overall

MNO

SP 2

VASP…

WLAN Wi-Fi

La

st

mil

e a

cc

ess

User

Billing transactions

billing



GSM [10]. In Figure 5, we adopt the traditional busi-ness modelling method when billing a mobile sub-scriber who is roaming into another network (foreign)away from his home network. The logical extensionsto the MNOs are the Home PN Operator (HPNO) andForeign PN Operators (FPNOs) which are responsi-ble for supporting user mobility on networking leveland service level as well. Any service the user dis-covers in any visiting PAN/PN could become avail-able to him/her as long as a billing agreement amongHPNO – FPNOs can be established.

Although the concept of PN operator still has notbeen clearly defined within the MAGNET frame-work, ideally it can be used for the sake of the billingprocess. Any roaming user currently away from hisoriginated PAN should be able to define/verify him-self in any foreign network. In other words, thereshould be a secure interaction among agents able toverify that this user is really what he claims. That isprobably the most secure way of authenticating a userwho cannot acquire a certain level of trust by anotherlocal user in the foreign network, in order to becomeattached to that.

In a service/content provider centric business modelwhere they also own the infrastructure e.g. MAG-NET.Care, shopping mall etc. billing can be per-formed by this provider; however we will certainlynotice a complication in the management of roamingfee among a plethora of small networks. The second

issue to address is that the MNOs should be preparedto cater to larger traffic size which would also bemore sporadic than what it is today. The solutionwould have to be a combination of access technologyand billing [10], not only facilitating the cost incurredfor the network operators but to control the traffic aswell. Among the use cases in PN we observe deeperconvergence between Internet and mobile communi-cations. Therefore, an integrated platform for charg-ing, accounting and billing is required. The InternetEngineering Task Force (IETF) and 3GPP have beenaddressing the requirements of stakeholders for the3G market.

IETF [11–13] are addressing the changing needs ofusers, ISPs and service/content providers. Users arewilling to pay additional charges in order to ensure abetter quality of the provided services. It is apparentthat each stakeholder has their specific requirementsfor the billing and charging. This makes the billingmethod complex to integrate, since the user’s demandis for a one-stop billing.

One of the options available is to outsource the activ-ity of billing and charging to a clearing house. 3GPPWorking Groups and UMTS Forums [14, 15] anal-ysed the users’ requirements and those of the otherstakeholders in the value chain in the UMTS network.

Users/Subscribers essentially require from a UMTSnetwork a single bill and charging which is simple

Figure 5 Proposed billing scheme for PNs

Content service

provider (s) (C/S Ps)

C/S P - FPNO

business

agreement

HPNO - FPNO1

roaming agreement

Foreign PN

operator 1

(FPNO-1)

Home PN

operator

(HPNO)

Roaming into local

foreign PAN/PNs

Roaming into local

foreign PAN/PNs

Home PN

user’s

subscription

C/S P - HPNO

business

agreement

C/S P - FPNO

business

agreement

HPNO - FPNO1

roaming agreement

Foreign PN

operator 2

(FPNO-2)



and easy to follow. MNOs’ requirement is to be ableto manage various billing models which cater to tra-ditional billing as well as dynamic profile basedaccess control. Within MAGNET we have come tothe conclusion that applications such as Diabetesrequire dynamic profile based access control. Theservice provider gives varying level of access to dif-ferent stakeholders, so the access to a patient’s recordto a doctor will be extensive and for instance to afriend it should be limited.

A layered charging architecture approach is struc-tured in three layers: transport, service and contentare some of the requirements of the MNO. The man-agement of each layer will be conducted separatelyand hence the charging can be applied autonomously.Theoretically, it seems feasible however that thismodel is not in line with the users’/subscribers’requirements.

Perhaps a pseudo bundle-charging would be of pref-erence to the user; here we are referring to a conceptwhere the user pays a single fee for a package. TheMNO offers a charge to the user which is a combina-tion of fees from services/contents in the packagewhich has been opted by the user.

However, from the independent content/serviceproviders’ point of view, there is an emergingdemand that each authorised player should be able toapply dynamically the desired pricing policy for itsservices’ usage. These requirements are building theplatform for the case we have presented in this paper.

One hypothesis that can be presented is that billingand charging for users/subscribers is a consequenceof perception. Which means, since users/subscriberswould like to see one-stop billing, they only pay theirmonthly bill for subscription to the network whichalso includes voice and low data rate services. Whilein the process of using other forms of contents/ser-vices they are charged instantly at the time of pur-chasing the service.

5 Conclusions

Personal network concept is leading towards deeperconvergence between telecommunications and otherindustries, hence allowing also service/contentproviders to drive the industry. Coherence in the con-fluence of business models of several different indus-tries will have to be managed. Internet businessmodel is the most comparable to the paradigm per-sonal network is moving towards. The complexity inthe management of access and interconnection wouldstill exist; nonetheless at this stage of R&D in MAG-NET it still does not seem that a new theory would be

warranted; whereas billing would require to be add-ressed as infrastructures could be afforded by service/content providers and we can witness multitudes ofsuch hotspots.

Due to Mobile Network Operators’ large client baseand position in the value chain as an operator, billingcould be managed by them. Clearing house is anotheroption available. Whichever model chosen, it shouldcomplement the chosen business model and providethe equilibrium between all stakeholders.

Acknowledgement

The authors of this paper would like to acknowledgeall the contributions that different companies and per-sons have made in relation to the MAGNET projectand especially the contributions to WP1.

References

1 User requirement for PN to drive the definition ofa valid architecture. (MAGNET Report, D1.1.1b)

2 Koutsopoulou, M, Kaloxylos, A, Alonistioti, A.Charging, Accounting and Billing as a Sophisti-cated and Reconfigurable Discrete Service fornext Generation Mobile Network. IEEE Semian-nual Vehicular Technology Conference (FallVTC2002), Vancouver, Canada, September 2002.

3 Henten, A, Saugstrup, D. The PN Concept in aBusiness Modelling Prespective. Copenhagen,Denmark. (CTI Working Papers, No. 95)

4 Prasad, R, Kaldanis, V et al. Paradigm Shift ofBusiness Models and its impact on Billing inPersonal Area Network: A Diabetes Case Study.Shanghai, MAGNET Workshop, November 2004.

5 PCOM:I3 internal report.

6 Prasad, R, Monti, M. Billing and Pricing for Per-sonal Network within MAGNET Project. Italy,WPMC 2004.

7 Buigues, P. Latest Progress in LLU as reflectedby sector enquiry. LLU hearing, 8 July 2002.(Downloadable from EC website)

8 Armstrong, M. Network interconnections inTelecommunications. Economic Journal, 108,545–564.

9 Laffront, J J, Rey, P, Tirole, J. Network Competi-tion: I. Overview and Nondiscreminatory Pricing.Rand Journal Economics, 29, 1–37, 1998a.



10 Panagiotakis, S, Koutsopoulou, M, Alonistioti, A,Kaloxylos, A. Generic Framework for the Provi-sion of Efficient Location-based Charging overFuture Mobile Communication Networks. IEEE13th International Symposium on Personal,Indoor and Mobile Radio Communications(PIMRC 2002), Lisbon, Portugal, September2002.

11 Aboba, B, Arkko, J, Harrington, D. Introductionto accounting management. 2000. (RFC 2975)2006, October 27 [online] – URL:http://www.ietf.org/rfc/rfc2975.txt?number=2975

12 Carle, G, Zander, S, Zseby, T. Policy-basedaccounting. 2002. (RFC 3334) 2006, October 27[online] – URL: ftp://ftp.rfc-editor.org/in-notes/rfc3334.txt.

13 Jonkers, H, Hille, S. Accounting Context: Appli-cation and Issues. 2000.http://www.aaaarch.org/doc06/file-11249.pdf

14 3GPP. 3rd Generation Partnership Project; Tech-nical Specification Group Services and SystemAspects; Service aspects; Charging and Billing(Release 5). June 2001. (3G TS 22.115 version5.1.0)

15 3GPP. 3rd Generation Partnership Project; Tech-nical Specification Group Services and SystemAspects; Telecommunication management;Charging management; Charging principles(Release 4). September 2001. (3G TS 32.200version 4.0.0)

16 WiFi access in airports for mobile subscribers.Available from: http://www.lufthansa.de

Rajeev R. Prasad received his BSc (1999) from the International School of Economics in Rotterdam, The

Netherlands, and his MSc (2002) in Corporate Finance and International Business from Aarhus School of

Business, Denmark. He was a strategy manager in PCOM:I3. Furthermore, he launched a new company,

Wireless Integrated Billing and Security (2003), licensing Billing and CRM software for WLAN and provided

solutions for the network architecture and deployment of WLAN.

Currently, Rajeev is Strategy Director in Sabita Holding, a Danish company responsible for investments in

wireless communication. Rajeev has authored several academic papers and has written a chapter in a book

“802.11 WLANs and IP Networking: Security, QoS, and Mobility”. His academic work has been in the fields of

Economics of Innovation, Economics of Network and Corporate Strategy.


Vasileios Kaldanis received his bachelor degree (1998) in Physics from the Aristotle University of Thessa-

loniki in Greece. In 2001 he joined the Centre for Telecommunications Research (CTR) in King’s College,

London as a research engineer, where in 2002 he obtained his MSc in Electronic research. Since 2002 he

has been working as a research engineer in the National Technical University of Athens (NTUA), where in

2005 he obtained his MBA in Techno-Economic systems. Since 2006 he is also a PhD candidate in the

Department of Electrical Engineering and Computer Science of NTUA in reconfigurable networks. He has

worked as a consultant for several mobile operators and other telecom companies as OSS/BSS support

engineer, and his academic research experience includes active participation in numerous IST European

projects (CAUTION, CAUTION++, MAGNET, MAGNET Beyond), publishing numerous papers, articles and

journals. His background experience is in the area of wireless/mobile IP core and access networks, system

implementation/integration, software development, service & application development, resource manage-

ment for GSM/GPRS/UMTS and 4G wireless systems, business modelling and techno-economic analysis

for mobile telecom & satellite solutions.




1 Introduction

Take the concept of pervasive computing and com-bine it with strong user focus and you get PersonalNetworks (PN) [1], [2]. PN is a collection of one’smost private devices referred to as personal nodes.From a technical point of view, the PN is seen to con-sist of devices sharing a common trust relationship.Security and privacy are the fundamental propertiesof the PN, as well as its ability to self-organize andadapt to mobility and changing network environ-ments.

The IST project MAGNET [3] vision is that PersonalNetworks (PNs) will support the users’ professionaland private activities, without being obtrusive whilesafeguarding privacy and security [4]. A PN canoperate on top of any number of networks that existfor subscriber services or are composed in an ad hocmanner for this particular purpose. These networksare dynamic and diverse in composition, configura-tion and connectivity depending on time, place, pref-erence and context, as well as resources available andrequired, and they function in cooperation with all theneeded and preferred partners.

The PN consists of clusters of personal nodes. Onecluster is special, because it is located around theuser. The clusters are connected with each other viaan interconnecting structure, which is likely to beinfrastructure based.

In order to protect the privacy of the user and theintegrity of the PN, security measures are used toencrypt the user’s data when it is sent outside of thedevice, i.e. using a wireless medium or the infrastruc-ture. The user can reach all of his or her devices usinga variety of underlying networking technologies,

which are invisible to the user. The user only sees theservices that are available in the PN and on foreignnodes that have been made available to the user.

Nonetheless, personal communications cannot berestricted to the services provided by the devices theuser owns, but the possibility to interact with otherusers’ PN has to be enabled in order to support theuser in his/her private and professional activities. Theconcept of PN Federations (PN-F) is an even morechallenging one since the relations between usershave to be managed and the security has to be re-inforced in order to not open security holes whileallowing authorized users to cooperate with you. PNFederation is a secure cooperation between a subsetof devices belonging to different PNs for the purposeof achieving a common goal or service by establish-ing an alliance. It can be established through inter-connecting infrastructures (namely infrastructurecase) or by direct communication between PN nodes(namely ad hoc case).

This paper presents the mechanisms that end up in theself-formation of a secure network of all the personaldevices around the user, the P-PAN. Besides, it willdescribe how the extension of these mechanismsenables the establishment and use of PN-Fs in thead hoc case. The paper is structured as follows: InSection 2 a survey of the work that has been done onthe personal networking area as well as the key dis-tinguishing points that differentiates it from the workdescribed in the paper will be presented. Section 3will sketch the mechanisms specified for the self-configuration of P-PANs putting emphasis on thesupport of the heterogeneity and the security. Theextensions needed to the aforementioned mechanismsas well as the newly introduced procedures that are

Extending Private Personal Area Networks to Personal

Network Federations in Heterogeneous Ad Hoc Scenarios

L U I S S A N C H E Z , J O R G E L A N Z A , L U I S M U Ñ O Z

Luis Sanchez is

a PhD Student

at the University

of Cantabria

(UC), Santander,

Spain

Jorge Lanza is a

PhD Student at

the University of

Cantabria (UC),

Santander,

Spain

Luis Muñoz is a

Professor at the

University of

Cantabria (UC),

Santander,

Spain

Personal Network (PN) is an emerging concept which combines pervasive computing and strong user

focus. The idea is that the user’s personal devices organize themselves in a secure and private per-

sonal network transparently of their geographical location or the access technologies used. The user

expects the network to be always ready for supporting his necessities without requiring too much user

involvement. Additionally, the PN must be ready to share the services it provides to the user with other

users that have been authorised in order to allow the collaboration between the PNs’ users. The PN

Federation concept is presented as a secure cooperation between a subset of devices belonging to

different PNs for the purpose of achieving a common goal or service by establishing an alliance. This

paper will present firstly the mechanisms developed for the self-configuration of the Private Personal

Area Network (P-PAN, the cluster of personal devices that are around the user) for describing after-

wards how an extension of these mechanisms can seamlessly support the establishment and use of a

PN Federation in the case that several P-PANs come together



P-PAN

Personalnode

Personalnode

Personalnode

Foreignnode

Interconnecting structure

Foreignnode

EdgerouterEdge

router

Edgerouter

Personalnode

Cluster(work)

Cluster(home)

Personalnode

Personalnode

Personalnode

GPSDevice

Glucometer

User 1’s P-PAN

User 3’s P-PAN

User 2’s P-PANUser 1’shome

User 3’shome

User 2’soffice

PN federation 2

PN federation 1

Interconnecting

infrastructure

Figure 1 Personal Network

Figure 2 Personal Network Federations



needed to establish and use a PN-F in the ad hoc casewill be introduced in Section 4. Finally, Section 5summarises the main conclusions from the paper.

2 Related Work

Starting with the PACWOMAN IST project [5], [6]the research on the personal networking area hasgathered quite a lot of attention. More and more, theperson is accompanied by a variety of devices withgrowing capabilities that provide to the user anincreasing number of services. Contrary to other ini-tiatives that explore fields like wireless personal areanetworking [7], mobile ad hoc networks [8] or self-configuration [9], [10] in isolation focus on optimiz-ing the characteristics of each field without havingmuch in mind about the others. The solution proposedhere presents an integrated approach that copes withthe different connectivity, networking and servicerequirements in order to accomplish the aforemen-tioned vision of an autonomous and self-organizedsecure Personal Network.

The traditional view of personal area networks [7],[11]–[13], where the problem is limited to a radiocoverage issue does not fully cope with the realrequirements of the user that wants his/her nodes tobe networked but without compromising his/her pri-vacy, intimacy, etc. It is needed to develop a networkthat extends its coverage depending on privacy andsecurity relations between the nodes, thus being ableto join the same collaborative network and offeringits services seamlessly to the user which owns all ofthem.

Most of the security frameworks that are nowadaysdeployed are based on third parties [14], [15] thatenable users to be authenticated to each other, and touse the information in identity certificates (i.e. eachother’s public keys) to encrypt and decrypt messagestravelling to and from. Personal Networking shouldnot depend on third parties giving the full control ofhis/her devices to the user. Some security solutions[16] work in a more distributed way exploiting theweb-of-trust feature. Nevertheless, it is not fully com-pliant with the requirements imposed by PersonalNetworks. The PN security architecture in generaland the secure P-PAN formation in particular arebased on bilateral long term trust relationshipsbetween PN devices that are bootstrapped directlyby the user [17].

3 Personal Networking Concepts

and Terminology

This section presents the main entities that comprisethe proposed Personal Network architecture, defining

as well, the terminology that will be used in thispaper when describing the architecture and the differ-ent solutions adopted to support the formation andoperation of Personal Networks.

As shown in Figure 1, the PN consists of clusters ofpersonal nodes. One cluster is special, because itis located around the user. The clusters are connectedwith each other via an interconnecting structure,which is likely to be infrastructure based. In order toprotect the privacy of the user and the integrity of thePN, security measures are used to encrypt the user’sdata when it is sent outside of the device, i.e. using awireless medium or the infrastructure. The user canreach all of his or her devices using a variety ofunderlying networking technologies, which are invis-ible to the user. The user only sees the services thatare available in the PN and on foreign nodes thathave been made available to the user.

In this section, the terms and key concepts that areused in the PN architecture will be presented anddefined.

• Device: Any communicating entity.

• Node: A device that implements IPv6 and/or IPv4.

• Personal Node: A node related to a given user orperson with a pre-established trust attribute. Theseattributes are typically cryptographic keys with apermanent (as long as not cancelled, redefined orrevoked) trust relationship.

• Private Personal Area Network (P-PAN)/Cluster:A network of personal devices and nodes, charac-terized by a common trust relationship, which cancommunicate with each other without using non-personal nodes. Nodes and devices in a cluster canbecome members of a P-PAN when a person entersan area where the cluster nodes are located.A P-PAN is often referred to as a personal bubblearound a person.

• Personal Network: A Personal Network includesthe P-PAN and a dynamic collection of remote per-sonal nodes and devices organized in clusters thatare connected to each other via InterconnectingStructures.

• Trust Relationship: is established when two partiescommunicate and determine with a measure of cer-tainty each other’s credentials to set up a securecommunication channel using encryption mecha-nisms. When devices and nodes want to establisha secure communication channel, they build a trustrelationship by whatever means possible.



• Imprinting: A procedure to bootstrap a trust rela-tionship between two nodes that basically consistof an authenticated key exchange.

• Gateway node (GW): A Personal Node within aCluster that enables connectivity with the Inter-connecting Structures.

• Interconnecting structures: Public, private orshared wired, wireless or hybrid networks suchas a UMTS network, the Internet, an intranet oran ad hoc network.

• Foreign node: A node that is not personal and can-not be part of the PN. Foreign nodes can either betrusted or not trusted. Whenever trusted, they willtypically have an ephemeral trust relationship witha node in a PN.

Figure 3 indicates how the PN entities defined in thissection relate to each other. Note that the P-PAN is amaterialization of the Cluster concept. The differencebetween the P-PAN and the rest of the PN Clusters isthat the user is in the surroundings. This differenceonly takes importance on the so-called ServiceAbstraction Level (see Section III). Therefore, in theremainder of the paper we will use both terms with-out distinction.

4 P-PAN Formation

The proposed scheme aims at being a fully dis-tributed approach for resilience and efficiency rea-sons which will act in a proactive way for havingalways the P-PAN formed for the user to not sufferfrom additional delay and which will assume only

security and privacy boundaries for the P-PANcontour.

4.1 Supporting Heterogeneity

The capability of working in a heterogeneous envi-ronment is a must for future personal networks. Thisheterogeneity will be mainly reflected in terms of thedifferent air interfaces that will coexist in these sce-narios requiring additional schemes to handle thisheterogeneity.

The concept of isolating the upper-layers from under-lying wireless technologies and thus providing realmulti-mode can be achieved by introducing a Univer-sal Convergence Layer [18]. The UCL will mainlyact as an enabler for backward and forward compati-bility by defining a common interface towards thenetwork layer while managing several different wire-less access technologies independently of their PHYand MAC layers. In this sense, the solution adoptedmakes it possible for the nodes to have a single IPaddress independently of the number of air interfaces

Figure 3 Personal Network principalentities relationship

Figure 4 UCL High-Level Architecture

PN

Cluster P-PAN

1

1

1

0..*

GW

0.. 1

0..*

Personal node

1

1 ..*

Personal device

0..*

0.. 1

UCL

Applications

TCP / UDP

IPv4 / IPv6

IEEE 802.11a/b/g BT IEEE 802.15.x

SecurityDLC SecurityOverlay



it has. This way the routing protocol placed in layer 3will be able to settle routes embracing multiple radiodomains in a completely seamless manner.

The combination of these two techniques, UCL plusad hoc routing protocol, enables the solution pro-posed to manage the heterogeneity that will appearin the P-PAN environment.

From a security perspective, one of the most impor-tant design goals of UCL is to make sure that use ofa heterogeneous radio specific legacy security systemdoes not cause any additional security vulnerabilities.In order to accomplish this, UCL includes a keyderivation function which generates radio specificlink keys from the long term PN keys generatedduring the P-PAN formation exercise. In addition tomaking parallel use of different radio systems secure,the presence of UCL also provides an opportunity toupgrade or even complement the legacy radio sys-tems used in MAGNET.

4.2 Enforcing the Security

This is the key feature to be provided for the P-PANformation. In this sense, the objective of the follow-ing mechanisms is to assure that only personal nodesare members of the P-PAN and that any personal nodein the area is also able to participate in the network.

As a result of the pairing procedure [17], the peersderive a long-term shared key that is subsequentlyused to secure the communication between them.Each device must store this information securely inthe form of a device record. A peer record mainlycontains the following information: (1) Peer identifier– a unique identifier associated to the device; (2) PNkey – the shared secret derived from the pairing pro-cess.

Nevertheless, this imprinting procedure is only thebaseline over which the trust is built in the P-PANformation. In our solution, we assume the supportgiven by the UCL [18], which deals with layer 2encryption so authenticity and integrity of receivedframes can be immediately assured.

Firstly, a beaconing process has been implementedin order to be aware of the immediate neighbours

continuously. The periodicity of the beacons is to bedesigned depending on the dynamicity of the cluster.In this sense, context discovery and context aware-ness techniques could be applied to the inter-beacontime.

The beacons are used for advertising the node pres-ence. Mandatory payload field: Node ID – 20 bytespublic identifier. Currently it is derived as a digestover the peer’s public DH (Diffie-Hellman) key usedduring the imprinting procedure. Optional payloadfield: Node name – Human friendly identifier. Usedfor UI purposes only.

Upon the reception of a beacon, it will be checked ifit is already registered in the neighbours table; if thepeer is already registered, the entry will be updatedby reinitializing the expiration timer associated (notethat there could be multiple entries for a single identi-fier, each of them with a different associated deviceso as to allocate multimode devices). If the neighbouris not already registered an authentication methodwill be called in order to assure that the discoverednode is really a personal node. It is important to notethat each entry is unique by the pair: identifier of theneighbour and device from which the beacon wasreceived. In this sense, it will be needed to performa different authentication process for each of the airinterfaces with which it is possible to communicatewith the neighbour.

The authentication is performed through a three wayhandshake (Request – Response – Success) in whichthe long-term shared key is used to verify the identitydenoted by the identifier field in the beacon received.

The same procedure is used for neighbour authentica-tion and exchange of link session keys used at UCLlevel for Intra P-PAN communications encryption.The following notations are used:

• | – concatenation• HMAC(key, data) – hashing function• gx – public Diffie-Hellman key• NX – nonce• E(key, data) – symmetric encryption

FF: FF: FF: FF: FF: FF: @ MAC of interface beacon is from ETH_Pbeacon

Length TLV payload structure

6 6 2 2 nbyte

PayloadMAC header

T L VT L V ···

Figure 5 Beacon Packet Format



Figure 6 Session and broadcast keys exchange protocol

Send beacon (ID 2)

SK request (N1, B1, T1)

Init EAP SK (ID 2, dev)

ACK

SK success (N2, B2)

SK response (N1, B1, N2, B2, T2)

Calculate session key

Calculate session key

Store/update peer_hw info (SK, B2)

Insert new node/

generate LMSK

Check data validity

(info store in EAP SK )

Check data validity

(info store in EAP SK)

Node 1 Node 2

Store/update peer_hw info (SK, B1)

Insert new node/generate LMSK

Node ANode B

bt0

eth1

device

11:11:11:11:11:11

12:34:56:78:9A:BC

MAC

bt0

eth1

device

11:11:11:11:11:11

MAC address bt0

eth1

device

CB:A9

MAC

bt0

eth1

device

22:22:22:22:22:22

CB:A9:87:65:43:21

MAC address

fc01:1111: aaaafc01:1111: bbbb

GWGJ909bt022:22:22:22:22:22fc01:1111: bbbb

CCCC111eth133:33:33:33:33:33fc01:1111: cccc

454

L2

key

PNPNeth1CB:A9:87:65:43:21fc01:1111: bbbb

BCAST

key

dev@MAC@IP@IP

Node A Pn neighbors table

QTQT909bt011:11:11:11:11:11fc01:1111: aaaa

CCCC222eth133:33:33:33:33:33fc01:1111: cccc

454

L2

key

XYXYeth112:34:56:78:9A:BCfc01:1111: aaaa

BCAST

key

dev@MAC@IP @MAC@IP

Node B Pn neighbors table

Node C

fc01:1111: cccc

eth1

device

33:33:33:33:33:33

MAC address

222

111

L2

key

XYXYeth112:34:56:78:9A:BCfc01:1111: aaaa

PNPNeth111:11:11:11:11:11fc01:1111: bbbb

key

dev@MAC@IP BCAST@MAC@IP

Node C Pn neighbors table

Figure 7 P-PAN formation exampleneighbours tables



Symmetric encryption is done using AES crypto-graphic algorithm with a key length of 256 bits.

• Node 1 receives a beacon from Node 2

• Node 1 sends EAP_request / MAGNET_SK(E(LMSK1_2, N1 | B1 | T1))

• Node 2 replies with EAP_response /MAGNET_SK (E(LMSK1_2, N1 | B1 | N2 | B2 |T2))

• Node 1 sends EAP_success / MAGNET_SK(E(LMSK1_2, N2 | B2)

where LMSK1–2 (Link Master Session Key) is calcu-lated as HMAC_SHA_256 (KPN, “MAC1+MAC2”).

Use of the MAC addresses of the candidate radios inthe derivation function ensures that different pairs ofhardware adaptors of a radio subsystem share differ-ent link keys even for the same pair of devices. Thisis particularly relevant in the presence of detachablewireless interface adaptors (USB or card based).

The SK1-2 (link layer Session Key) is computed asHMAC SHA-256(g1-2, N1 | N2) and is valid for T2seconds (T2 ≤ T1). This procedure is run any time anew neighbour is discovered by a peer and wheneverthe derived session keys expire.

Besides the authentication process, upon the additionof a new entry, the layer 2 session key generation andexchange procedure will be triggered.

On a heterogeneous scenario like the depicted one theinformation necessary is stored in tables upon suc-cessful finalisation of the authentication and keyexchange procedure.

Finally, if the layer 2 session key exchange has suc-cessfully ended, the nodes’ broadcast keys (the onefor the air interface associated) will be exchanged soeach one will have the counterpart broadcast key inorder to use it for decryption.

Figure 7 shows an example of which information thenodes would manage after a P-PAN is formed.

For the authentication and layer 2 keys distribution,Extensible Authentication Protocol (EAP) primitivesare used to transport the information exchanged.Advanced Encryption Standard (AES) is used forencrypting the frames at the UCL. Additionally, asshown in Figure 8, packets are signed to enforce theintegrity of the information.

Authenticity, integrity and privacy are assured byusing this format. Together with the flat addressingscheme described in Section 4.3, this allows the UCLto filter personal traffic and prevent any imperson-ation attack. This is, if a packet comes from or isdirected to a personal IP address, it will only beallowed to pass through the UCL security checks ifsuch packet has been encrypted using a personal linksession key.

Once the aforementioned procedures have beenaccomplished all the communications between per-sonal nodes within the P-PAN are assured as longas each of the components of the network is trustedthrough them.

4.3 Intra P-PAN Communications

The self-configuration of the P-PAN starts with PNaddress autoconfiguration for which stateless auto-configuration will be used. To this end, it is specifiedthat the PN address consists of a concatenation of a40 bit PN prefix and a 64 bit Interface ID which ismapped from a MAC address using the IEEE EUI-64format [20].

As already introduced, the P-PAN will be an IP net-work composed by personal nodes that are able tocommunicate using different kinds of PAN radiotechnologies. On the other hand, Mobile Ad hoc Net-works (MANETs) offer the capacity of a self-config-uring and self-healing network able to deal with thedynamics inside the P-PAN.

The solution adopted is to use a proactive ad hocrouting algorithm which takes advantage of the bea-coning process used for neighbour detection for linkappearance and breakage and that is able to maintainan updated vision of the network in all the nodesbelonging to it. In this sense, a personal node willalways know which are the rest of the members of theP-PAN it is currently in.

Intra P-PAN communications is then enabled toembrace a heterogeneous multihop path as long asthe combination of the UCL and the ad hoc routingprotocol allows it.

Figure 8 Data PDU Format after UCL

DATA (IP+TCP/UDP+Data)HASH(CRC,

SHA-256, ..)

DATA (IP+TCP/UDP+data)

AES(SK)



5 PN Federation Formation and Use

In [19] the PN-F cycle was divided into two mainstages, the first one dealing with the management andcontrol, in which the definition and the participationof the PN-F is agreed, and the second one dealingwith the networking, in which the actual mechanismsfor securely interconnect devices belonging to differ-ent PNs take place.

This section will sketch how an extension of the pro-cedures specified for the P-PAN formation in Section3 can support the two steps that comprise the net-working stage of the PN-F cycle, namely PN-FFormation and PN-F Use, in the ad hoc case.

5.1 Trust Establishment

The imprinting procedure leverages a node to nodetrust that is supervised by the user. A similar app-roach can be followed in the case of PN-Fs in whichthe trust is established between PNs and both usersare involved. This is the first extension necessary.The primary keys exchanged in this case would beshared between all the nodes belonging to both PNs.This way, whenever two nodes that are members ofdifferent PNs have been paired, they will have a pre-shared secret that they can use to authenticate eachother, similarly to the way two personal nodes wouldauthenticate each other. These pre-shared keys wouldhave different characteristics from the personal keys.In this sense, while a personal primary key sharedbetween two personal nodes is only destroyable uponuser revocation, the key shared between two PNs canbe ephemeral and destroyable upon expiration or dueto misbehaviour of one of the users.

In the primary personal keys case, the nodes store thekeys using the Node ID as index. Then when the twonodes meet, the appropriate primary key is selectedbased on the Node ID information embedded in thebeacon packets (Figure 4). For the PN-PN case, thekeys should be stored based on a common PN Identifierthat all PN nodes would share. Additionally, this infor-mation should also be embedded within the beaconpackets. This is another extension needed to supportPN-Fs. Thus, mandatory fields for the bacon payload(Figure 4) are finally both the Node ID and the PN ID.

Following this approach, a node has N-1 primary keys(where N is the size of the PN). Each of them is associ-ated to each of the other N-1 personal nodes in the PN.Also, it would have 1 primary key per PN with whoma previous “imprinting” procedure has been carriedout. The same primary key is used to derive the ses-sion keys with all the nodes in the neighbouring PN.

Another possibility that would not require the usersto meet beforehand in order to exchange the primary

keys is the use of a typical PKI structure such thateach PN Identifier is signed by a CertificationAuthority (CA) that is trustable for both users. In thiscase, whenever two nodes from different PNs meet,they firstly would have to exchange a shared secretbased on the information stored in each other’s cer-tificates. Then, the procedure would be the same asif the secret would have been shared using a pairingprocedure as the one used for personal nodes.

It is important to note that this trust establishmentis only meant to enable the authentication of nodes.Further authorization to make use of the services pro-vided by the user PN should be based on this authen-tication. Besides, the solutions proposed enable nodeauthentication but if really sensitive informationmight be disclosed, user authentication should beput on top.

5.2 Mesh Connectivity Establishment

Taking into account the extensions described in theprevious section, it is easy to see that in the ad hoccase of a PN-F, i.e. when several P-PANs cometogether, the result at connectivity level will be a fullsecure mesh. Making use of the same mechanismsused for authentication, session key exchange andencryption in the P-PAN formation, nodes belongingto different PNs that are in the same radio domainwill be able to authenticate each other and exchangethe link session keys to be used afterwards. The maindifference is that whenever a beacon arrives, the firstfield to be evaluated would be the PN Identifier. If itindicates that the beacon comes from a personal node,then the Node ID field would be necessary forrequesting the appropriate primary key. Conversely,if it comes from another PN member, the PN Identi-fier would be the one used for selecting the appropri-ate primary key to be used in the subsequent phases.

The data is sent using the same security enforcementmechanisms (see Figure 8) independently of who isthe recipient of the packet (a personal or a foreignone), since there is no reason not to protect the com-munications with others in the same way as you pro-tect your internal communications.

5.3 Network overlay establishment

Once we have established a connectivity level thatprevents impersonation and supports privacy,integrity and authenticity, a network overlay can beset down. This network overlay includes the nodesfrom the different P-PANs forming the PN-F.

The first extension to be included in order to supportthe formation of a PN-F is the specification of anaddressing scheme. In this case, the proposed solutionis to have a dedicated addressing space for each PN-F



Figure 9 Ad hoc network overlay establishment forPN-F formation

a PN participates in. Similarly to the case of Intra P-PAN communications, the flat addressing does notonly ease the routing procedures but is also used toprevent impersonation and to filter down the trafficthat should not be allowed. In this sense, upon thereception of a packet at the edge of one of the in-volved P-PANs, the gateway node can, after decrypt-ing and assuring the integrity of the incoming packet,check the IP header and decide whether to allow thepacket to pass or not, depending on whether thepacket comes from a PN that is a member of thisfederation or not.

Similarly to the P-PAN formation, the network over-lay establishment will be based on the deployment ofa MANET like ad hoc routing that would enable theend-to-end path discovery and maintenance betweenPN-F members belonging to different P-PANs. Forthe P-PAN case, a proactive routing protocol waschosen due to the requirement of quickly and alwaysready networking. Nevertheless, the use of a pro-active routing protocol for the PN-F case would betoo costly, mainly taking into account the possiblelarge size that an ad hoc PN-F could have (e.g. meet-ing room, conference, etc.). In this case, the use of areactive routing protocol seems more appropriate. Theextension required is then two-fold: on the one handto implement the reactive routing protocol and on theother hand to develop the mechanism at the networklevel of P-PAN nodes such that whenever a packet isissued to another personal node, the proactive routingis used, whereas the reactive one is called wheneverthe packet is destined to a PN-F node. The use ofdifferent addressing spaces enables this filtering.

Nodes will have multiple network identifiers (i.e. IPaddresses) that they will use whenever they want toaccess a specific service provided under the auspicesof a particular PN-F. In the ad hoc case of PN-F anetwork overlay will be established such that P-PANsof the PN-F members will form a network identifiedby a pre-established addressing space.

6 Conclusions and future work

Personal Networking is a promising research and eco-nomic field that imposes a big set of challenges thathave to be overcome before it could be fully opera-tional. This paper has described the mechanisms thatenable the self-configuration of a Private PersonalArea Network around the user overcoming the het-erogeneity while reinforcing the required security.Besides, it has been shown how these mechanismsonly require minor extensions to support the estab-lishment and use of PN Federations whenever differ-ent persons with their P-PANs meet together.

Contrary to other descriptions of cluster or PersonalArea Network [4], [8] that limit the concept to a mat-ter of radio coverage (e.g. 10 m range), the conceptof cluster proposed in this architecture stands on anopportunistic, distributed, multihop and proactiveapproach based on the trust relationships establishedbetween the cluster constituents. Further, it copeswith the heterogeneity support, dynamic adaptation,infrastructureless environment survival and privacyrequirements imposed by the P-PAN concept. Theclusters will be as large as possible (as long as a newpersonal node or device is reachable through a PANair interface, the cluster will add a new wireless hopto its structure), adding new personal nodes anddevices as soon as they appear in the cluster sur-roundings.

The mechanisms described in this paper focuses onthe connectivity and network level of the PN archi-tecture [21]. Further procedures at service level areneeded for a full-blown Personal Network and Per-sonal Network Federation, but they were out of thescope of this paper.

Additionally, no mention of air interfaces has beenincluded into the paper since, as already mentioned,the architecture and solutions proposed have beendesigned in order to support the heterogeneity interms of radio interfaces that nowadays exists in thewireless communications area. The Personal Net-working paradigm extends the short-range communi-cations concept that is followed in typical PANapproaches. Unlike PANs, with a limited geographi-cally coverage, PNs have an unrestricted geographi-cal span, and may incorporate devices into the per-sonal environment regardless of their geographiclocation. As already introduced the final aim of thePNs is to be an enabler for the ubiquitous computingstanding on a strong user focus.

It is also important to note that a proof-of-conceptimplementation has been successfully integratedand used for both the validation of the techniquesdesigned and the analysis of its conformance to the

P-PAN1 P-PAN2



personal networking system specifications. Thedescription of this implementation can be found in[22].

Finally, and taking the aforementioned proof-of-con-cept implementation as a baseline, future work isfocused on building a set of Pilot Services on top ofa PN platform. The envisaged Pilot Services offer aframework to assess usability and acceptance of PNservices and provide user feedback to the researchactivities. Pilot services will help the project assessthe market potential of the overall PN architectureand services, produce exploitation plans and build thebusiness. The feedback from the pilot will be invalu-able for the research activities to further guide theresearch and pursue the long term MAGNET vision.

Acknowledgements

This paper describes work undertaken in the contextof the IST-FP6-IP-027396 ‘My personal AdaptiveGlobal Net and Beyond’ IST-MAGNET Beyond pro-ject. MAGNET Beyond is a worldwide R&D projectwithin Mobile & Wireless Communication beyond3G. MAGNET Beyond will introduce new technolo-gies, systems and applications that are at the sametime user-centric and secure. MAGNET Beyond willdevelop user-centric business model concepts forsecure Personal Networks in multi-network, multi-device, and multi-user environments. Please visitwww.ist-magnet.org

The authors would like to acknowledge the collabora-tion of their colleagues from the MAGNET BeyondConsortium.

References

1 Niemegeers, I G, Heemstra de Groot, S. FromPersonal Area Networks to Personal Networks: Auser oriented approach. Journal on Wireless andPersonal Communications, 22, 175–186, 2002.

2 Niemegeers, I, Heemstra de Groot, S. Personalnetworks: Ad hoc distributed personal environ-ments. Med-HocNet, IFIP Conference on Ad-HocNetworks, September 2002.

3 FP6-IST-IP-507102. My personal AdaptiveGlobal Net. IST-MAGNET project. 2006, Octo-ber 31 [online] – URL: www.ist-magnet.org

4 Gustafsson, E, Jonsson, A. Always best con-nected. IEEE Wireless Communications, 10 (1),49–55, 2003.

5 Power Aware Communications for Wireless Opti-Mised personal Area Network, PACWOMAN.IST-2001-34157. 2006, October 31 [online] –URL: http://www.imec.be/pacwoman

6 Muñoz, L, Agüero, R, Choque, J, Irastorza, J A,Sánchez, L, Petrova, M, Mähönen, P. Empower-ing Next-Generation Wireless Personal Commu-nication Networks. IEEE Communications Maga-zine, 42 (5), 64–70, 2004.

7 IEEE 802.15 Working Group for WPAN. 2006,October 31 [online] – http://www.ieee802.org/15/

8 IETF Mobile Ad hoc NETworks (MANET) work-ing group. 2006, October 31 [online] –http://www.ietf.org/html.charters/manet-charter.html

9 IETF Zero Configuration Networking (Zeroconf)working group. 2006, October 31 [online] –http://www.zeroconf.org/

10 UPnP™ Forum. 2006, October 31 [online] –www.upnp.org

11 Shivers, O. BodyTalk and the BodyNet: A Per-sonal Information Infrastructure, Personal Infor-mation Architecture Note 1. Cambridge, MA,MIT Laboratory for Computer Science, Decem-ber 1, 1993.

12 Zimmerman, T G, Smith, J R, Paradiso, J A, All-port, D, Gershenfeld, N. Applying Electric FieldSensing to Human-Computer Interfaces. CHI’95Human Factors in Computing Systems, Denver,May 9–11, 1995, ACM Press, New York.

13 Zimmerman, T G. Personal Area Networks(PAN): Near-Field Intra-Body Communication.Cambridge, MA, MIT Media Laboratory,September 1995. (M.S. thesis)

14 Austin, T. PKI. John Wiley, 2001. (ISBN 0-471-35380-9)

15 Liberty Alliance Project. 2006, October 31[online] – http://www.projectliberty.org/

16 Zimmerman, P. PGP Source Code and Internals.The MIT Press, 1995. (ISBN 0-262-24039-4)

17 IST-507102 – MAGNET. Final version of theNetwork-Level Security ArchitectureSpecification. Deliverable D4.3.2, March 2005.



18 Sanchez, L, Lanza, J, Muñoz, L, Perez Vila, J.Enabling Secure Communications over Heteroge-neous Air Interfaces: Building Private PersonalArea Networks. 8th International Symposium onWireless Personal Multimedia Communications,Aalborg, Denmark, September 2005.

19 Hoebeke, J, Holderbeke, G, Moerman, I, Jacobs-son, M, Prasad, V, Wangi, N I C, Niemegeers, I,Heemstra De Groot, S. Personal Network Federa-tions. In: Proceedings of the 15th IST Mobile &Wireless Summit Communications Summit,Mykonos, Greece, June 2006.

Luis Sanchez received the Telecommunications Engineering Degree by the Telecommunications Engineer-

ing School of Santander, University of Cantabria (UC), Spain, in 2002. Since 2001 he has been a

researcher at the Communications Engineering Department at that university, where he is also pursuing

his PhD. He has participated in several international research projects (e.g. WINE, 6HOP, PACWOMAN,

MAGNET, MAGNET Beyond) corresponding to the 5th and 6th Framework Programme of the IST initiative.

His research interests are focused on: 1) Personal Networking; 2) Heterogeneous Mobile Networks; and

3) Ad hoc networks.


Jorge Lanza received a degree in Telecommunications Engineering from the University of Cantabria (UC),

Spain, in 2000. Since then he has been a researcher at the Data Transmission and Mobile Networks group

of that university, where he is currently working toward a PhD in communications engineering. His research

activities focus on ad hoc networks over wireless technologies, especially putting emphasis on protocol

design and performance analysis of TCP/IP protocols over real test-beds, such as muti-hop wireless

environments. As a member of the Technical Observatory for Smart Cards at the University of Cantabria

(OTTIUC), he also works with highly regarded manufacturers, banking entities and mobile operators

acquiring experience in smartcard technology, highlighting a patent request concerning security and user

authentication mechanisms through mobile phones. Current research in combined mobility and security for

the wireless Internet is being carried out based on the merging of wireless technologies and smartcards for

current and next generation networks.


Luis Muñoz is Professor at the University of Cantabria. He received the Telecommunications Engineering

Degree by the Telecommunications Engineering School of Barcelona, Polytechnical University of Cataluña

(UPC), Spain, in 1990 and the PhD also by the UPC in 1995. He joined the University of Cantabria in 1990

first as Assistant Professor of the Electronics Department and from 1996 as Lecturer of the Communica-

tions Engineering Department. He is head of the Data Transmission and Mobile Networks group belonging

to DICOM. He started to work in the field of Data Transmission and Mobile Networks in 1990, first in topics

related to modulation, equalisation techniques and channel coding; later in mobile networks with voice and

data integration, designing and carrying out projects as TETRA for power utilities, security systems and

telecontrol with real time needs. He has participated in projects of the 4th Framework of the EU R&D

Programme, such as ACTS and at present he is participating in the 5th Framework IST. His group has

strong relations with the Spanish Telecom operators and manufacturers companies belonging to these

sectors. In parallel to this activity Dr. Luis Muñoz serves as consultant of different companies.


20 Thomson, S, Narter, T. IPv6 Stateless AddressAutoconfiguration. IETF RFC 2462. URL –http://www.ietf.org/rfc/rfc2462.txt, December1998.

21 IST-507102 – MAGNET. Overall secure PNarchitecture. Deliverable D.2.1.2, October 2005.

22 Hoebeke, J, Holderbeke, G, Moerman, I, Louati,W, Girod Genet, M, Zeghlache, D, Sanchez, L,Lanza, J, Alutoin, M, Ahola, K, Lehtonen, S, JaenPallares, J. Personal Networks: from concept toa demonstrator. In: Proceedings of the 15th ISTMobile & Wireless Summit CommunicationsSummit, Mykonos, Greece, June 2006.


1 Introduction

The next generation of wireless communications sys-tems, commonly known as fourth-generation (4G)network [1], is envisaged to encompass a multitude ofcellular and wireless networking technologies whichinclude Wireless Personal Area Network (WPAN),Wireless Local Area Network (WLAN) and third-generation (3G) cellular network. These wireless net-working technologies are seamlessly interconnectedby the Internet Protocol (IP) backbone network. Inessence, 4G aims to transform communications archi-tectures from traditional vertical stovepiped to hori-zontal integrated systems [1]. Personal Networks [2]are one such network architecture that can fulfill theaim of 4G with user-centric perspectives. It is adynamic network building on the above mentionedwireless networking technologies, to facilitate per-sonalized ubiquitous communications anywhere atanytime. Figure 1 shows the network architecture of apersonal network which begins from a WPAN bubblethat can be expanded or shrunk depending on theuser’s demands and environment. The WPAN ex-pansion can physically be made via interconnectingstructures, e.g. Universal Mobile TelecommunicationsSystem (UMTS) [3-4] and the Internet, to remotenetworks such as home area networks, corporate areanetworks or vehicular area networks. A WPAN is anetwork of devices which could consist of a mobilephone, a PDA, a notebook PC, a digital camera, etc.All or a parts of these devices are carried around bya person in everyday life for both work and pleasure.

This paper considers the first step toward buildinga personal network by enabling the co-operationbetween the UMTS, and WPAN and WLAN tech-nologies. This co-operation poses a new set of prob-lems. Current cellular and wireless networking tech-nologies consider terminals only in isolation. In per-sonal networks, we no longer have single terminals,but a very dynamic WPAN wanting to establish co-operation with UMTS so that it can connect with

remote devices or remote WPANs. That means,current technologies are insufficient or have to beenhanced to accommodate new requirements. Themajor issues that need to be addressed are self-orga-nization, establishing and maintaining quality of ser-vice for particular applications, routing and mobilitymanagement. The work presented in this paper willaddress all of these issues. Firstly, the state-of-the-arttechnologies are evaluated in view of building a per-sonal network. We will point out the limitations withthe current state-of-the-art technologies. Then, wepropose solutions to these limitations within therealm of an interconnecting architecture for personalnetwork. Finally, we present a number of ongoingkey projects related to personal networks.

2 State-of-the-art Wireless

Technologies

In this section, we briefly describe the state-of-the-artwireless technologies that are suitable for buildingpersonal networks, namely, WPANs, WLANs andUMTS.

2.1 Wireless Personal Area Networks

A WPAN is a short-range (typically, transmissionrange is limited to 10 m), low-cost and low-powerconsumption technology. Unlike UMTS, WPANoperates in the unlicensed Industrial, Scientific andMedical (ISM) frequency band at 2.4 GHz. The IEEE802.15 working group is standardizing different ver-sions of WPAN:

• IEEE 802.15.1 (Bluetooth) [5]• IEEE 802.15.3 [6]

2.1.1 IEEE 802.15.1 (Bluetooth)

The Bluetooth specification has been made the IEEE802.15.1 standard [5]. Hence, Bluetooth and IEEE802.15.1 are synonymous. Throughout this paper,we use the term Bluetooth. Two or more Bluetooth

Personal Networks – An Architecture for 4G Mobile

Communications Networks

A N T H O N Y L O , W E I D O N G L U , M A R T I N J A C O B S S O N , V E N K A T E S H A P R A S A D ,

I G N A S N I E M E G E E R S

Anthony Lo is

Assistant Pro-

fessor at Delft

University of

Technology,

The Netherlands

Weidong Lu is a

PhD student at

Delft University

of Technology,

The Netherlands

Martin Jacobsson

is a PhD student

at Delft Univer-

sity of Tech-

nology, The

Netherlands

Venkatesha

Prasad is a

Research

Scholar at Delft

University of

Technology,

The Netherlands

A personal network is a network architecture which builds on various wireless networking technologies.

It is responsible for glueing these wireless networking technologies to serve users. This paper considers

the co-operation of several key technologies to realize a personal network; namely Wireless Personal

Area Network (WPAN), Wireless Local Area Network (WLAN) and Universal Mobile Telecommunica-

tions System (UMTS). This co-operation poses a new set of problems as these technologies were not

designed to interwork with each other. In this paper, we present an architectural framework on which

one can build a personal network using these wireless technologies. We also discuss each of these

problems and propose solutions toward building a personal network demonstrator.



devices sharing the same frequency-hopping se-quence form a piconet, which is a star topology. Thesmallest unit of a WPAN is known as piconet. Withina piconet, a Bluetooth device can play either one ofthe two roles: master or slave. Each piconet may onlycontain one master device and up to seven slavedevices. Communication in a piconet is organized insuch a way that the master device polls each slaveaccording to a certain polling algorithm. A slavedevice is only allowed to transmit after being polledby the master device as depicted in Figure 2. Differ-ent piconets employ different frequency-hopping

sequences to prevent mutual interferences. Bluetoothoffers gross bit rates of up to 3 Mb/s.

Bluetooth defines not only a radio interface, but awhole communications stack that allows devices tofind each other and advertise the services they offer.The core Bluetooth protocol stack, which consists ofLayer 1 and 2, is illustrated in Figure 5. BluetoothNetwork Encapsulation Protocol (BNEP) provides anEthernet-like interface to the upper layer. Communi-cations at the Logical Link Control and AdaptationProtocol (L2CAP) layer in a piconet can only be

Figure 1 Personal Network

Figure 2 Bluetooth Piconet Architecture Figure 3 IEEE 802.15.3 Piconet Architecture

Ignas

Niemegeers is

Professor at

Delft University

of Technology,

The Netherlands

Home area network

Vehicle area network

Interconnecting structures

Corporate area network

Wireless personal area network

Device

Node B Node B

UMTS

Internet

Slave Slave

Slave

Master

Slave

data

data

data

data

control

control

control

control

PNC

data

data

data

data

data

control

control

control

control



between the master device and a slave device. Themaster device acts as an Ethernet bridge at the BNEPlayer forwarding packets that are not destined foritself.

2.1.2 IEEE 802.15.3 High Data Rate WPAN

Unlike Bluetooth, IEEE 802.15.3 [6] offers highgross bit rates of up to 55 Mb/s. An IEEE 802.15.3piconet is a distributed topology with a central con-troller known as Piconet Controller (PNC). The PNCdiffers from the Bluetooth master in that it is respon-sible for scheduling the communication between thedevices but the data traffic may not pass through thePNC. That means, the devices in the piconet cancommunicate on a peer-to-peer basis as shown inFigure 3. Each piconet may only contain one PNCdevice and up to 255 slave devices.

The IEEE 802.15.3 standard only defines Layers 1and 2, namely, the Physical layer and the MediumAccess Control layer as depicted by the center blockdiagram in Figure 5.

2.2 Wireless Local Area Networks

Currently, IEEE 802.11 [7] is the most mature andwidely deployed WLAN technology. It also operatesin the unlicensed frequency band of 2.4 GHz. TheIEEE 802.11 standard defines two modes, namely,infrastructure and ad hoc. In the former mode, theIEEE 802.11 devices form a star topology with anaccess point as the central controller. For non-real-time services, the devices communicate with theaccess point through a random access technique whilea polling scheme is used for real-time services. In thead hoc mode, the devices communicate with eachother directly on a peer-to-peer basis as shown inFigure 4. The IEEE 802.11 standard only definesLayers 1 and 2 as shown in the right block diagramof Figure 5.

Each of the above-mentioned state-of-the-art wirelesstechnologies provides its own mechanism for WPANformation or self-organization. In personal networks,a WPAN consists of heterogeneous devices. In thiscase, the WPAN formation mechanism defined ineach wireless technology is insufficient. However,in the next section, we will describe the design ofa personal network gateway architecture which canbe used in the formation of such a heterogeneousWPAN.

2.3 UMTS

The UMTS network architecture [3-4] is depicted inFigure 6 which consists of a User Equipment (UE)(the UMTS term for mobile station) and two indepen-dent land-based network segments: the UMTS Ter-restrial Radio Access Network (UTRAN) and thecore network. The latter is composed of the ServingGPRS Support Node (SGSN) and Gateway GPRSSupport Node (GGSN) which are interconnected viaan IP network. The SGSN keeps track of the locationof individual mobile stations and performs securityfunctions and access control. The GGSN encapsulatespackets received from external IP networks androutes them toward the SGSN. The UTRAN consistsof the Radio Network Controller (RNC) and Node B(i.e. the base station) connected by an asynchronoustransfer mode network. The RNC is in charge of theoverall control of the logical resources provided byNode Bs. A UE communicates with the Node Bthrough a radio interface based on Wideband CodeDivision Multiple access (WCDMA) technology. The

Figure 4 IEEE 802.11 Ad Hoc Mode NetworkArchitecture

Figure 5 Protocol Architecture of Bluetooth, IEEE 802.15.3 andIEEE 802.11

Figure 6 UMTS Network Architecture

data

data

datadata

data

data

BNEP

L2CAP

802.15.3 MAC

LMP

Baseband

Radio

802.15.3

Physical

802.11

Physical

802.11 MAC

Bluetooth IEEE 802.15.3 IEEE 802.11

Layer 2

Layer 1

Internet

UTRAN Core Network

Node BRNC GGSNSGSNMTTE

R

User equipment

Node B



UE in turn consists of two disjoint entities, namely,the Terminal Equipment (TE) and the Mobile Termi-nal (MT). The TE and MT entities can reside in dif-ferent physical modules interconnected by the R-ref-erence point. The TE hosts the application and user-interaction, while the MT is in charge of all theUMTS communications-related tasks. Figure 7 showsthe protocol stacks of UMTS, which comprise theuser plane and the control plane. The user plane con-sists of a layered protocol structure providing userinformation transfer, along with associated informa-tion transfer control procedures. The control planeconsists of protocols for control and support of userplane functions.

Currently, the notion of WPAN, which comprises agroup of TEs associated with a single MT, does notexist in the UMTS standards. What has been definedin the standard is a protocol for point-to-point com-munication between an MT and a TE over a serialphysical link which can be a cable. Point-to-PointProtocol (PPP) is used to establish such communica-tions between MT and TE, where the MT serves as amodem. However, in our context, we have a numberof TEs which are grouped into a Bluetooth or anIEEE 802.15.3 WPAN, and the MT functions as apersonal network gateway which will be described in

the next section. Instead of using PPP for communi-cation, the group of TEs and the MT are networkedusing Bluetooth or IEEE 802.15.3.

3 Personal Network Architecture

A key component in the WPAN is the Personal Net-work Gateway (PNG) which seamlessly connects aWPAN or a WLAN to UMTS as shown in Figure 8.Unlike other devices in the WPAN, the PNG is multi-modal, i.e. it contains different protocol stacks. Onthe WPAN/WLAN side, it houses the Layer 1 andLayer 2 of Bluetooth, IEEE 802.15.3, and IEEE802.11, and on the other side it is the UMTS userplane and control plane radio access protocol stackas depicted in Figure 9. The UMTS user plane isconnected to the Bluetooth, the IEEE 802.15.3 andthe IEEE 802.11 at the IP layer. The PNG can alsoassume the role of master or PNC in Bluetooth or802.15.3, respectively.

In addition to providing UMTS connectivity, the PNGalso facilitates communication between devices whichare equipped with different technologies. For exam-ple, a Bluetooth-enabled device can communicatewith an 802.11-enabled device via the PNG. In thispaper, we assume that there is only one PNG in

Figure 7 UMTS Protocol Architecture

IP

Upper layers

IPv4 or IPv6

PPP

L1

TE MT Node B RNC SGSN GGSN

PPP

L1

PDCP

RLC

MAC

UMTS

PHY

UMTS

PHY L1

L2

L1

L2

L1

L2

PDCP

RLC

MAC

GTP-U

UDP

L1

L2

UDP

L1

L2

UDP

GTP-U GTP-U

L1

L2

UDP

Ra) User plane

UMTS

PHY L1

L2

GMM/SM

RLC

MAC

RRC

UMTS

PHY

PPP

L1

GMM/SM

RANAP

RLC SSCP

MTP3

L1

L2

L1

L2

L1

L2

L1

L2

MTP3

UDPSSCP

GTP-O

RRC RANAP

MAC

TE MT Node B RNC SGSN GGSN

b) Control plane

User-level IP Transport-level IP

L1

PPP

IP IP IP IP

IPv4 or IPv6

IP

L1

L2

IP

UDP

GTP-O



a WPAN or WLAN and the rest of the devices in theWPAN or WLAN are single-mode, i.e. either Blue-tooth, 802.15.3 or 802.11. In the scope of this paper,a user’s WPAN is not limited to Bluetooth- and802.15.3-enabled devices but may also include802.11-enabled devices due to the popularity of theWLAN technology. The PNG operates as a bridgein order to link Bluetooth-, 802.15.3- and 802.11-enabled devices together. Using a bridge, the differentdevices appear to be connected on the same subnet-work. Bridging is recommended because it providesefficient communication in a small sized network suchas WPAN. If two communicating devices are morethan one hop away, then multi-hop communication isutilized. This will be described in subsection 3.3.1.

3.1 Self-Organization

WPAN is a self-organized ad hoc network which isautomatically formed with little or no user interven-tion. The WPAN formation mechanism is providedby Layer 1 and Layer 2 of the WPAN. The WPANformation mechanism of Bluetooth and IEEE 802.15.3is provided by the Inquiry and the Association proce-dures, respectively. For the IEEE 802.11 ad hocmode, any device within radio range can be directlyaddressed without forming a subnetwork by using theScan procedure. The formation of a WPAN, whichcomprises devices of different wireless technologies,will be coordinated by the PNG using the WPAN for-mation mechanism of its wireless technology becauseit has multiple interfaces. The procedure is describedin subsection 3.1.1.

Once the WPAN is formed, it can operate either as astand-alone ad hoc network or as a subnetwork of theinterconnecting structure. In the latter, the PNG actsas the gateway and provides seamless connectivity tothe UMTS network which is in turn connected to theInternet. Before any device in the WPAN can sendand receive traffic from the interconnecting structure,it must be able to obtain a valid IP address and con-figure the PNG as the default router.

3.1.1 Personal Network Gateway Discovery

The PNG provides connectivity to the interconnect-ing structure, and therefore also needs to acquireunique IPv4 or IPv6 addresses for the WPANdevices. In order to use the PNG, devices in a WPANmust be able to find it even if the PNG is several hopsaway from the devices. Hence, the PNG discoverymechanism must also facilitate route constructionbetween the device and the PNG in the event of amulti-hop scenario. The PNG discovery can be real-ized proactively or reactively. In the latter approach,the PNG discovery is triggered by a WPAN device,while the former is initiated by the PNG. To leveragethe advantages of the two approaches, a hybrid

approach is appropriate for PNG discovery. The PNGdiscovery protocol comprises two messages: PNGAdvertisement and PNG Solicitation.PNG Advertisements are periodically broadcast intothe WPAN by the PNG. The period between two con-secutive PNG advertisements must be set to an opti-mum value so that the WPAN is not flooded to avoidwasting WPAN radio resources. If a device in aWPAN wants to learn about the PNG immediately,it can broadcast a PNG Solicitation which triggersimmediate PNG Advertisements. These two mes-sages could be defined as new Internet ControlMessage Protocol (ICMP) message types.

3.1.2 Address Auto-configuration

Two addressing schemes can be envisaged here con-sidering both IPv4 and IPv6. We call the first of thesetwo schemes “Private Address Auto-configuration”and the second one “Global Address Auto-configura-tion”.

Figure 8 Personal Network Gateway

Figure 9 Personal Network Gateway Protocol Architecture

Node B

PNG

WPAN

UMTS

network

UMTS

user plane

UMTS

control planeWLAN

WPAN

IP IP

WPAN

layer 2

WPAN

layer 1

PDCP

RLC

MAC

PHY

PNG

discovery

GMM/SM

RRC



Private Address Auto-configuration

In this scheme, the WPAN is assigned a singleglobally unique IP address. That means, the WPANappears as a single point of presence on the Internet.The globally unique IP address is allocated to thePNG and the rest of the devices in the WPAN areassigned with private IP addresses which are notglobally unique. Hence, a Network Address Trans-lator (NAT) [8], which is located in the PNG, isemployed to translate private IP addresses to theglobal routable IP address for packets emanatingfrom any device in the WPAN, and vice versa. BothIPv4 and IPv6 support auto-generation of privateaddresses, viz. RFC 3927 [9] and RFC 2462 [10],respectively. The PNG can obtain a globally uniqueIP address through the UMTS Packet Data Protocol(PDP) context activation which is defined in the Ses-sion Management (SM) layer of the UMTS controlplane. The PDP context can be viewed as a recordthat holds parameters that characterize a certainconnection. In other words, it is a virtual connectionbetween the PNG and the GGSN, which is character-ized by the IP address and the quality of service pro-file. The PDP context is stored in the PNG, the SGSNand the GGSN. With an active PDP context, the PNGis visible to the GGSN and it can send and receivedata packets. Figure 10 illustrates the steps involvedin the PDP context activation.

Step 1: The PNG generates the Activate PDPContext Request message and sends it to SGSN.

Step 2: The SGSN checks the Activate PDPContext Request message and generates theCreate PDP Context Request messagewhich is sent to the GGSN to establish a GTP tunnelbetween the SGSN and the GGSN. The tunnel is usedas the packet routing path between the GGSN and theSGSN.

Step 3: The GGSN allocates an IP address for thePNG, which is carried by the Create PDP Con-text Response message.

Step 4: Finally, the SGSN informs the PNG of theallocated IP address through the Activate PDPContext Accept message.

Global Address Auto-configuration

In this scheme, each device in the WPAN is allocateda globally unique IP address. Therefore, each WPANdevice is visible to external nodes on the Internet. Inthis case, the PNG does not need to perform addresstranslation for the WPAN devices. This addressingscheme does not favor IPv4 because of limited IPv4address space. For allocating globally unique IPaddresses, IPv4 supports only stateful address alloca-tion technique, while IPv6 defines two techniques,namely stateless address allocation and statefuladdress allocation.

For the stateful address allocation technique, theoperation is similar to the first scheme except that thePNG needs to perform the PDP context activation foreach device in the WPAN in order to get a globallyunique IP address. The PNG also needs to performPDP context activation to obtain an IP address for theUMTS interface. In IP-based networks, the statefuladdress allocation can be accomplished by means ofDynamic Host Configuration Protocol (DHCP).However, DHCP has been designed with the assump-tion that the DHCP client and server is one hop away.

Figure 10 PDP Context Activation Procedure

Figure 11 IPv6 Stateful Address Allocation

PNG UTRAN SGSN GGSN

1. Activate PDF context request

2. Activate PDF context accept

3. Create PDF context request

4. Create PDF context response

PNG UTRAN SGSN GGSN

2. Activate PDF context request

5. Activate PDF context accept

3. Create PDF context request

4. Create PDF context response

WPAN device

6. PNG advertisement

1. PNG solicitation



In WPAN, the DHCP client and the server may bemultiple hops away. Hence, DHCP is not suitable forproviding stateful address configuration in a multi-hop network. The PNG messages should be designedto support the stateful address configuration option.Figure 11 illustrates the operation of stateful addressconfiguration. As shown in step 1 of the figure, thePNG Solicitation is used by the device in theWPAN to request for an IP address, which in turntriggers the PNG to perform PDP context activation(step 2 to step 5). Finally, the IP address allocatedduring PDP context activation is conveyed to the WPANdevice via the PNG Advertisement message.

As mentioned, the stateless address allocation tech-nique is only applicable to IPv6 since no equivalenttechnique exists in IPv4 for generating globallyunique IP addresses. For this technique, each devicein the WPAN is able to generate its own IPv6 addressby concatenating a subnet prefix with an interfaceidentifier. In IP-based networks, such a subnet prefixis contained in the IPv6 Router Advertisement mes-sages which are transmitted by an IPv6 router. Simi-larly to DHCP, IPv6 Router Advertisement is notdesigned for multi-hop networks. The PNG messagesshould also support stateless address configuration inaddition to stateful address configuration. The PNGcan be configured to support either the stateful or thestateless address configuration. The interface identi-fier can be a Bluetooth, an IEEE 802.15.3 or an IEEE802.11 MAC address. The stateless address allocationis depicted in Figure 12. The WPAN device is respon-sible for ensuring that the interface identifier isunique by broadcasting an IPv6 neighbor soli-citation message to perform duplicate addressdetection (step 1 of Figure 12). However, the dupli-cate address detection cannot be used unchanged ina multi-hop scenario because the message can onlyreach devices one hop away. The duplicate addressdetection in multi-hop scenario is also present inmobile ad hoc networks. Several mechanisms havebeen proposed [11]. Therefore, these proposed mech-

anisms can be used by personal networks. If noaddress duplication is detected, then the WPANdevice will use the interface identifier and issue aPNG Solicitation message (step 2). In order toget the IPv6 subnet prefix for the entire WPAN, thePNG performs the PDP context activation (step 3).Once the PDP context activation is completed, theGGSN can send an IPv6 router advertise-ment message (step 5) on the newly established PDPcontext. Alternatively, the PNG may issue an IPv6router Solicitation message to GGSN (step4). After the PNG receives the router advertisementmessage, it broadcasts the PNG Advertisementmessage to the devices in the WPAN (step 6). At thesame time, the PNG constructs its IPv6 address byconcatenating a randomly generated interface identi-fier and the subnet prefix. The interface identifier israndomly generated because the UMTS networkinterface does not have an equivalent IEEE MACaddress. The WPAN device also generates its IPv6address by concatenating its interface identifier andthe subnet prefix.

3.2 Quality of Service

Quality of Service (QoS) is defined as a set of servicecharacteristics that the network is requested to meetwhen transporting a sequence of data packets. Theservice characteristics can be expressed in terms ofthroughput, delay, loss, bit error rate, or as a relativepriority of access to the network. End-to-end QoS inpersonal network spans across different domains:WPAN, UMTS and IP QoS-enabled interconnectingstructures such as the future Internet. In such a hetero-geneous environment, the end-to-end QoS will relyon the coordination of QoS mechanisms in differentdomains along the end-to-end communication path.Each of these domains (Bluetooth, IEEE 802.15.3 orUMTS) has its own QoS provisioning mechanisms.The challenge in QoS provisioning in personal net-work is to seamlessly interwork the QoS mechanismin each domain. The QoS provisioning functionalityof each domain is identified and interworked as

Figure 12 IPv6 Stateless Address Allocation

PNG UTRAN SGSN GGSNWPAN/WLANdevice

3. PDP context activation procedure

1. Neighbor solicitation

2. PNG solicitation

4. Router solicitation

5. Router advertisement

6. PNG advertisement



shown in Figure 13. The interworking between QoSfunctionality takes place at the PNG and the UMTSGGSN. The PNG provides interworking between theWPAN QoS functionality and the UMTS QoS func-tionality while the GGSN deals with interworkingbetween the UMTS QoS functionality and the QoSfunctionality in external IP networks.

The QoS management modules for UMTS are speci-fied in [12-13]. UMTS achieves QoS managementusing a layered architecture with bearer servicesestablished at different layers between UMTS QoSmanagement modules. The QoS management inWPAN is responsible for setting up a bearer (or con-nection) according to the requested QoS parameters.

The application in a WPAN device triggers therequest for network service with particular QoSrequirements. The application QoS requirements aresent to the IP QoS manager of the WPAN device foran IP level service. The IP QoS manager translatesthe QoS parameters for the WPAN QoS manager,which sets up the bearer with the required QoSbetween the device and the PNG. The WPAN QoSmanager comprises admission control and schedulingalgorithms. The admission control takes care of theradio resource allocation based on the availabilitywhile the scheduling algorithm schedules data trans-mission according to the QoS requirements. Thesame application QoS requirements, which arereceived by the WPAN QoS manager at PNG, aresignaled to the IP bearer service manager. This inturn sets up the UMTS bearer service by initiatingthe PDP context activation with the required QoS.The PDP context which contains the application QoSrequirements is translated for the IP bearer servicemanager by the UMTS bearer service manager atthe GGSN. The IP bearer service manager uses therequirements for controlling the QoS with externalIP networks.

For QoS control with an external IP network, UMTSspecifies the use of DiffServ at the IP bearer servicemanager in the GGSN. For the QoS signalingbetween WPAN and UMTS, a number of QoS signal-ing protocols, e.g. Resource reSerVation Protocol(RSVP) can be used at the IP QoS manager and theIP bearer service manager in the WPAN device andthe PNG, respectively. For illustration purposes, wehave chosen RSVP because it can provide accurateand complete description of application QoS require-ments. The basic idea is to employ RSVP as a localresource reservation protocol between a WPANdevice and the PNG, and to use the QoS descriptioncontained in the RSVP messages for interworkingwith the QoS functionality in other networks.

The application in a WPAN device sets its QoSrequirements in an RSVP PATH which is processedby the WPAN QoS manager in the WPAN deviceand conveyed to the PNG. Upon receiving the RSVPmessage, the PNG performs the following tasks:translates RSVP parameters into PDP context para-meters; initiates PDP context activation procedures ifRSVP PATH message is received; and negotiates thePDP context characteristics with the UMTS network.Conversely, if the QoS requirements are initiated bythe external IP network, the PNG performs the fol-lowing tasks: translates PDP context into RSVPparameters, constructs and sends an RSVP PATH tothe recipient; and completes the PDP context modifi-cation when the RSVP RESV is received.

Figure 14 shows the signaling flows. We assume thata PDP context already exists, which is set up duringthe IP address allocation. The existing PDP context isreferred to as the primary PDP context and supportsbest-effort data only. The QoS request is initiated bythe external IP network. The IP bearer service man-ager of GGSN maps the requested QoS into PDP con-text parameters and triggers the GGSN-initiated PDP

Figure 13 Personal Network QoS Management

Router

ApplicationApplication

WPAN/WLANdevice

IP QoSmanager

WPAN QoSmanager

IP bearer servicemanager

PNG

SGSN

UTRAN

GGSN

WPAN QoSmanager

UMTS bearerservice

manager

UMTS bearerservice

manager

UMTS bearerservice

manager

Remoteterminal

IP QoSmanager

IP QoSmanager

IP bearer servicemanager

QoS in PDP context



context modification (steps 1 and 2 in Figure 14).Upon receiving the Modify PDP ContextRequest message, the requested QoS requirementsare translated into an RSVP PATH message which issent by the PNG. Once the PNG receives the RSVPRESV message, it triggers the Modify PDP Con-text Accept message (steps 4 to 6 in the figure).

3.3 Routing

Routing in personal networks can be categorized intotwo levels: intra WPAN routing, and WPAN andremote area network routing as shown in Figure 15.

3.3.1 Intra WPAN Routing

Intra WPAN routing deals with the communicationbetween two devices in the same WPAN. That is, thecommunication does not involve the interconnectingstructures. In a WPAN, if direct communication isnot feasible due to out of radio range, then multi-hopforwarding is utilized. Multi-hop communication isan issue that belongs to the area of Mobile Ad hocNETwork (MANET) routing which has been anactive area of research for the last decade. TheMANET research efforts are mostly concerted bythe MANET Working Group [14] of the InternetEngineering Task Force (IETF). However, MANETmostly deals with large-scale, military-typed ad hocnetworks. Furthermore, MANET assumes that thewireless technology is homogeneous, i.e. all thedevices use the same radio access technology, e.g.IEEE 802.11. Conversely, multi-hop communicationin a WPAN could be over different wireless technolo-gies. For instance, a WPAN is composed of a Blue-tooth-enabled device, an IEEE 802.11-enabled deviceand a PNG which has multiple interface includingBluetooth and IEEE 802.11. When the Bluetooth-enabled device and the 802.11-enabled device wantto communicate with each other, the communicationis only possible via the PNG. In addition, the multi-hop communication is over different network topolo-gies. The size of WPAN/WLAN is relatively small ascompared with the scenarios considered in MANET.Instead of designing new routing protocols, which is

not the scope of this paper, state-of-the-art routingprotocols can be adopted and customized to suit theneed of personal networking. The idea of PNG-assisted routing seems appealing. As mentioned insubsection 3.1, the PNG co-ordinates the formationof WPAN and naturally, it becomes the central con-troller of the WPAN. Hence, it is fully aware of thedevices in the WPAN. During the PNG discoveryphase, routes are constructed between the PNG andany device that wants to join the WPAN. Therefore,the PNG can be used as a default router. For instance,if device A wants to send packets to device B in theWPAN, then device A sends the packets to the PNGwhich in turn forwards to device B using the routeconstructed during the PNG discovery phase. WithPNG-assisted routing, the ad hoc routing protocol isenergy-efficient and simple since the device does notneed to build and maintain routing table to otherdevices in the WPAN. An energy-efficient ad hocrouting protocol is needed because WPAN devicesare usually battery-powered. The PNG-assisted rout-ing can result in non-optimum routes and high trafficload at the PNG. However, the performance will notbe critical since the network size of WPAN is small.

3.3.2 WPAN and Remote Area Network Routing

This level of routing deals with the communicationbetween a WPAN and another WPAN or a remotearea network such as a corporate area network. Thecommunication may involve interconnecting struc-tures or go via ad hoc networks. In either case, thecommunication entry and exit point of the WPAN isthe PNG or the gateway for the remote area network.

3.4 Mobility Management

Mobility management is responsible for tracking thedynamics of the personal network and users. Severaltypes of mobility are identified within personal net-works, viz. terminal mobility, network mobility andsession mobility. To date, solutions for each type ofmobility have been investigated and developed sepa-rately. In personal networks, we require a unifiedmobility solution that is efficient and can support all

Figure 14 QoS Signalling with RSVP

PNG UTRAN SGSN GGSNWPAN/WLANdevice

3. RSVP PATH

Primary PDP context

1. update PDP context

request

2. modify PDP context reqest

4. RSVP RESV

2. modify PDP context reqest

5. modify PDP context accept 6. update PDP context

accept



three types of mobility simultaneously. In this sub-section, we will describe such a solution.

Terminal mobility arises due to devices joining orleaving the WPAN. The WPAN generally residesin the same location as the user who will be mobile.As the user moves around his/her environment, newdevices may be encountered and attached to theWPAN, similarly other devices may becomedetached. The IETF has standardized network-layersolutions to support terminal mobility, namelyMobile IPv4 [15] and Mobile IPv6 [16]. The IETFhas also standardized and has been working on differ-ent solutions such as the Stream Control Transmis-sion Protocol (SCTP) [17] and the Host IdentityProtocol (HIP) [18]. For a comprehensive surveyof terminal mobility solutions, we refer to [19].

When the entire WPAN moves as a unit and changesits point of attachment to the interconnecting struc-ture, it is referred to as network mobility. For exam-ple, a WPAN switches its connection from WLANto UMTS. The IETF has established a working groupcalled NEtwork MObility (NEMO) [20] to standard-ize solutions for network mobility. NEMO aims atextending existing solutions to support networkmobility in IPv6.

Session mobility concerns the transfer of an ongoingsession from one device to another. A session is anactive transport connection (i.e. TCP) between two

communicating devices. The need for a session trans-fer arises when a device is detached or a new andmore powerful device joins the WPAN. Sessionmobility is inherent to personal networks. The solu-tions for terminal mobility and network mobility donot cater for session mobility.

Hence, we propose a session mobility solution whichleverages the advantages of Virtual Network AddressTranslation (VNAT) [21] and Mobile IP. The solu-tion can also deal with terminal mobility and networkmobility. Currently, IP addresses are used to identifythe end-point of TCP connections. As each device isassigned a different IP address, it is impossible totransfer the TCP connection to another device with-out breaking and restarting the connection on anotherdevice. VNAT decouples the TCP end-point identifi-cation from the IP addresses by using virtual add-resses. The virtual address is then mapped to a corre-sponding IP address at the Network layer. VNAT,however, relies on external servers for obtaining IPaddresses, which incurs extra overhead. Instead ofrelying on virtual addresses and external servers, wecan achieve a similar decoupling effect by using thehome IP address of a device in combination withMobile IP.

The protocol architecture, which contains the VNATand Mobile IP layers, is shown in Figure 16. VNATis composed of the connection translation layer andthe session transfer management layer. The VNAT

Figure 15 Routing in Personal Networks

Corporate area network


INTRA WPAN routing

WPAN and remote area network routing

Internet

Node B

PNG

WPAN

UMTS

network

Gateway



connection translation is responsible for mapping theIP address used for TCP end-point identification intothe IP address of the new device after the sessiontransfer. The VNAT session transfer managementfacilitates the automatic migration by securing andkeeping alive sessions during the migration process.Mobile IP is then used by the new device to informthe home agent of the old device to tunnel packetsbelonging to this session to the new device.

Figure 17 shows an example of a session transferfrom device A to device B in the WPAN. The sessionis a TCP connection set up between device A anddevice C which resides in the user’s WPAN andhome area network, respectively. On device A anddevice C, the end-point of the TCP connection isidentified by the IP address of A, IP address of C andport numbers as shown in the bottom diagram of thefigure. The port numbers are not shown in the figure.Then, session S is transferred to device B. In orderto keep the session alive, the end-point identifierremains the same as on device A and device C. Ondevice B and device C, the VNAT Connection Trans-lation module is responsible for mapping IP addressA into the IP address of B, and vice-versa. The figurealso illustrates terminal mobility. Device B is a for-eign device in the WPAN, then it will obtain a care-of IP address. If Mobile IP route optimization isemployed, the IP home address of B, which is usedby the VNAT Connection Translation module ismapped into the care-of address as shown in Figure17. The advantage of using Mobile IP is that func-tions for network mobility easily can be incorporated.

5 Personal Network Related

Projects

In this section, we present some of the current per-sonal network projects in Europe and around theworld. Note, it is not in the scope of this paper topresent an exhaustive list.

5.1 IST MAGNET Beyond

My personal Adaptive Global NETwork and Beyond(MAGNET Beyond) [22] is an Information SocietyTechnologies (IST) project, which builds on theachievements and results of its predecessor, i.e. theMAGNET project. The objective of MAGNETBeyond is to design and develop the concept of per-sonal network that supports context-aware resource-efficient, robust, ubiquitous personal services in asecure, heterogeneous networking environment formobile users.

5.2 Personal Distributed Environment

The concept of personal networking is also beingdefined and developed by the Mobile Virtual Centre

of Excellence (MVCE) at the University of Strath-clyde [23]. The personal network concept is knownas the Personal Distributed Environment (PDE). Theobjective of the PDE is to provide virtual personalnetwork connectivity in a dynamic and heterogeneousenvironment, irrespective of the location of devicesincluded in the personal network.

5.3 IBM Personal Mobile Hub

IBM has built a Personal Mobile Hub (PMH) [24]which serves as a gateway between a WPAN and theInternet. The functionality of PMH is similar to thePNG. The WPAN consists of devices worn by userssuch as medical sensors, wrist-watch computers, etc.

5.4 MOPED

In [25], the authors presented a networking modelthat treats a user’s set of personal devices as aMOPED, an autonomous set of MObile grouPEdDevices (MOPED), which appears as a single entityto the rest of the Internet. All communication trafficfor a MOPED user is delivered to the MOPED, wherethe final disposition of the traffic is determined.

5.5 MyNet

MyNet project [26] is a collaboration between Nokiaand the MIT User Information Architecture group.MyNet aims to study and develop a network architec-ture, tools and applications for simple, secure, per-sonal overlay networks.

5.6 Freeband PNP 2008

Personal Network Pilot 2008 (PNP 2008) [27] is aproject sponsored by the Dutch Research Councilunder the Freeband Communication program. Theproject aims at developing the personal networkingconcept into practical technology and demonstrators.

6 Conclusion

The paper has addressed the major issues of self-organization, establishing and maintaining QoS andmobility management in personal networks. The per-sonal network is built on top of WPANs and UMTS.

Figure 16 VNAT and Mobile IP Protocol Architecture

Transport

VNAT

connection

translation

Mobile IP

IP

WPAN/WLAN

layer 1/2

Application

VNAT session

transfer management



The WPAN technologies considered are Bluetoothand IEEE 802.15.3. IEEE 802.11, which is a WLANtechnology, is also considered due to its popularity.A key component in the WPAN and UMTS co-opera-tion is the PNG which seamlessly connects a WPANto a UMTS network. The PNG enables the WPAN todynamically configure globally unique IP addressesand to provide routing information for devices in theWPAN. As for QoS, the QoS provisioning function-ality in WPAN and UMTS was identified and inter-working QoS management modules were designedto allow seamless QoS establishing by applications.We proposed a unified solution to deal with terminalmobility, network mobility and session mobility inthe personal networks.

Acknowledgement

This work was partially funded by the IST MAGNETBeyond and the Freeband PNP2008 projects.

References

1 Roberts, M L et al. Evolution of the Air Interfaceof Cellular Communications Systems toward 4GRealization. IEEE Communications Surveys andTutorials, 8 (1), 2006.

2 Niemegeers, I G, Heemstra de Groot, S M.Research issues in ad-hoc distributed personalnetworking. Wireless Personal Communications,26 (2-3), 2003.

Home area network

Device C

Gateway


A

address

C

address

VNAT

connection translation

B

address

C

address

Mobile IP

B

care of address

C

address

After session transfer

TCP end-point identifier

device B

A

address

C

address

VNAT

connection translation

B

address

C

address

Mobile IP

B

care of address

C

address

After session transfer

TCP end-point identifier

device C

Node B

UMTS

Session S

House

PNG

WPAN

Device A

Device B

Internet

Figure 17 Session Mobility with VNAT and Mobile IP



3 3GPP. Available from http://www.3gpp.org

4 Kaaranen, H, Naghian, S, Laitinen, L, Ahtianen,A, Niemi, V. UMTS Networks: Architecture,Mobility and Services. New York, John Wiley,2001.

5 IEEE. Part 15.1: Wireless Medium Access Con-trol and Physical Layer (PHY) Specification forWireless Personal Area Networks (WPANs).2002. (IEEE 802.15.1)

6 IEEE. Part 15.3: Wireless Medium Access Con-trol (MAC) and Physical Layer (PHY) Specifica-tion for High Rate Personal Area Networks(WPANs). Sept 2003. (IEEE 802.15.3)

7 IEEE. Wireless Medium Access Control andPhysical Layer Specification. 1999. (IEEE Stan-dard 802.11)

8 Tsirtsis, G, Srisuresh, P. Network Address Trans-lation – Protocol Translation (NAT-PT). 2000.(IETF RFC 2766)

9 Cheshire, S, Aboba, B, Guttman, E. DynamicConfiguration of IPv4 Link-Local Addresses.2005. (IETF RFC 3927)

10 Thomson, S, Narten, T. IPv6Stateless AddressAutoconfiguration. 1998. (IETF RFC 2462)

11 Weniger, K, Zitterbart, M. Address Autoconfigu-ration in Mobile Ad Hoc Networks: CurrentApproaches and Future Directions. IEEE NetworkMagazine, 18 (4), 2004.

12 3GPP. Quality of Service (QoS) Concept andArchitecture. 2006. (TS 23.107)

13 Chakravorty, R, Pratt, I, Crowcroft, J. A Frame-work for Dynamic SLA-based QoS Control forUMTS. IEEE Wireless Comms. Magazine, 10 (5),2003.

14 Mobile Ad hoc Networks (MANET). 2006,December 1 [online] – URL:http://www.ietf.org/html.charters/manet-charter.html. (Work in Progress)

15 Perkins, C. IP Mobility Support for IPv4. 2002.(RFC 3344)

16 Johnson, D, Perkins, C, Arkko, J. Mobility sup-port in IPv6. 2004. (IETF RFC 3775)

17 Stewart, R, Xie, Q, Morneault, K. Stream ControlTransmission Protocol. 2000. (IETF RFC 2960)

18 Moskowitz, R, Nikander, P. Host Identity Proto-col (HIP) Architecture. 2006. (RFC 4423)

19 Le, D, Fu, X, Hogrefe, D. A Review of MobilitySupport Paradigms for the Internet. IEEE Com-munications Surveys and Tutorials, 8 (1), 2006.

20 Devarapalli, V, Wakikawa, R, Petrescu, A,Thubert, P. Network Mobility (NEMO) BasicProtocol. 2005. (IETF RFC 3963)

21 Su, G, Nieh, J. Mobile Communnication withVirtual Network Address Translation. ColumbiaUniversity, 2002. (Technical Report)

22 IST MAGNET Beyond. Available from:http://www.ist-magnet.org

23 Dunlop, J, Atkinson, R C, Irvine, J, Pearce, D.A Personal Distributed Environment for FutureMobile Systems. Proceedings of IST Mobile Sum-mit, 2003.

24 Husemann, D, Narayanaswami, C, Nidd, M. Per-sonal Mobile Hub. Proceedings of the 8th Inter-national Symposium on Wearable Computers(ISWC), 2004.

25 Kravets, R, Carter, C, Magalhaes, L. A Coopera-tive Approach to User Mobility. ACM ComputerCommunication Review, 31 (5), 2001.

26 MyNet. Available from:http://research.nokia.com/research/projects/mynet-uia/index.html

27 Freeband PNP 2008. Available from:http://www.freeband.nl/



Anthony Lo received his combined BSc/BE degree with first class Honours in Computer Science and Elec-

tronics Engineering in 1992 and his PhD degree in Protocol and Network Engineering in 1996, all from La

Trobe University, Australia. He is currently an assistant professor at Delft University of Tecnology in the

Netherlands. Prior to that, he was a Wireless Internet Researcher at Ericsson EuroLab, where he worked on

research and development of UMTS and beyond 3G systems.


Weidong Lu received his Bachelor’s degree from Southeast University, Nanjing, China in July 2001 and his

Master’s degree from Delft University of Technology, Delft, The Netherlands in August 2003. In November

2003, he joined the Wireless and Mobile Communications Group in the Telecommunications Department

at Delft University of Technology as a PhD student. His research topic is the Self-organization of Personal

Networks and his research interests include: Wireless networking and ad hoc networks, Mobility Manage-

ment and QoS in Wireless Networks.


Martin Jacobsson graduated with an MSc in Computer Science from the University of Linköping, Sweden in

2002. In 2003, he joined the Wireless and Mobile Communications group at Delft University of Technology

where he is working towards a PhD degree. He has participated in several Dutch and European funded

research projects. His PhD research includes ad hoc and self-organization wireless networking techniques

in combination with infrastructure-based networks for personal networks.


Venkatesha Prasad got his bachelor’s degree in Electronics and Communication Engineering from the Uni-

versity of Mysore, India in 1991. In 1994 he received the M.Tech degree in Industrial Electronics and the

PhD degree in 2003 from the University of Mysore, India and Indian Institute of Science, Bangalore, India,

respectively. During 1994 and 1996 he worked as a consultant and project associate for ERNET Lab of

ECE at Indian Institute of Science. While pursuing his PhD degree, from 1999 to 2003 he was also working

as a consultant for CEDT, IISc, Bangalore for VoIP application developments as part of a Nortel Networks

sponsored project. From 2003 to 2005 he headed a team of engineers at the Esqube Communication

Solutions Pvt. Ltd. Bangalore for the development of various real-time networking applications. From 2005

till date he has been with the Wireless and Mobile Communications group at Delft University of Technology

working on the EU funded projects MAGNET/MAGNET Beyond and PNP-2008, as well as guiding students.


Ignas Niemegeers received a degree in Electrical Engineering from the University of Gent, Belgium, in 1970.

In 1972, he received an MSEE degree in Computer Engineering and in 1978 a PhD degree from Purdue Uni-

versity in West Lafayette, Indiana, USA. From 1978 to 1981 he was a designer of packet switching networks

at Bell Telephone Mfg. Cy, Antwerp, Belgium. From 1981 to 2002 he was a professor in the Faculty of

Computer Science and Electrical Engineering at the University of Twente, Enschede, The Netherlands. From

1995 to 2001, he was Scientific Director of the Centre for Telematics and Information Technology (CTIT) of

the University of Twente, a multi-disciplinary research institute in ICT and applications. Since May 2002,

he holds the chair in Wireless and Mobile Communications at Delft University of Technology in The Nether-

lands, where he is heading the Centre for Wireless and Personal Communications (CWPC) and the Depart-

ment of Telecommunications. He is an active member of the Wireless World Research Forum (WWRF) and

IFIP TC-6 Working Group on Personal Wireless Communications. He has been involved in many European

research projects, in particular ACTS TOBASCO, ACTS PRISMA, ACTS HARMONICS, RACE MONET, RACE

INSIGNIA and RACE MAGIC. Presently, he is participating in the IST projects MAGNET on personal networks,

and EUROPCOM on emergency ad hoc networks.



1 Introduction

Service discovery allows services to advertise them-selves so that clients can query and possibly accessthem. Service discovery is most often local and hencelimited in range. With the distribution of services inincreasingly large-scale environments, new para-digms for service discovery, creation and provision-ing are needed to cover wider areas and prepare forthis evolution. Wide-area service discovery hasbecome a requirement and has consequently attractedmuch interest1).

Wide-area service discovery is a key enabler forwide-area pervasive environment and especially forPersonal Networks (PN) that have gained interestrecently in the research community [1]. The PN con-cept extends the Personal Area Network (PAN) byincluding remote personal devices in the networkarchitecture (Figure 1). The PN can be viewed as acollection of geographically scattered clusters thatcan communicate via interconnecting structures or inad hoc mode. A PN cluster is a network of devicescharacterized by a common trust relationship and

Wide-Area Publish/Subscribe Service Discovery

– Application to Personal Networks

W A S S E F L O U A T I , D J A M A L Z E G H L A C H E

Wassef Louati is

a PhD student at

INT, Evry, France

Djamal Zeghlache

is Professor and

Head of Wireless

Networks and

Multimedia Ser-

vices Dept., INT,

Evry, France

This paper explores the use of event-based communications to design publish/subscribe service

discovery architectures that are adequate for wide-area pervasive environments. Some service

discovery protocols and paradigms lack this event-based communication capability. They also often

present weaknesses in expressiveness, extensibility and flexibility. Some service discovery frameworks

include eventing but are limited to local area discovery. All these features and capabilities are needed

to create the next generation of mobile and large-scale Internet services. A Wide-area Publish/

subscribe Service Discovery (WPSD) framework meeting such requirements is presented and

analyzed in terms of design and implementation characteristics. The applicability of WPSD to service

discovery and management in Personal Networks is also examined.

1) The wide-area service discovery is the current focus of the IRTF P2PRG Content, Resource and Service Discovery (CORE)Subgroup. The purpose of this subgroup activity is to define a research agenda within the P2PRG community to evaluate existingresearch, identify requirements, and develop solutions for wide-area P2P content, resource, and service discovery.

P-PAN

Cluster(work)

Cluster(home)


Figure 1 A Personal Network



located within a limited geographical area (e.g. homecluster, office cluster, etc.). The PAN surrounding theuser is a special cluster called Private PAN (P-PAN).

Wide-area service discovery can support many appli-cations including service control, dynamic servicecreation, context awareness and network manage-ment. Personal networks can also considerably bene-fit from wide-area service discovery frameworks.

Clients (e.g. PN user, PN application, PN manage-ment system, etc.) need in fact to be notified aboutservices that match their queries at service announce-ment times. Clients may also require to be notifiedabout modification of state and access informationof services due to mobility and dynamic changes. Toperform such notifications, service discovery shouldembed eventing mechanisms.

Existing architectures supporting eventing (e.g. UPnP[2], Jini [3]) are suitable for small-area and local dis-covery but cannot support wide-area and remote dis-covery because they use synchronous or multicastcommunications that suffer from scalability problems.

In the context of wide-area environments, eventingis known as event-based communication or as apublish/subscribe model. In this model, subscribersdescribe the events (data, messages, etc.) that theywant to receive from publishers, without havingprevious knowledge of each other’s locations. Themodel ensures the event flow from publishers to sub-scribers so that notifications can be received by thesubscribers. The strength of this model is its loosecoupling between publishers and subscribers in time,space and synchronization. This characteristic makesevent-based communication a scalable communica-tion model that is widely used in many large-scaleInternet services and mobile computing environ-ments. A natural evolution for service discovery isthus to integrate event-based communication to pro-vide eventing capability and achieve wide-area ser-vice discovery.

The objective of this paper is to address such servicediscovery frameworks and especially analyze a Wide-area Publish/subscribe Service Discovery (namedWPSD) framework as a potential solution.

Many applications, especially PN applications andservices, can benefit from WPSD. These applicationsneed to be informed immediately about changes incontext data, services states and PN networking andmanagement.

The paper conducts its analysis of wide-area publish/subscribe service discovery by first presenting an

overview of event-based middleware. This descrip-tion is followed by definitions and requirements onthe proposed WPSD service discovery framework. Ageneral architecture for WPSD is then presented anda specific implementation reported. Finally, the appli-cability of WPSD to Personal Networks is examined.

2 Event-Based Middleware

An event-based middleware provides publish/sub-scribe communication between components of large-scale distributed systems. In an event-based middle-ware, events represent the basic or fundamental com-munication mechanism [4].

Event consumers express their interest, in receivingevents, in the form of an event subscription. Eventconsumers are commonly called Subscribers. Eventproducers produce events, which will be delivered toall interested Subscribers. Event producers are com-monly called Publishers. Thus, an event-based sys-tem has an asynchronous many-to-many communica-tion model, where the Publishers and the Subscribersare decoupled. The event delivery is performed byevent brokers that form an overlay network thatroutes the events from the Publishers to the Sub-scribers. The path is set up by advertisements andsubscriptions sent by Publishers and Subscribersrespectively.

The event-based middleware architecture has fourlayers (Figure 2) [4][5]. The lowest layer is the net-work layer. On top of this layer, an overlay networkimplements application-level routing (content-based,peer-to-peer (P2P), etc.). The overlay handles therouting between the application-level nodes (eventbrokers, peer nodes, etc.) and has a self-organizingmechanism that supports the node dynamic changes(when joining and leaving the overlay).

The event-based middleware layer provides the pro-grammer an interface to handle the events (advertise-ment, subscription, publication, etc.) and managesalso the storage, event filtering and notification oper-ations in the event brokers.

Figure 2 Event-based middleware architecture

Services

Event-based middleware

Overlay routing network

Network



The event-based middleware layer can benefit fromextensions offered by the service layer that can pro-vide additional services such as security and QoS.An event-based system can be extended to supportQoS-aware paths between the Publishers and the Sub-scribers such as in [6]. Introducing security into thesystem can be achieved for example by making theevent brokers OASIS-capable (Open Architecturefor Secure Interworking Services) [7] to perform aboundary policy-based access control to the middle-ware. The broker would use the OASIS policy todecide if the advertisement/subscription can beadmitted [8].

3 Wide-Area Publish/Subscribe

Service Discovery

Event-based middleware has a layered architecturedesigned to offer multiple services. Service discoverycan be one possible offered service. Even thoughsome event-based systems can be tailored for servicediscovery, most adaptations suffer from significantoverhead. The event content must be adapted to a ser-vice description and rendezvous brokers must be pre-viously set up by the advertisement to perform notifi-cation. The best solution, as adopted by WPSD pro-posed in this paper, is to build an event-based systemdesigned for service discovery from the start. Theevent content would be explicitly defined (serviceinformation) and the routing of events between thePublishers and the Subscribers carried out at service(event) delivery time.

3.1 Design Requirements

To design a Wide-area Publish/Subscribe ServiceDiscovery (WPSD) system, we define a set of systemrequirements resulting from a mix of our experiencein wide-area service discovery [9][10], the event-based middleware requirements of [4], event-basedproprieties [11] and the problem statement specifiedrecently by the P2PRG CORE subgroup in [12]. Therequirements consist of:

Multiple event support

A WPSD system should notify the subscribers whennew events occur:

• Appearance of new services• Departure of services• Dynamic changes in service state• Changes in mobile service location• etc.

Reliability

A WPSD system must be resilient to entity failures.The failure of a part of the system must not causetotal system failure. The system must be always

aware of the location of the nomadic services andtheir clients, especially with the emergence of wire-less technologies. The system should also supportthe mobility of the brokers.

Scalability

The scalability is an important requirement for anyWPSD system since the number of involved devices,users and services is in continuous increase.

Dynamic update

In a WPSD system, a subscriber should not receivean interrupted stream from the publisher. When a ser-vice is dynamic and/or mobile, its service informationupdate should not be performed by removing the oldinformation and announcing the new one. Only anupdate due to the new information should take place.

Security

A WPSD system should support security and privacyprotection. Especially, the security mechanismsshould protect the system from attacks related to thedistributed architectures.

Interoperability

Many service discovery technologies can be used atthe edge of a WPSD system. Thus, WPSD should beable to interact with different service discovery sys-tems. This leads to the following interoperabilityrequirements:

• A linking mechanism (i.e. gateway) should be usedto make the translation between the WPSD and thelocal discovery systems when service descriptionsemantics differ.

• Gateways at the edges should have an API thatsupports all operations (addition, removal, dynamicupdate, security, etc) to enable interaction.

• The WPSD system should use an expressive andextensible service description language to be capa-ble of translating descriptions to/from other lan-guages.

There is a trade-off to achieve between expressive-ness in the service description language and scalabil-ity [13]. A highly expressive service descriptionlanguage would require more processing and wouldmake the system less scalable. The presented designin this paper takes into account this trade-off by striv-ing for a balance between expressiveness and scala-bility.

3.2 Design Decisions

In this section, we describe how to integrate the pub-lish/subscribe model in a wide-area service discovery



leading to a WPSD system. In this description, wedefine the required additional operations, the in-volved entities and their interactions.

To start with, we must select the right model thatshould be applied to service discovery. Mainly, thereare two categories of publish/subscribe models: topic-based and content-based. Each form has its own man-ner of expressing the event subscriptions. The topic-based model uses a predefined set of topics or sub-jects to identify the events. In content-based model,the events do not have such restricted structure.

In topic-based model, the topics are not expressive.They just use predefined labels and this is not suffi-ciently rich to specify service subscriptions. In thecontent-based model, a service subscriber can makemore specific service subscription using richerdescription formats (e.g. attribute/value pairs, XML,

etc.). WPSD hence adopts the content-based modelsince it is more suitable for service discovery applica-tions.

In a WPSD system, a service announces its serviceinformation in the form of a pair (name, name-record). The name denotes the service description(e.g. attribute/value pairs) and the name-recorddenotes the access information to the provider ofthe service, such as its IP address. By handling nameto address mapping, a WPSD system acts also as anaming and name resolution system.

The producer and consumer in a WPSD system arethe service Announcer and the service Subscriber.The event brokers of WPSD, called Name Brokers(NB), are name information databases. The NBsroute names from the Announcers to the Subscribersusing a name-based routing paradigm. The routing

Figure 3 WPSD architecture

IP network

NB overlay network

A

A

S

S

Subscription

Announcement

Notification

Subscriber

Announcer

NNB - Notification Name Broker

NB - Name Broker

Application-level router (e.g. peer)

IP node

S

A

S

Application-level routing network (e.g. P2P)



occurs on an application-level routing that can beachieved using a P2P network or any other type ofoverlay (Figure 3).

The Subscriber announces a subscription name toperform one of the three subscription types below:

• Name subscription to be notified about names thatmatch exactly all the subscription name attributes;

• Supername subscription to be notified about anyname that match the subscription name;

• Name-record subscription to be notified about anymodification of the name-record information (dueto Announcer mobility or dynamic changes in envi-ronment) of the subscription name.

The name subscription is performed using the usualname announcement mechanism of the current ser-vice discovery architecture. The name-record of asubscription name is called subscriber name-record.The Notifications are performed after name informa-tion database updates by the Announcers (addition,removal, and dynamic update). For a given nameupdate session, an NB holding the name databasechanges is called Notification Name Broker (NNB).The location and the number of NNBs for a givensubscription depend on the used name-based routingof the current service discovery protocol. The notifi-cations of the name update are sent by an NNB to theappropriate Subscribers identified by their name-records.

For example, suppose Alice wants to be notifiedabout all her cameras (in her home, office, etc.) anddynamic changes in their states. Alice submits fromher Subscriber a supername subscription like [service= camera] [owner = Alice [cluster = *]]. One or moreNBs that receive the supername will play the role ofNNBs. When a new camera is added or the states ofexisting cameras change, the Announcers related tothe cameras announce or update the names. TheNNBs notify Alice’s Subscriber about these newchanges.

WPSD introduces the ability to make dynamic up-dates of the announced name information. WhenAnnouncers or Subscribers move, their access infor-mation (i.e. name-records) is updated in the NBs. Thecorresponding NNBs notify the Subscribers so thatthey are always aware of the Announcers’ locations.

3.3 WPSD Scalability

WPSD is intrinsically scalable since Announcersand Subscribers are decoupled, as in any publish/subscribe system. Many other factors improve the

scalability of WPSD. The NBs self-configure into adistributed application-level overlay network to per-form name-based routing. By integrating naming androuting in the application-level, the scalability can bemaintained while highly expressive names are used.

The scalability is also improved by the load balancingensured by WPSD. The load induced by subscriptionand announcement is quite balanced and reduces therisk of an NB becoming a bottleneck node in the net-work. This is due to the NB location that depends onthe name to be routed. Since the NNB locationdepends on the subscription names, the notificationsare issued from different NNBs. This distributes natu-rally the load and avoids congestion in the NB net-work.

Scalability and load balancing occur naturally whena Distributed Hash Table (DHT) is used as the appli-cation-level routing network in the WPSD system.DHTs, such as Chord [14], CAN [15], Pastry [16]and Tapestry [17], create structured P2P networksthat, given a key, can efficiently lookup the peer atwhich the corresponding value is stored. The DHTstorage and lookup processes scale logarithmicallywith the number of peers in the network. For DHT-based designs, the NB location may be determinedby hashing names (announced or subscription names)and looking up the NB peers responsible for theresulting keys (i.e. the hash values). This will beillustrated with more details in section 3.6.

3.4 WPSD Security

In [18] the authors present a number of securityissues and requirements in wide-area publish/sub-scribe systems. Based mainly on this analysis, severalsolutions have been proposed to secure publish/sub-scribe systems. Some of these solutions can beadopted and added to secure a WPSD system.

Three main security mechanisms should be add-ressed. First, an access control mechanism should bedesigned at the boundary of the NB network. Onlyauthenticated and authorized Announcers and Sub-scribers should be able to access the NB network. Toaddress access control, the authors in [19] present anaccess control mechanism based on publication/sub-scription message content. This is a suitable solutionfor WPSD since it is a content-based publish/sub-scribe model.

Second, a mutual trust should be provided betweenthe Subscribers and the Announcers. The mostadvanced solution, proposed by the authors in [20], isto introduce the notion of scoping for structuring pub-lish/subscribe systems. Scopes delimit groups of Pub-lishers and Subscribers on the application level and



control the dissemination of notifications within thebroker network. Hence, they are suitable for buildinggroups of trust (i.e. groups whose members belong tothe same authentication domain). The trust relation-ships in scopes are established using public keyinfrastructures and access control methodologies.WPSD can adopt this solution as well, especiallywhen the NBs belong to different administrativedomains.

Finally, confidentiality and integrity of the announcednames, the subscription names, and the inter-brokermessages should be provided. The common solutionis to encrypt and sign the messages, as has beenachieved by the work in [8] and [20].

3.5 The Support of the WPSD Model

The integration of the publish/subscribe model intoan existing wide-area service discovery system canlead to a WPSD system. To support this integration,the selected system should be flexible, extensible andscalable. Several architectures can support the inte-gration of the WPSD model as long as their exten-sions fulfill all previously mentioned requirements.In fact, this depends mainly on the infrastructuremodel of the service discovery protocol. Wide-areaservice discovery protocols ([21]–[25]) have nor-mally a structured directory2)-based infrastructuremodel [26].

There are two major existing directory structures forwide-area service discovery: flat and hierarchicalstructures. In a flat structure, exemplified by proto-cols such as INS/Twine [21] and Superstring [22],directories have P2P relationships. Hierarchical struc-tures, including SSDS [23], CSP [24] and GloServ[25] have a DNS-like directory structure.

Hierarchical structures overcome the overloadingproblem of directories by spawning the load on thechild nodes. Nevertheless, they cannot scale easily toWPSD integration. The announcements and the noti-fications would propagate up and down through thehierarchy and can turn the root node into a bottle-neck. Even if several hierarchies may co-exist, thehierarchies would not handle efficiently subscriptionnames that include orthogonal attributes. For exam-ple, imagine a WPSD system having two hierarchiesof directories: one based on service location and theother based on the service property. A user may sub-scribe to its own services located in a given geo-graphical area. The hierarchy described would nothandle this subscription in a scalable manner, sinceboth hierarchies will be involved in the notification.

Service discovery protocols having flat structuresgenerally rely on DHT. These protocols inherit DHTscalability, efficiency, reliability and robustness.Hence, DHT-based service discovery protocols aremore suitable for WPSD integration than hierarchicalapproaches. The next section describes an implemen-tation example of a WPSD system using INS/Twine,a typical DHT P2P-based service discovery protocol.

3.6 Implementation Example: WPSD Based

on DHT P2P-Based Service Discovery

A viable framework to implement a WPSD system isINS/Twine [21] that attracted our interest for its flexi-bility, expressiveness and ability to use any kind ofidentities to describe objects, nodes, services whileproviding information on their location. This systemresolves names or identities into IP addresses for net-working purposes.

INS/Twine is a P2P-based service discovery protocolwith an expressive, responsive and robust systemdesigned for dynamic and mobile environments.Services are described using a hierarchical attribute-value pair naming scheme such as XML. The descrip-tion is called intentional name. The INS/Twine archi-tecture consists of a network of Intentional NameResolvers (INRs) that provide name resolution andname-based routing. The service providers and theclients are located at the edge of the INR network.

The INR name database has a data structure called aname-tree. The INRs self-configure into an applica-tion-level distributed overlay network over a DHTnetwork to disseminate and route name informationto the appropriate locations.

INS/Twine performs hash-based partitioning of ser-vice information among a subset of INRs. The query-ing has a similar mechanism. INS/Twine extracts eachunique prefix subsequence of attributes and valuesfrom a name (an announced name or a name query).Each subsequence, called a strand, is then used toproduce a separate key. INS/Twine uses a DHT pro-cess to map the keys to appropriate resolvers intowhich the service information or query is forwarded.

WPSD can be designed by extending the INRs of theINS/Twine framework. The resulting INRs play therole of Name Brokers (NBs). The Subscribers corre-spond to clients and the Announcers to serviceproviders.

For name announcement, the name is added in thename-tree of the corresponding NBs. This is per-

2) The directory, also often called broker or resolver, is the entity that stores service information and processes announcements andqueries.



formed by adding the corresponding name-recordto all leaf values of the name in the name-tree (seeFigure 4).

The dynamic update of a name and its access infor-mation consist of finding the corresponding name-record in the name-tree. If the dynamic update con-cerns the access information of the name, the updateaffects the name-record itself. If the name needs to beupdated (i.e. updating a given value in the name), thename-record is detached from the old value and asso-ciated to the new one.

For name subscription, the subscriber name-recordwill be associated to the leaf values of the subscrip-

tion name. In the case of the supername subscriptionthe association will involve all the supername chil-dren values. The name-records of a matching name orsupername will be associated to the values having thesubscriber name-record. This association will triggerthe notification action. The detachment of an associa-tion, due to name removal or update, will also triggerthe notification.

The notification can be achieved by any NB corre-sponding to a key computed from the strands of asubscription name. To achieve scalability, one singleNB is selected to play the role of the NNB. This NNBcorresponds to the longest strand of the subscriptionname.

Figure 4 Name database management in INS/Twine-based WPSD

service accessibility owner

root

camera private public Alice Bob

resolution data-typ cluster cluster

600*800 640*480

picture Car Home

format

jpgSubscriber

name-record

Name-record1 Name-record2

Name-record1: of Alice name ([service=camera[resolution=600*800]

[data-type=picture[format=jpg]]][accessiblility=private][owner=Alice

[cluster=car]])

Name-record2: of Bob name ([service=camera[resolution=600*800]

[data-type=picture]][accessiblility=public][owner=Bob

[cluster=Home]])

Subcriber Name-record: of Alice supername subscription ([eservice=camera]

[accessibiltity=*][owner=Alice[cluster=*]])



Figure 4 gives an example of a name-tree manage-ment. The figure illustrates a dynamic update, asupername subscription and two announced names(related to the two depicted name-records). For thedynamic update, the resolution attribute of the cameraservice has been updated from a camera resolutionvalue of 640*480 to the 600*800 value.

The name-tree depicts a supername subscription([service = camera][accessibility = *][owner = Alice[cluster = *]]). The supername is composed of fourstrands: (service, camera), (accessibility), (owner,Alice) and (owner, Alice, cluster). The subscriberwill be notified of Alice’s name as the correspondingname-record is associated to the values having thesubscriber name-record. The NNB corresponds tothe fourth strand, as it is the longest one.

4 WPSD Applicability for Personal

Networks

In this section, we will show the applicability of theWPSD system for Personal Networks (PNs) (Figure5) as a communications middleware between local

service discovery frameworks with eventing capabili-ties. WPSD can also assist PN establishment and con-text information delivery.

4.1 WPSD-Based PN Establishment

PN networking entails the establishment of secureinter-cluster communication using tunnels. To estab-lish tunnels between clusters, one needs to be capableof locating the clusters and their points of attachmentto intermediate networks. To provide this informa-tion, the concept of a PN Agent has been introduced[27] to assist PN establishment. A possible approachis to use tunnel managers to send tunnel configurationparameters to tunnel end-points needed to connectdistant PN clusters. Tunnel establishment based onpolicies relying on names is referred to as name-based tunnel establishment [28][29]. A scalable pub-lish/subscribe naming system such as WPSD canenable PN establishment by embedding the PNAgent.

This can be achieved as follows. The tunnel endpoints, depicted in Figure 5 as access routers, shallinclude the WPSD Announcers. The Announcers

Figure 5 WPSD for Personal Networks

P-PAN

Cluster(work)

Cluster(home)

PN/user

A S

WPSD system

A S

A

S

A

A

A

Cluster Gateway

Edge router

Tunnel

WPSD/UPnP Transcoder

NB

Subscriber

AnnouncerA

S

Tunnelmanager

S



declare the names of the attached clusters. The clusternames include the PN identifier and the cluster identi-fiers (e.g. [PN = Bob [Cluster = Home]]). The PNAgent maintains information about the cluster namesand their attachment points (the IP addresses of theaccess routers are in the name-records).

The tunnel management, using an embedded WPSDSubscriber, performs cluster name subscription andsubscribes to the name-record of each registered clus-ter name to become aware of the cluster dynamicsand mobility induced changes.

When a cluster changes its attachment point, the clus-ter name is removed from the Announcer in the oldtunnel end-point and is announced by the prospectivenew tunnel end-point. Once the Subscriber is notifiedof these changes (by obtaining the IP addresses of theold and new tunnel end-points), the tunnel managersets up the new tunnel and tears down the old one.

Clusters can frequently merge or split. For examplethe home cluster splits into the P-PAN and the homecluster. The P-PAN can merge with other clusters,e.g. office. During these dynamic changes, theAnnouncers in tunnel end points can perform updatesof the cluster identifier attribute value. The advantagehere is that the update can occur without completelyremoving or announcing again the cluster name. Theactive tunnels can be maintained thus avoiding addi-tional tunnel tear downs and set ups.

4.2 UPnP Extension for Wide-Area Service

Discovery Using WPSD

Universal Plug and Play (UPnP) [2] is a widelydeployed local service discovery protocol. WPSDcan be used as communication middleware betweenUPnP-enabled PN clusters. With this combination, aPN user can access, from any location, any UPnP ser-vice in any given UPnP-enabled PN cluster.

This interoperability requires the design of interwork-ing components called Transcoders. The Transcodertranslates UPnP service descriptions into WPSDnames and vice versa. The Transcoder binds also theUPnP operations (announcement, discovery, andeventing) into WPSD operations and vice versa.

By using WPSD, the discovery can benefit from themost recent dynamic updates of the service descrip-tions. The UPnP Client of the Transcoder can sub-scribe to all UPnP services in its associated clusterusing the UPnP eventing functionality. When receivingthe updated service description from UPnP, the WPSDAnnouncer of the Transcoder can infer the updatedattributes and achieve the value updates. Hence, theUPnP service database remains always up to date.

In addition, UPnP eventing can be concatenated toWPSD eventing if the UPnP services cannot bedirectly accessed remotely. The distant nodes wouldsubscribe to the remote UPnP service through WPSD.

UPnP is just an example to illustrate the interoper-ability benefits of WPSD. WPSD can cooperate withother local area service discovery supporting theeventing function such as Jini [3].

Interoperability of WPSD with popular service dis-covery protocols and frameworks should ease itslarge-scale deployment and adoption as a wide-areaservice discovery architecture.

4.3 Context-Awareness Using WPSD

The publish/subscribe feature of WPSD can enablecontext awareness in Personal Networks if nodessubscribe to context information services offered bynodes acting as context sources. It would be sufficientto use a name Announcer that publishes names thatcorrespond to key context information advertised bythe context sources. The context data consumerswould build from these names a subscription nameand simply subscribe to automatically receive notifi-cations about changes in context from the WPSDsystem. Services and applications can thus be madecontext aware and can adapt according to dynamicchanges in the overall PN architecture and con-stituents.

5 Conclusion

This paper examined the introduction of publish/subscribe paradigms into current service discoveryframeworks to achieve flexible, extensible andexpressive wide-area service discovery architectures.This was accomplished through the design of a gen-eral and generic service discovery architecture namedWPSD that relies on event-based communications.The applicability of WPSD to Personal Networks(PNs) was examined along with interoperabilityrequirements with other discovery frameworks.WPSD can enable service discovery betweenremotely located PN clusters and can facilitate PNestablishment and context information delivery.WPSD can easily interoperate with popular servicediscovery protocols to fulfill the requirements ofwide-area service discovery in pervasive environ-ments. These WPSD features should ease large scaledeployment and adoption.

6 Acknowledgments

The authors would like to thank the reviewers for theirvaluable comments. This work was partially fundedby IST Integrated Project MAGNET No. 507102.



7 References

1 My personal Adaptive Global NET. IST-MAG-NET consortium. Available from: http://www.ist-magnet.org/

2 UPnP Forum Website. http://www.upnp.org/

3 Jini Website. http://www.jini.org

4 Pietzuch, P R. Hermes: A Scalable Event-basedMiddleware. Queens’ College, University ofCambridge, UK, 2004. (PhD thesis)

5 Fiorentino, C, Cilia, M, Fiege, L, Buchmann, A.Building a Configurable Publish/Subscribe Noti-fication Service. IFIP International Conferenceon Distributed Applications and InteroperableSystems (DAIS’05), Athens, Greece, June 2005.Springer-Verlag. (LNCS 3543)

6 Carvalho, N, Araújo, F, Rodrigues, L. ScalableQoS-Based Event Routing in Publish-SubscribeSystems. IEEE Internacional Symposium on Net-work Computing and Applications (NCA ‘05),Boston, MA, USA, 2005.

7 Bacon, J, Moody, K, Yao, W. A model of OASISrole-based access control and its support foractive security. ACM Transactions on Informa-tion and System Security (TISSEC), 5 (4),492–540, 2002.

8 Belokosztolszki, A, Eyers, D M, Pietzuch, P R,Bacon, J, Moody, K. Role-based access controlfor publish/subscribe middleware architectures.International Workshop on Distributed Event-Based Systems (DEBS’03), ACM SIGMOD, SanDiego, CA, USA, June 2003. ACM.

9 Louati, W, Girod Genet, M, Zeghlache, D. UPnPextension for wide-area service discovery usingthe INS/Twine Framework. IEEE InternationalSymposium on Personal Indoor and Mobile RadioCommunications (PIMRC’05), Germany, Septem-ber 2005.

10 Ghader, M, Olsen, R, Prasad, V, Jacobsson, M,Sanchez, L, Lanza, J, Louati, W, Girot-Genet, M,Zeghlache, D, Tafazolli, R. Service Discovery inPersonal Networks; design, implementation andanalysis. IST Mobile Summit 2006, Myconos,Greece, June 2006.

11 Meier, R, Cahill, V. Taxonomy of DistributedEvent-Based Programming Systems. The Com-puter Journal, 48, 602–626, 2005.

12 Buford, J, Ross, K, Kolberg, M. CORE SubgroupProblem Statement. Internet Draft, January 2006.draft-irtf-p2prg-coreproblem-statement-00

13 Carzaniga, A, Rosenblum, D, Wolf, A L. Chal-lenges for Distributed Event Services: Scalabilityvs. Expressiveness. Engineering DistributedObjects (EDO ‘99), ICSE 99 Workshop, LosAngeles, CA, May 1999.

14 Stoica, I, Morris, R, Karger, D, Kaashoek, M F,Balakrishnan, H. Chord: A scalable peer-to-peerlookup service for internet applications. In: Pro-ceedings of the 2001 ACM SIGCOMM Confer-ence, 149–160, San Diego, CA, August 2001.

15 Ratnasamy, S, Francis, P, Handley, M, Karp, R,Shenker, S. A scalable content addressable net-work. In: Proc. of the ACM SIGCOMM Confer-ence, 161–172, San Diego, CA, 2001.

16 Rowstron, A, Druschel, P. Pastry: Scalable,decentralized object location, and routing forlarge-scale peer-to-peer systems. Lecture Notesin Computer Science, 2218, 329–350, November2001.

17 Zhao, B Y, Huang, L, Stribling, J, Rhea, S C,Joseph, A D, Kubiatowicz, J D. Tapestry: Aresilient globalscale overlay for service deploy-ment. IEEE Journal on Selected Areas in Com-munications, 22 (1), 41–53, 2004.

18 Wang, C, Carzaniga, A, Evans, D, Wolf, A. Secu-rity Issues and Requirements in Internet-scalePublishsubscribe Systems. In: Proceedings of theThirty-Fifth Annual Hawaii International Confer-ence on System Sciences (HICSS’02), 303.

19 Miklos, Z. Towards an access control mechanismfor wide-area publish/subscribe systems. Interna-tional Workshop on Distributed Event-based Sys-tems (DEBS’02), Vienna, Austria, July 2002.

20 Fiege, L, Zeidler, A, Buchmann, A, Kehr, R K,Mhl, G. Security Aspects in Publish/SubscribeSystems. International Workshop on DistributedEvent-Based Systems (DEBS’04), May 2004.

21 Balazinska, M, Balakrishnan, H, Karger, D.INS/Twine: A scalable peer-to-peer architecturefor intentional resource discovery. InternationalConference on Pervasive Computing, 195–210,Zurich, Switzerland, August 2002. Springer-Ver-lag.



22 Robinson, R, Indulska, J. Superstring: A ScalableService Discovery Protocol for the Wide-AreaPervasive Environment. Proc of the 11th IEEEInternational Conference on Networks, ICON’03,699–704, Sydney, Australia.

23 Hodes, T D, Czerwinski, S E, Zhao, B Y, Joseph,A D, Katz, R H. An Architecture for SecureWide-Area Service Discovery. WirelessNetworks, 8, 213–230, March 2002.

24 Lee, C, Helal, A. A Multi-tier Ubiquitous ServiceDiscovery Protocol for Mobile Clients. Proceed-ings of the 2003 International Symposium on Per-formance Evaluation of Computer and Telecom-munication Systems (SPECTS 2003), Montréal,Canada, July 2003.

25 Arabshian, K, Schulzrinne, H. GloServ: Globalservice discovery architecture. In: MobiQuitous,319–325. IEEE Computer Society, June 2004.

26 Zhu, F, Mutka, M, Ni, L. Service Discovery inPervasive Computing Environments. IEEE Perva-sive Computing, vol. 4, pp. 81-90, 2005.

27 Louati, W et al. Networking in Personal Networks.Workshop on Applications and Services in Wire-less Networks (ASWN’05), Paris, France, June 29– July 1, 2005.

28 Louati, W, Zeghlache, D. A Dynamic VPN man-agement architecture for Personal Networks.ASWN’05, Paris, France, June 29 – July 1, 2005.

29 Murakami, H, Olsen, R L, Schwefel, H P, Prasad,R. Managing Personal Network SpecificAddresses in Naming Schemes. WPMC’05,Aalborg, Denmark, September 2005.

Wassef Louati received the MS Degree in Computer Science in 2004 from Pierre et Marie Curie University

(Paris, France). Currently, he is a PhD student at the "Institut National des Telecommunications" (Evry,

France) in the wireless networks and multimedia services department. His main research interests include

naming, service discovery and peer-to-peer architectures with a current focus on Personal Networks.


Djamal Zeghlache graduated from SMU in Dallas, Texas in 1987 with a PhD in Electrical Engineering and

the same year joined Cleveland State University as an Assistant Professor. In 1990 and 1991 he worked with

the NASA Lewis Research Centre on mobile satellite terminals, systems and applications. In 1992 he joined

the Networks and Services Department at INT where he currently acts as Professor and Head of the Wire-

less Networks and Multimedia Services Department. He is an active member of the IEEE communications

Society and a member of the IEEE Technical Committee on Personal Communications. He acted as co-

technical chair of the ASWN 2001, 2002, 2005 Workshops and Technical Chair of the Wireless Communi-

cations Symposium for Globecom 2003. He acts as lead scientist for INT in the European project MAGNET

Beyond. He is also an expert group member of the eMobility Platform at the European level for framework

program 7 and involved in WWRF working groups 2, 3 and 6. His interests and research activities span a

broad spectrum of issues related to wireless networks and services. The current focus besides resource

allocation is on dynamic adaptation and configuration of wireless networks and services based on context

awareness and service discovery using P2P and autonomic networking paradigms. An ongoing activity

relates to personal networks seen as a wide area extension of wireless personal area networks involving

remotely located personal clusters. A key objective is to address the challenge of establishing overlay

networks and service overlays for these networks at run time to enable dynamic and context aware

adaptation of services and applications.




1 Introduction

Adaptation to the environment and circumstances isan important issue for any person. People are doingthis constantly without thinking about it, e.g. avoid-ing walking into objects, reacting in certain waystowards different people, switching off the light whenwishing to sleep and so on. Humans do this withoutgiving too much thought about many of these deci-sions we make every day, and even consider manyof them as trivial tasks. For electronic devices this isdifferent. Such devices need instructions in how theyshould behave. Even though several techniques existtoday, most with its origin in Intelligent AutonomousSystems such as robotics that implements both intelli-gence and autonomy, such technology is still at theemerging state in the world of communication. Themain reason for this is that whereas robots and otherintelligent autonomous systems are typically smalland closed systems with a specific purpose to serve,communication technology typically has more gen-eral purposes, and may have to adapt to a variety ofsituations. Furthermore, specific purposed systemsare typically equipped with the necessary sensorsand software to enable them to do their job, whereasmobile communicating devices may have to deal withthe fact that they do not always have directly avail-able the information needed, but will have to discoversources that provide the information needed, and thenobtain the information over the network. This factgives the mobile system an advantage over the specif-ically designed closed systems, as it becomes possi-ble to use all kinds of information imaginable, butwith the price of complexity of achieving a gener-alised methodology to achieve this.

Much research has been ongoing to make devicesintelligent in this way, as for European projects

examples are [E-SENSE], which is a project that aimsto develop a system that captures the environmentalsituation of a given user by the usage of sensorsorganised in networks (sensor networks). The projectMAGNET Beyond [MAGNETB] is focused on con-structing a platform providing context informationwithin the framework of Personal Networks, whichin turn will support context aware applications.DAIDALOS [DAIDALOS] is another project work-ing on a platform to support context aware applica-tions. Other specific projects such as LAICA [Gia-como05 et al.] aims to use ambient intelligence tocontrol human and vehicle flows in real life. In fact,all this activity, and also earlier activities within thisfield shows the potential of this concept. However,there is a long range of problems related to using con-text information to achieve the desired behaviours.

It is the main purpose of this paper to provide anoverview of the challenges that need to be solvedprior to successful deployment of context awareapplications. The key solution to many of these chal-lenges is given as personalisation using profile infor-mation as it will be explained in the paper. This linksclosely to the network paradigm Personal Networks,in which the basic concept is also explained in thepaper. However, there are some implications totaking this approach, as will also be described.

The rest of the paper is organised as follows: The restof the introduction is focused on introducing PersonalNetworks, context, context awareness and how theserelate to each other. Following this, an introduction toa Context Management Framework is given in Sec-tion 2, where requirements to such a system are listedand a high level architecture is described. Challengesin managing some common context information is

Challenges and Solutions in Achieving Personalisation

Through Context Adaptation

R A S M U S L O L S E N

Rasmus L. Olsen

is a PhD student

at Aalborg Uni-

versity, Denmark

This paper introduces the general concept of personalisation by context adaptation of client appli-

cation. To achieve this, there is a need for a framework that gathers, stores, processes context data

and in any other way makes the access to any relevant information easy for any client application

that wishes to utilise context information. This paper addresses some of the fundamental challenges

needed to be overcome before a true realisation of such a vision can be deployed. The general

solution to many of these challenges is achieved through personalisation using profile information.

First is presented an introduction to the Personal Network environment which this concept has been

developed for. This is followed by a description of a Secure Context Management Framework, which

plays a central role in the personalisation concept. Since different context information introduces

different technical challenges, a short introduction to some commonly used information is described

and how it relates to the personalisation concept. Finally is given a description of how different choices

of architectural and security nature impact the adaptive behaviour desired for client applications.



Vehicular cluster

Private PAN (P- PAN)

Local

foreign

devices

Smart building

Home network




Corporate cluster


Internet, UMTS, WLANAd Hoc, etc



Internet, UMTS, WLANAd Hoc, etc

PN2

PN3

PAN

Personal networks (PN)

PN Federation

Home network

Vehicular area network

PN1

Corporate network

Home network

Personal area network (PAN)

Personal network

PAN

Figure 1.1 Conceptual illustration of how the PAN concept has migrated into the PN concept and further tofederation of PNs [MAGNETB]



then described. Section 3 describes how profile infor-mation can be used to personalise context awareness,and finally in Section 4 implications on context aware-ness are discussed in terms of profile location and pri-vacy policies. The paper concludes in Section 5.

1.1 Introduction to Personal Networks and

Federation

Devices within a short geographical range being ableto communicate and at the same time having a per-sonal relation to each other can be perceived as a Per-sonal Area Network (PAN) [Pereira00]. This concepthas now existed for some time, and a natural exten-sion to the PAN concept is called a Personal Network(PN). A Personal Network was first introduced in[Niemegeers02]. Whereas a PAN is limited in its geo-graphical distance, a Personal Network (PN) can spana larger area, potentially globally, covering severalnetwork domains. A PN is hence a network that con-nects the user’s PAN to remote networks, like otherPANs, office networks, or home networks. Figure 1.1illustrates the concept of PNs, showing its hetero-geneous collection of networks and how it relates tothe PAN concept.

In a Personal Network, a user is able to connect tohis or her devices and services using whatever infra-structure is available for communication, e.g. theInternet, WLAN, UMTS or GSM. Personal networksare dynamic just as PANs, in the sense that they arecreated, maintained and destructed in an ad hoc man-ner, e.g. when a user moves around a building, nodesbecome a part of the network ad hoc, and may alsoleave the network as they move out of range or forother reasons are no longer useful to the user. Theheterogeneous network composition that is a naturalconsequence of enabling this vision is also a charac-teristic of a PN.

For Personal Networks the communication betweenclusters is governed by security mechanisms thatensure the user’s privacy and protects the deviceswithin the network from outside attacks. This meansthat security is a key issue in Personal Networks,since sensitive information regarding the user, andprivate services will be accessible to the user any-time, anywhere, while required to be protectedagainst intruders.

Considering that the PN is personal to the user, thecase where the user wishes to share resources or ser-vices with friends, colleagues, family members orothers of interest, clearly requires additional securityfeatures. The concept of sharing resources and ser-vices between PNs is considered as PN federation[Niemegeers05]. A PN federation can be establishedeither manually by one individual inviting relevant

users to a pre-selected subset of resources and/or ser-vices available in the initiating user’s PN. It couldalso be context triggered proactively, e.g. as the usergoes to a meeting, the PN federation between meetingparticipants is established automatically.

In fact, the context triggered PN establishment is justone example of what context information can be usedfor within the scope of Personal Networks. Also,service discovery or provisioning could benefit fromthe knowledge of the given situation of the user andservice. For example finding the nearest availableprinter in the office building requires knowledge ofthe position of the user and the service, or switchingthe video stream from a 3G mobile phone to a nearbylaptop before the 3G mobile phone runs out of batterypower. However, whether a user wishes to have a PNsystem to make such decisions and how exactly itshould react will vary. Hence, there will be an impor-tant interaction between context information, and userprofile data that instructs a client application how touse it.

1.2 What is Context and Context

Awareness?

In this paper the wider definition of context suggestedby [Dey00] will be used as a basis for a definition oncontext:

Context is any information that can be used tocharacterize the situation of an entity. An entity is aperson, place, or object that is considered relevantto the interaction between a user and an applica-tion, including the user and application themselves.

Based on this definition, context in this paper will befocused around the following sub groups:

• User context: Any attributes describing a user’scontext, e.g. the activity, position, status and so on.

• Device context: Any attributes describing the situa-tion of a device and its environment, e.g. batterystatus, OS environmental parameters such as avail-able memory, CPU usage and so on.

• Network context: Any information regarding thenetwork, e.g. link state, end-to-end delay and band-width, topology information, network types etc.

• Environmental context: Any information about theenvironment, e.g. the ambient temperature, lightintensity, sound level etc.

These groups describe well, although still on quite ahigh level, what is relevant to Personal Networks(and other networks in general).



Being context aware [Dey, Pascoe99, Schilit94]means that an entity is aware of the situation it is in,and potentially can react properly to the situation, oras a minimum make decent decisions. In this paper,context awareness is defined as:

Being context aware is the ability of an entity to beaware of the current circumstances and be awareof relevant information that may influence anydecision and behavioural change based on thecontext it may or may not take.

In practice, context awareness can be divided intotwo groups:

• Proactive: Those entities providing proactivecontext awareness use context information pro-actively to react on changes, e.g. by changing para-meters and configuration to fit into the new situa-tion or notifying a user of something.

• Reactive: Those entities reactively using contextinformation typically need to know the current situ-ation to make a decision here and now. An exampleis context aware service discovery, see [CASD106,ACANCASD].

This paper describes how context awareness can beachieved in both the proactive and reactive sense.A range of challenges need to be solved in order tomake this concept strong enough for deployment,which is addressed in this paper. The key solution tothese issues is found in personalisation using profileinformation, as will be explained in the paper. Thereason behind this is that as individuals, users aredifferent and want and expect different things froma system and the applications running on a system.Proactive context awareness may be seen as annoyingby some, while assisting others. The settings that thepaper is based on, is Personal Networks, which isintroduced in the following section, for which person-alisation plays a key role.

1.3 Context Description and Semantics

A particular problem in dealing with context informa-tion is how to describe and model it. Obviously, onecan take any arbitrary approach if interaction withexternal systems is of no concerns. But especially forcontext, this is hardly ever the case, since context perdefinition largely also constitutes information about theworld external to an entity. This makes it necessary forthe entity to understand how other systems describe theworld. It is important for a context description that it is

• Based on a proper context model, i.e a modelthat well describes the relevant objects and theirattributes in order to be useful for client applications;

• That the description language is extensible, i.e. it iseasy to add new objects to the description withouthaving to change any interfaces;

• Understandable by all involved parties, i.e. theclient application and the context provider.

Of existing technologies that can be mentioned,ontology is one of the more important methodologiesused to describe context. Ontologies allow formaldescriptions of concepts and attributes which allowefficient modelling between instances of concepts.Ontologies also allow for reasoning in the sense thatthe device can deduce certain things based on havingknowledge of something else. Ontologies are hierar-chically organised, which makes them useful for anobject oriented model approach. One of the mostimportant ontologies is OWL (Ontology Web Lan-guage) [W3OWL]. OWL comes in three flavours,OWL Lite, OWL DL and OWL Full. Even thoughthey are increasingly expressive, they are alsoincreasingly complex, and whereas OWL Lite andOWL DL are decidable in time, the full version mayhave endless loops. OWL is based on XML and RDF(Resource Description Framework) which also makesthis ontology better understandable for client applica-tions. Notification 3 [W3N3] is another ontology,which is basically a non-XML serialisation of RDF.

1.4 The Need for a Context Management

Framework in Personal Networks

To enable context awareness in an applicationundoubtedly requires additional intelligence imple-mented within the application, concerning the obser-vation of context and the reaction to changes thatoccur. Since there are many applications that maybenefit from context information, it is beneficial tohave a system that can monitor and manage suchinformation to avoid applications replicating the samefunctionality of managing context information. Usinga context management system allows the applicationsand their developers not to worry about problemsrelated to management of context information, whichamong other things include the discovery of informa-tion sources and maintenance. A system performingthese functionalities is sometimes known as ContextManagement system [PERNETS06]. The applicationwould however still need to worry about reacting tocontext information (and interfacing the context man-agement system). Hence, the division of intelligencerequired for context awareness would roughly be sothat all discovery and monitoring functionalities aresuited for the context management, while the reactionand usage logic will need to reside on the client.Having all management logic residing in a contextmanagement system is advantageous for a number ofreasons:



• Clients will achieve a common view of the world.This is an important aspect from a user’s point ofview. Imagine that this is not the case and there isno SCMF, then any application, service or compo-nent being context aware would necessarily have toobtain all information by itself. Now, if there is anylogic used in this process to e.g. derive higher levelcontext information, the same context data wouldnot necessarily be derived in the same way, leadingto different context awareness, in spite of wantingthe same context awareness.

• Client applications will use less processing poweron inferring, monitoring etc. However, in the caseof multiple clients needing access to the same kindof information computer resources can clearly besaved, since instead of having the multiple clientsdoing the same job, only one entity will be doingthe job.

• There is always the potential for saving communi-cation when considering multiple client applica-tions accessing the same information over the net-work, compared to the case when each client appli-cation is responsible for obtaining the informationrequired on its own. Proactive versus reactive up-dating schemes along with various caching strate-gies have an impact on the performance here, butis not further investigated in this paper.

• Clients will not need to worry about discovery andmaintenance of the particular information (whichalso may be optimised with multiple clients), andthis allows an easy interface to the information.

How the information is used is of course of no con-cern to the context management framework. Anydecisions and reactions to the information are clearlyapplication dependent, i.e. this kind of logic must beperformed in the client.

2 Secure Context Management in

Personal Networks

This section describes the requirements and challengesthat a Secure Context Management Framework(SCMF) as envisioned for Personal Networks will face.There are many other initiatives investigating this topic,such as [E-SENSE], ACAN [ACAN], but in this paperthe focus is on the solution proposed in the MAGNETBeyond project, as this is directly related to PersonalNetworks. Only a short introduction to the concept isgiven in this paper, see [PerNets06, CMF06] for moredetailed descriptions. Following the introduction to theSCMF, an overview is given of how and why contextinformation is specifically challenging to manage usingexamples of some commonly used information.

2.1 Requirements to a Secure Context

Management Framework

For a context management framework to be reallyuseful, it must fulfill a set of requirements. Thesecome from the client applications and the individualuser. A brief summary of the key requirements thathave been derived for the context management frame-work in [CMF06] are listed below.

§1 It must be possible to add and remove contextinformation which the context managementframework will be monitoring by the most effi-cient approach.

§2 The client application should have efficientaccess to the context information in the sensethat it does not need to worry about 1) the dis-covery of the context source, 2) what means itshould use to obtain the data, and 3) whether thedata is trustworthy or not.

§3 The client application should not need to worryabout the dynamics of a cluster, i.e. discoveringnew potential sources of information, makinghandover to new and potentially better sourcesetc.

§4 The SCMF needs to be scalable in order to copewith, not only a full sized PN, but also poten-tially large PN federations.

§5 The SCMF should be able to cope with missingor ambiguous context data. This is an indirectrequirement from §2.

§6 The SCMF needs to support proactive contextawareness, i.e. it will need to be able to sendevent messages to client applications to invokereactions at the application.

§7 The SCMF must use standardized data formats,i.e. the potential sources of information use aplethora of different data formats, and for allapplications to understand its output, one or morestandardized output data formats must be used.

§8 The privacy of the user must be ensured by anymeans, either by disallowing the information tobe given by the SCMF or by making the outputdata anonymous.

§9 All access to context data must be authenticatedin order to verify that the information provided isnot falling into the wrong hands.

§10 Data integrity and confidentiality must be keptat all times to ensure that the user will continue



trusting such a system. If the user ever loses trustin the SCMF, the user will most likely not use itanymore, and over time, the system will be lessand less used as it will gain a bad reputation insociety.

§11 Data freshness and non repudiation must bedetected by the system.

In particular the requirements regarding security areof the utmost importance, as context informationpotentially consists of very sensitive material aboutthe user. Using this information to enhance clientapplications functionality and behaviors has its bene-fits, but also its downside if the system leaks theinformation to the wrong persons.

2.2 Overview of the Secure Context

Management Framework Architecture

To meet all the functional requirements set to asecure context management framework a set of func-tional entities can be set up in a framework. For theproject MAGNET Beyond, the components shown inFigur 2.1 have been defined [PerNets06]. Instancesof these components are running in one client, but asmultiple instances of the SCMF are distributed acrossthe PN, the components shown in Figur 2.1 are capa-ble of interacting with other SCMF entities, herebyenabling exchange of context information efficiently.

A hierarchy of the different roles an SCMF entity cantake ensures the required scalability of the system.

In brief, the Context Management Interface (CMI)handles all incoming requests and reformatting ofinput and output between the internal data structuresused by the other components and the exterior world.Also queries involving different kinds of scope, e.g.scope in time and location (network as well as spa-tial), are handled by the CMI. The Context AwareSecurity Module (CASecM) ensures that the datainside the framework is secured from outside, thatauthentication and authorization are ensured, and thatthe privacy of the user is kept according to the user’swishes. The Context Access Manager (CAM) keepstrack of where the data can be accessed most effi-ciently, e.g. if the information should be fetchedthrough from the P&S module, DSA or through thenetwork. The Processing & Storage unit is responsi-ble for storing gathered data, inferring higher levelcontext information, deriving additional metadata tothe context information itself and other more proces-sor requiring operations. The Data Source Abstrac-tion contains the functionality to access and interactwith local context data sources.

The above described components are not necessarilyimplemented on one node, although optimally theywould be. Acknowledging that not all devices have

Local

Context Management Interface (CMI)

Queries SubscriptionsResponses

Ctx management framework

Processing&

Storage (P&S)

Data source(sensors)

Data source(OS metrics)

Retriever

Data source(PHY/MAC

parameters)

Retriever

Data source(....)

Retriever

DSA manager

Local

Local

Network

Data source

abstraction

(To other node’sCAM module)

(To the samenode’s CAM module)

CAM ⇔ CAM

Notifications

Context AccessManager (CAM)

(To the samenode’s CAM module)

Retriever

Context AwareSecurity Module

(CASM)

(To same node’sCAM module)

CMI ⇔ CAM

CAM ⇔ DSAM

CAM ⇔ P&S

Figure 2.1 The components as defined in theMAGNET Beyond project [PerNets06]



the same computational resources available, threedifferent entity types have been defined [PerNets06]:

• Basic Context Node (BCN) which implements alightweight version of Figur 2.1, with none or moreretrievers depending on the number of data sourcesavailable. This will have only the required func-tionality to efficiently access context informationwithin the PN.

• Enhanced Context Node (ECN) which implementsa heavier version of Figur 2.1 which has moreintelligence attached to it to allow inferring andderivation of higher level context and additionalmeta data for the context.

• Context Management Node (CMN) which is anECN, but with the special role of having an over-view of all context within a cluster.

When these entities interact on a local networking level,the BCN or ECN will need to ask the CMN of any con-text information it does not have itself locally from itsown DSA. Once known it may or may not obtain itdirectly from the source by subscribing, eventing orrequest methods. The exchange of information can bedone either by value, or by reference pointers.

All interaction on global level will initially gothrough the CMN at first, while later subscription,eventing or request methods may be communicatedbetween the requesting and providing entities.

2.3 Challenges in Dealing with Context

Information

Managing context information is not only about thedistribution, but also a matter of ensuring that the out-put given to the client application is unambiguousand that the probability of any error is minimised.In the following, a brief overview is given of whatproblem needs to be taken care of to some commonlyused context information [Olesen et al. 2006]. Somemore general issues, which relate to the fact that theinformation is typically distributed in a cluster, PNor PN federation, are such as:

• There may be multiple sources for obtaining theparticular context information which may lead to in-consistency between the true and the used value, orselecting the wrong data source would potentially leadto misbehaviour in the context aware application.

• Mismatch between what is read and what is theactual value. This situation occurs in particularwhen accessing the information remotely, whichmay change value due to some external process.Through various means, it is however possible to

estimate the probability of such error [mmpr06],which can be used to determine whether the infor-mation should be used or not.

• A particular context information is given in a syn-tax not understandable by either the SCMF or theapplication itself. A solution to this could be toattach transformation operators to the context itself,or pointers to where a transformation service canbe found to translate the information into thedesired syntax.

2.3.1 Location Information

Geographical locations has many potential usages,e.g. searching for nearby objects, showing directionto objects or even indicating how other context shouldbe perceived. However, some issues may occur sincethis information can come from various sources:

• Location can be described either as a textualdescription, e.g. Room 203, or given as a set ofcoordinates. This implies that location given indifferent formats cannot easily be compared fore.g. distance calculation. A potential solution couldbe to put an attribute to the context information onwhere and how a location can be transformed. Thiswould also allow transformation between differenttypes of coordinate systems used.

• Bad tracking capabilities, e.g. if relying on a GPSsignal in an indoor scenario. If the system useddoes not provide the functionality to inform aboutthis, how can the context management then figureout to change the source of information?

• Security and privacy issues related to providing thelocation to a client application. It may not be thatsimple due to restrictions in the user’s privacy oreven legislation of a country (which may evenchange as the user travels around the world).

If there is uncertainty about the location, this will havesevere impact on the user, e.g. if the system directs theuser in the wrong direction, finding objects that are100s of kilometres away etc. Hence, it is important fora context management system to address these issues,and as a minimum have a clear policy of what to do insuch situations for then to allow application develop-ers to decide what the application should do.

2.3.2 Time

Time is yet another important aspect in context infor-mation that needs some attention. In many cases,reacting on context will almost always depend onthe right time (and place). In most cases, the SCMFwould rely on the local clock on the device (if avail-able), but there are some issues related to that



• The clock may not be synchronised. If the devicedoes not include automatic synchronisation withsome known source such as GPS, it will rely on theuser having to set the clock manually. How does thesystem detect if the user did not do this properly?Or if it has switched to summer time? One way toaddress this issue is to check what other deviceswithin a cluster consider to be the local time.

• The user may have switched time zone in relationto a travel.

• Relative time relations may not easily be describedas absolute values, such as “information must beless than one hour old”. What will happen if theinformation is one hour and two seconds old,because of update delay, is that the information isconsidered invalid, while in fact it may not be.One potential solution is to describe time relationsusing fuzzy membership functions. However, thisrequires an agreed definition of the membershipfunctions that is available for all nodes within thenetwork before such a solution can be used.

2.3.3 Device Context

Dynamic and static information about the device and itslocal environment, e.g. the OS and its environmentalparameters such as memory, cpu usage, storage spaceavailable etc., can be considered as device capabilities.There are hardly any issues regarding getting this infor-mation as these are typically provided by the OS oraccessed through device profiles. How this is done ishowever system specific, something which needs to betaken into account when implementing such a system.

2.3.4 Network Context

The network state is fundamentally challenging con-text information to address. Since a PN or PN federa-tion contains a long range of network technologies,using a plethora of communication protocol stacks,each with their own states for different purposes, andhaving components working in different levels of theOSI stack model, one would need to focus on a subsetof information to handle. It is however important tomake the SCMF so extensible that mapping new inter-esting information into the system would not be aproblem. Two parameters that are commonly used are:

• Link and end-to-end bandwidth: What is availablefor the application may have an impact on whatservice to choose, e.g. what video stream serviceshould be chosen. However, determining end-to-end bandwidth across a heterogeneous networksuch as a PN remains a challenge to be solved.

• Link and end-to-end delay: Similar to the band-width end-to-end delay may not be easily deter-

mined due to the heterogeneity in the networkcomposition.

A context management framework may help toexchange information between nodes and clusterswithin a PN, allowing a bigger picture of the networkstate in a PN. This would however put additionalrequirements to the model used to describe a fullblown PN, which may not be so simple. Furthermore,the dynamicity within a cluster is rather rapid com-pared to the time it may take to distribute the infor-mation, so the SCMF designer must strike the bal-ance between the level of information distributed inthe PN and the delay of doing so with the dynamicityof the information. If the changes are happening toofast to be efficiently distributed, then perhaps itmakes no sense to distribute it at all, but with theresult that nodes in remote clusters will have anincomplete knowledge of all context in the PN.

2.3.5 Higher Level Context Information

Higher level context information is defined here assynthesized, inferred or derived information, which isbased on one or multiple context information. Inferencerules or other logic operations are used to figure outtypes of context information that sensors or othersources cannot provide. Some examples relate espe-cially to user activity, such as sleeping, walking, in ameeting, generally available or similar. Since this groupof information is based on other context information,some unique problems occur within this group:

• If one or more context information is missing,potentially out of date or in other way not com-pletely trusted, the output, i.e. the inferred informa-tion, will also suffer from not being completelycertain. If at least the probability of uncertainty canbe estimated through the derivation, the applicationmay be able to determine whether to trust theinferred information or not, and adapt its reactionpattern to that.

• The composition and weight of the different con-text information used may be individual and per-sonal, i.e. depending on who the user is, where heis, what role the user currently has, a set of contextinformation may be perceived in one way or an-other. For example how and what context informa-tion to determine if a user is busy or not, maydepend on what job he or she has.

• Since multiple information may be required, thetime to access and get all information may becomea performance factor to be considered. Also theamount of computational resources may need to becarefully considered. On the one hand, the compu-tations are time-wise best performed on the same



Figur 3.1 Adaptation concept on context processingusing profile data [MBD121]

User profile

Context

Contextprocessing

Learning

History

Higher level ctx

Quality ofcontext

User’s choice

Conflict freecontext….

Context

Adaptation

node as the requesting application, while resource-wise, a distribution of the computations and deriva-tion may be the best option.

Without doubt, profiles become an important partwhen deriving higher level context information,which is also one of the aspects covered in Section 3.

2.4 How Personalized Profiles can Remedy

Context Issues

There are of course many more cases, where contextinformation gives ground to potential conflicts ormisbehavior if not carefully analysed prior to theusage. It is out of the scope of this paper to performa deep analysis of all these problems, which wouldalso be a rather tough job, since context may consistof nearly unlimited amounts of information. However,handling these issues is not trivial either, and one mustacknowledge that a solution which works for one per-son may not be suitable for another. For example withlocation, switching to another tracking system whengoing indoors may be fine for Person A, but for Per-son B who does not even like to be tracked, this maynot be needed at all. Perhaps there is a reason thatPerson C has set the clock differently, e.g. becausehe likes to be in control of shifting time zones etc.

By any means, in managing context it is very impor-tant that the user has the possibility to be in control atall times. Whether the user chooses not to use a con-text management system, or to have partial or fullcontrol of how the system manages context, or evenallow the system to do all things automatically simplybecause other options become too cumbersome to theuser, is something that the user needs to decide.

The user may even be selective in what to controland what not to. Such choices may depend on severalthings, such as whether the information is trivial tothe user, e.g. the ambient temperature, or if there areprivacy policies related to it, such as his/her currentgeographical location.

Either way, knowledge about the user’s preferences,will undoubtedly help the SCMF decide what to do,what not to do and what to ask the user on doing, andthis is what the paper will discuss in the followingsections, together with the impact of the user’s deci-sions on the system.

3 Personalisation of Context Aware

Behaviour

To address the problem of individuals and personalviews on context management, profile informationwill play a critical role in the acceptance of any con-text aware behaviour. The direct influence of profile

information in this sense may be divided into twogroups;

1 Within the SCMF, i.e. those profile data that willinfluence the way context is interpreted. These willnecessarily depend on who the user is and to someextent what role the user has.

2 To the client application, i.e. the profiles willinstruct the client application how to react to con-text or context situations.

Common to any context aware client application (i.e.any software entity that uses context information toalter its behaviour), is that there is a need for anagreement between the system and the user, on howcontext is understood and how the clients shouldreact. For one person a given situation may require aspecific reaction, while another person would preferto have the application react otherwise. In fact, eventhe same person may require different reactions to thesame situation, given that the user has different rolesin the situation. User profile data is in this matterrequired to instruct a client application on how theparticular situation or context should be perceived,and how the client should react or relate to it. It isimportant that all clients have access to the sameprofile data, which is why they in this way share thesame view of context. If this is not the case, the userwill risk different instances of the same class of con-text aware clients behaving differently to the samesituation. This is not desirable and leads to the needfor a common understanding of what role the usercurrently has, what the current context is, and whatthe user’s preferences and settings are.

3.1 Personalisation at SCMFLevel

Figur 3.1 illustrates a general description of the rela-tions between various information sources and pro-cessed context information. This will be a part ofwhat is ongoing in the P&S module, depicted in Fig-ure 2.1. There are two main active components in thisdiagram, the processing unit and the learning unit.The processing unit takes relevant input from the dif-



ferent sources, shown to the left, and performs anylogic, statistical or other necessary operation to deriveor infer the required output, with some examplesshown to the right. The learning box, below the pro-cessing unit, is responsible for adapting user profilesbased on processed context and user input, using anymethodologies suitable for doing so. These twoblocks form a closed feedback loop system, with theuser profiles as target for manipulation. The purposeof this interaction is to enable the context processingunit to adapt its derivations and inference metrics sothat in future it will output better results.

Understanding context information, and in particularhigher-level context information, depends on who theuser is and what role the user has. Imagine a bakerand an office clerk. If the processing unit is supposedto determine whether a user sleeps or not, differentinformation can be taken into account for each of thepersons’ jobs. Information such as user location, timeof day, ambient brightness level and sound level maybe used to infer the sleep status of a user by followingpseudo rule, e.g.:

• IF (location is bedroom) AND (time is midnight)AND (ambient brightness is dark) AND (ambientnoise level is low) THEN user sleeps.

Such a sentence may work well for the typical usecase of an office clerk, except sometimes he sleeps athis girlfriend’s or he takes a nap during the day in theweekend. For a baker, the time may not last throughthe whole night, as he would typically have to wakeup at e.g. 3 a.m. Hence he would have a differentsleep pattern that the above rule would not necessar-ily capture correctly. User profiles may then beapplied by the processing unit to instruct how contextdata should be interpreted and used to infer higherlevel context data.

3.2 Personalisation at Client Application

Level

As already stated, any application or service that ispersonalised will need some profile on how this isdone. In this section we distinguish between thoseprofiles that are static and those that are adaptive.The difference between these two types will have animpact on the complexity of the data structures neces-sary to describe the profile, as the static only dependson the profile information, while the adaptive alsodepends on context.

3.2.1 Static Personalisation

By static personalisation is meant the kind of person-alisation that does not depend on context. This couldbe preference values, e.g. for

• Background colour or image of an applicationscreen;

• Default notification sound;

and other behaviour or descriptions that are more orless static. These should of course be able to changewith the user’s will. What is primarily relevant forthis kind of information is:

1 What client the data is relevant for: It is neces-sary to know what client a specific set of profiledata is associated with.

2 A set of attribute names and values: For eachbehaviour a set of attributes and value pairs isneeded to instruct the client what behaviour is beingdescribed, and to what value it needs to be set.

3.2.2 Adaptive Personalisation

Dynamic adaptation of services and applicationbehaviour are in many cases a desired functionalityfor the user and will in most cases require contextinformation, since the world that we are surroundedby is constantly changing. If the user has to manuallysetup the way a service or application should behaveevery time a parameter changes, he or she wouldneed to do so constantly. Some examples of whatadaptation of behaviour could mean are:

• Adjusting the contrast level of the display automat-ically to the ambient light intensity and screencolour so that the user can actually see what’s onthe screen. This is not necessarily an easy operationfor the user to do.

• Change of notification methodology due to envi-ronmental restrictions, e.g. silence ringing whengoing to the cinema, which people tend to forgeteven with the reminders often provided, orincreased sound volume in noisy environments.

• Adjustments of video resolution when streaming;in cases where the network conditions are changingoften as in the wireless domain, it may be prefer-able to shift to a lower resolution to maintain asmooth view.

• Output stream re-direction to more capabledevices, e.g. in case of the video stream, whichmay be viewed originally on the user’s mobilephone, may automatically be switched to her TVscreen or laptop for better view, performance ordue to lack of resources on the current device.

Common for these examples, and for all other cases,is that the adaptation is user and application specificand depends on context information. In fact, to



achieve this kind of operations, a set of informationneeds to be available for the application to do so,primarily:

1 What client this information is relevant for: Thefollowing information must somehow be linkedtogether with a specific client type, i.e. the contextadaptation profile data for a calendar applicationwill most likely not be the same as for a touristguide application, simply because they focus ondifferent objectives, require different inputs etc.

2 What context to react on: This depends on theservice/application and to what extent this informa-tion is available.

3 How to react on it: This depends on the service/application and who the user is, and what role theuser has. In some cases, context information maynot be available, or only related context is avail-able, in which some instructions are needed for theapplication to know what to do then.

4 How to deal with uncertain information: If con-text has been accessed remotely, there will alwaysbe attached some uncertainty, whether this infor-mation has changed during the update. How theapplication should react, if the probability of usingoutdated context information is very high shouldalso be specified. Such action depends not only onthe service/application and their requirements tothe context used, but also on what kind of personthe user is.

In the above examples, adjusting and changing thevideo stream depends strongly on what the user reallydesires, how much the user wishes to be in control(which may also depend on how technically knowl-edgeable the user is), the current role of the user etc.Such information needs to be accessible by the clientanywhere within the PN, and furthermore, if two sim-ilar clients, e.g. two calendar applications are runningon different nodes within the PN, they should adaptsimilarly.

It is clear from the two different situations, i.e.static and adaptive personalisation, that the adaptiverequires much more information, simply becausethere is a need for additional information on whatinformation is needed, and how it should be treated.This overhead of information is something that willimpact the performance of the overall system be-haviour, e.g. reaction time, network traffic, process-ing power (and system resources in general) etc., andhence one of the major challenges with this concept isto minimise the costs of these metrics and still be ableto personalise in a satisfactory way.

3.3 Dealing with Uncertain Information:

Being Optimistic or Pessimistic

No matter what kind of personalisation is being utilised,the client of the context management framework willneed to consider one particular issue; namely that theinformation accessed may be remotely accessed. Thisfact may induce inconsistencies between what isobtained and what is actually there, i.e. some event mayhave changed the value of the information accessed.The Secure Context Management Framework will,based on different statistical information, be able toestimate the probability of a mismatch between thereturned context information, and the true value (whichis a particular problem that always occurs when access-ing remote information, see [mmpr06]). This informa-tion is valuable to any client application, since it can beused to determine whether the information should beused or not. In this way an optimistic or pessimisticclient behaviour can be obtained. Optimistic means thatthe client may take chances on context information,while others would prefer or even require a more pes-simistic system that requires the system to be very con-fident on the context information before being used.

To exemplify the difference, client A may require apessimistic view on information such as location datafor directing a blind person to and down the stairs inhis home, the client may take a pessimistic view toavoid the person being directed wrongly down thestairs hereby avoiding injuries to the person. ClientB, who needs network context to provide a link to aresource for the person’s PN, may be more optimistictowards the information given, since the consequencesof invalid information are less dangerous to the user.The loss in taking a pessimistic view on context is thatclient A will become more reluctant to adapt to a givencontext than an optimistic client. However, client B,who takes an optimistic view of location data, risksthe adaptation being error prone, but will more likelyadapt to the changes detected by the SCMF.

The level of optimism may depend on the individualapplication, but may also be a general systemattribute, depending on what type of person and roleof the user, that clients per default will take. The indi-vidual part will need to be a part of a service profile,while the general system attribute can be a part of theuser profile data.

Whether it will be per client or a global attribute, thereis a need to define the level of optimism that a clientor user wishes to have (which could also be based onthe role of the user, and to the extreme based on othercontext information). As context information will beequipped with an indicator of how likely it is of beingwrong when accessed (the mismatch probability, see[mmpr06]). A simple methodology, as for the user, is



to define an indicator, optimism defined in the interval[0;1], which simply describes the level of probabilitythat the context information has to have before thesystem will consider it useful. A 0 indicates a com-pletely pessimistic system (which can also be viewedas a non context aware system), and a 1 indicates afully optimistic system. Using the Fuzzy Logic [Left-eri et al.] operator, Not, the level of pessimism is alsodefined as not(optimistic) which equals one minusoptimism. The user or a user’s role would necessarilyhave to be somewhere in between zero and one, inorder to be really useful.

4 Client Access to the Data and Its

Implications

In this section, it is analysed how different strategiesof accessing the profile data may influence thedesired context aware behaviour. First is given a dis-cussion on a centralised versus decentralised solution,followed by a discussion on how security and privacypolicies impacts implicated entities.

4.1 Local Versus Remote Storage of

Profile Data

The personalisation concept is, as stated, based onprofile information, which must be present some-where within range of the application. The SCMFoffers two basic approaches for accessing this infor-mation, which are A) a distributed approach, and B)a centralised approach.

To achieve this, two basic principles on where tostore the necessary information can be used, seeFigure 4.1:

A) A distributed solution: All the information is dis-tributed within the PN and is accessible throughthe PN:- Consistency problems may become a problem

when maintaining the data. Different distributionalgorithms and strategies may be applied withvarying impact on read access delay, inconsis-tency probability and generated network traffic,which need to be weighted against the impor-tance of keeping data relatively fresh in the PN.

- Maintenance is harder since the information isdistributed and may involve more complexoperations to ensure that all nodes have accessto the data. Updates may not necessarily hap-pen in one go, since not all nodes in a PN maybe accessible at that time.

- This approach ensures data is available at alltimes, hence suitable for PNs, where connectiv-ity to infrastructure is not always present.

- Data can be kept inside the PN, which is a plusfor users who do not trust or for other reasonsdo not want to put this kind of data on a centralrepository, and potentially place their trust inthird party entities.

B) A centralised solution: All the information isaccessible through one specific node, e.g. a GUPserver [3GPPGUP] or a private node in e.g. theuser’s home, either through SCMF or directly:- This minimises consistency problems with the

data, i.e. updates are done on one master copy,and replicates may exist in the various clusters.

- Maintenance of the data is easy, since this isdone in one place (leading back to consistencyproblems).

- Such a system relies heavily on that node beingaccessible at all times. If this is not the case,then the information is not known. This canbe remedied to some degree by storing locallycached information, while updates may not bedetected easily, leading to inconsistency prob-lems.

- Potential third party solutions, such as GUP,need to be entrusted by the user. Some usersmay like this, and some may not like it verymuch, since potentially sensitive personal datawill be stored at this server.

Whether to take the centralised or decentralisedapproach may influence the experience of personali-sation for the user, as the metrics consistency, readaccess, availability and so on are important for howthe system behaviour is experienced by the user.Since the choice here will have an impact of consis-tency between the obtained and actual data (whenremotely accessed), this may influence the level ofoptimism/pessimism the user desires.

Figure 4.1 Approaches for a client to access SCMF maintained profiles[MBD121]

A)

B)

Client

SCMF

Profile

Context

dataGUP Profile



4.2 Security Implications on Accessing

Context and Profile Information

Whatever approach the client takes in accessing thepersonalisation information, it will eventually need toaccess context information (those who personalisesby adaptation). Since there are two fundamental dif-ferent client types, push and pull, these two typeswill be described in the following:

4.2.1 Pull Types of Clients

In this case it is the user that initializes the communi-cation with the service, invoking a client applicationetc. This can be done either locally on the node, orremotely through e.g. the Service ManagementFramework (in that case it would be the SMN). Eitherway, authorization and authentication are necessaryoperations conducted when invoking the client. Inthis setting, the client will hereafter have the possi-bility of keeping the credentials for authorization andauthentication during a session, when accessing theinformation needed through the SCMF. This will ofcourse require additional functionality in the client,but so will the adaptive behaviour desired.

4.2.2 Push Types of Clients

For Push types of clients the user does not initiate thecommunication with the client. This also means thatin most cases the clients that require access to contextand profile data do not necessarily have the creden-tials to gain access to the information. This may meanthat, depending on the local security policy (and theoverall user’s security policy and levels), context orprofile data may not be accessible to the requestingentity.

4.2.3 Impact of Anonymous Context Information

on Clients

In either case, the client service or application mustbe able to accept that profile and/or context informa-tion may be only partially or not accessible at all, dueto security policies alone. Another obstacle that mustbe accepted by the client, is that the information maybe available but has been anonymised, e.g. a user thatdoes not like to reveal his exact GPS position to anunknown, but useful service may (according to thepolicy) reveal a blurred and/or textual position, e.g.the name of the city he is currently in. This must beacceptable to the service in order to respect the pri-vacy of the user.

This potential issue of anonymising context and pro-file data to the client will without doubt impose diffi-culties in enabling the personalisation concept that isdesired. Even without this issue at hand, a client willpotentially be able to utilise a plethora of information(the total set of context information and profile datagoes towards infinity, considering that definitions of

these are so broad). Furthermore, the fact that therecan be alternative and anonymous name spaces forboth context data and user profiles indicates the diffi-culty of implementing something that is more than aprototype. Hence, the work done here will be basedon standards and possible extensions of these.

Considering anonymous context and profile data tobecome standardised (which seems reasonable alonefrom the fact that everybody then will be using theseformats, hereby truly anonymising the user), a clientcould implement standardised behaviours for theanonymous data provided. This could be the case forboth push and pull clients. Furthermore, there wouldalso be a need for a default way to handle when theinformation is missing due to policy restrictions, notaccessible information or because the level of opti-mism/pessimism dictates that the information shouldnot be used. Hence, the road ahead for ensuring thatthis does not become a problem, is a standardisedsemantics for anonymised context data.

5 Conclusion

Adaptive behaviour is a key issue in current andfuture applications and communication means. Oneway of achieving this adaptive behaviour is to enableapplications, services or other subsystems with theability of being context aware. It is important for anyentity to react properly to its environment, to knowwhat is going on around it, hence the term contextaware. As described in this paper, this is also the casefor Personal Networks, whereas a Secure ContextManagement Framework is envisioned to assist theclient application and its developer to overcome cer-tain aspects in the management of this information.This framework was briefly described along with aset of sub problems related specifically to contextinformation that needs to be addressed prior to a suc-cessful deployment of the framework. The key solu-tion in addressing many of these issues is foundthrough personalisation of context management,which is done through profiles. This concept can existon two levels; the SCMF level and the applicationlevel. The profiles on the SCMF level instructs theSCMF on what it should do in certain situationswhere there are possibilities for ambiguous data con-flicts, invalid data, how to interpret sets of contextinformation, and so on. On the application level, theprofiles instruct the application on how to be contextaware, i.e. what reaction pattern the applicationshould take in a certain situation.

The main point of this paper is that any context basedreaction may be individual from person to person, i.e.one person may have trust in the system or wish notto interact with a potential cumbersome system setup,



while other people would prefer to be in control at alltime or not allow such systems at all to operate forprivacy concerns. Furthermore, certain groups of peo-ple would prefer certain reaction patterns, while oth-ers prefer other reactions, and so on. Rules and infer-ring of context information are also a matter of whothe person is, what role the user has, and how he/sheprefers the system to behave.

In fact, the problem can be boiled down to the ques-tion of whether a user would like to have this kind ofproactive behaviour or to what level the user wouldaccept it and still find it useful. The answer to thisquestion is clear: let the user decide. The technicalsolution to how to achieve context awareness withoutbeing obtrusive to the user is on the other hand nottrivial, but will require that the user accepts a moreproactive system. This may perhaps be the most chal-lenging task of all; to convince users worldwide thatsuch a system is trustworthy and actually beneficialto the user.

The key challenges that need to be dealt with prior tothe success of context aware systems, are manifold;the model and description of context is not trivial anddepends in many cases on more specific use cases.Descriptions need to be accessible and standardisedfor all involved parties in the process, otherwise thecontext aware system will not be able to understandthe information provided by surrounding systems orvice versa. A good formal description of contextrelies on a good context model, whereas ontologiesseem to be the road ahead. But as the world is quitecomplex, with many relations between world objects,a good model is not easy to achieve. Indeed muchwork has been done in this area, but undoubtedlythere is still more work to be done. A good model ofcontext and their relations to others, may also be thekey to resolve for conflicting context informationwhich surely is a problem for context aware systems.However, as already stated, context is difficultbecause it constitutes many different types of infor-mation, neither resolving the context conflict will bean easy task to do, but it is nevertheless an importantfunctionality for context aware applications and ser-vices in order to function properly.

A particular problem with any type of context awareclient is that it will from time to time need informa-tion about the user that may be rather private. Nodoubt, the user will require an absolute secure systemwhich enforces privacy policies, which is a challengein itself, but the real challenge for context awarenessto become a true success is whether the user willaccept that a system uses information about the usersuch as her current geographical position, her currentactivity, mood or other information that may reveal

the situation the user is in. As a minimum, contextaware systems will have to ask the user once in awhile whether this information is okay to use or notfor this and that purpose. Over time, the context man-agement system may learn from this input and askless and less the user for advice. It is at all timesimportant that the user is able to maintain control ofsuch privacy rules. A particular problem that needsto be addressed is proactive context awareness, whichthe user does not necessarily want, and maybe seemslike spam to the user. In other cases the user mightdesire such behaviour to some degree. Since mostusers will not like to constantly change the settings ofthese behaviours, there is an important need to strikethe right balance between user interaction, systemand application behaviour. If this balance is notachieved, then there is a great risk for context awaresystems to fail on the market.

Acknowledgement

The author of this paper would like to thank everyonein the MAGNET Beyond project for fruitful discus-sions. In particular the author would like to thankMartin Bauer and Luis Sanchez for their contributionto and development of the secure context manage-ment framework.

References

[3GPPGUP] Salsano, S. Presentation of the IST Sim-plicity project. 3GPP SA1 meeting, Povoa de Varzim,Italy, July 2005. (Doc. num. S1-050725)

[ACAN] Khedr, M, Karmouch, A, Liscano, R, Gray,T. Agent-Based Context-Aware Ad Hoc Communica-tion. In: Karmouch, A et al. (eds.). MATA 2002,LNCS 2521, 105–118, Springer Verlag, 2002

[ACANCASD] Khedr, M, Karmouch, A. ACAN Ad-Hoc Context Aware Network. Proceedings of the2002 IEEE Canadian Conference on Electrical andComputer Engineering, 1342–1346, 2002.

[CASD106] Olsen, R L et al. Experimental analysisof the influence of context awareness on service dis-covery in PNs. Proceedings of IST Summit 2006,Mykonos, Greece.

[CMF06] Bauer, M et al. Context managementframework for MAGNET Beyond. Invited paper,Joint MAGNET Beyond, e-SENSE, DAIDALOS andCRUISE IST workshop, Myconos, Greece, June 2006.

[DAIDALOS] DAIDALOS. Available from:http://www.ist-daidalos.org/



Rasmus Løvenstein Olsen received his MSc in Electrical Engineering from Aalborg University in 2003 with

focus on antenna control for satellite communication. He has been very active in the pico-satellite program

of Aalborg University called AAU-Cubesat. In 2004 he started in the IST project MAGNET where he has

been working with Context Aware Service Discovery for Personal Networks. He is currently engaged in the

project MAGNET Beyond where he is active in the design of a Secure Context Management Framework for

Personal Networks. He is also pursuing a PhD degree on the topic of context aware service discovery.


[Dey00] Dey, A K. Providing Architectural Supportfor Building Context-Aware Applications. GeorgiaInstitute of Technology, 2000. (PhD thesis)

[Dey] Dey, A K, Abowd, G D. Towards a betterunderstanding of context and context-awareness.College of Computing, Georgia Institute of Technol-ogy. (GVU Technical Report GIT-GVU-99-22)

[E-SENSE] E-SENSE project website. Availablefrom: http//www.ist-esense.org/

[Giacomo05 et al] Cabri, G, Ferrari, L, Leonardi, L,Zambonelli, F. The LAICA Project: SupportingAmbient Intelligence via Agents and Ad-Hoc Mid-dleware. Proceedings of the 14th IEEE InternationalWorkshops on Enabling Technologies: Infrastructurefor Collaborative Enterprise (WETICE’05). Lin-köping University, Sweden, 13–15 June 2005, 39–44.

[Liscano03] Liscano, R, Ghavam, A. Context Aware-ness and Service Discovery for Spontaneous Net-working. AdhocNow 03, 2003.

[Lefteri et al] Tsoukalas, L H, Uhrig, R E. Fuzzy andNeural Approaches in Engineering. John Wiley,1997. (ISBN 0-471-16003-2)

[Niemegeers02] Niemegeers, I G G, de Groot, S H.From Personal Area Networks to Personal Networks:A user oriented approach. Journal on Wireless andPersonal Communications, 22, 175–186, 2002.

[Niemegeers05] Niemegeers, I G G, de Groot, S MH. FEDNETS: Context-Aware Ad-Hoc NetworkFederations. Wireless Personal Communication,Springer, vol 33, June 2005

[MAGNETB] MAGNET Beyond project website.Available from: http//www.magnet.aau.dk/

[MD2.2.1] Ghader, M et al. Resource and ServiceDiscovery: PN Solutions. MAGNET Deliverable2.2.1, IST-507102, Dec. 2005.

[MD2.2.3] Olsen, R L et al. Service, resource andcontext discovery system specification. MAGNETDeliverable 2.2.3, IST-507102, Dec. 2006.

[MDB121] Olesen, H et al. The conceptual structureof user profiles. MAGNET Beyond deliverable 1.2.1,IST-507102, September 2006.

[mmpr06] Olsen, R L, Schwefel, H P, Hansen, M B.Quantitative analysis of access strategies to remoteinformation in network services. To be published atGlobecom, 2006.

[Olesen et al., 2006] Olesen, H et al. Scenario con-struction and personalization of PN services basedon user profiles and context information. Invitedpaper, Joint MAGNET Beyond, e-SENSE, DAIDA-LOS and CRUISE IST workshop, Myconos, Greece,June 2006.

[Pascoe99] Pascoe, J, Ryan, N S, Morse, D R. Issuesin developing context-aware computing. Proceedingsof the International Symposium on Handheld andUbiquitous Computing, Karlsruhe, Germany, Sept.1999, Springer Verlag, 208–221.

[Pereira00] Pereira, J M. Fourth Generation: now, itis Personal. Proc. PIMRC 2000, 1009–1016, Septem-ber 2000.

[PerNets06] Sanchez, L, Lanza, J, Olsen, R L, Bauer,M, Genet, M G. A generic context managementframework for Personal Networking environments.Proceedings of First International Workshop on Per-sonalized Networks, San Jose, California, July 2006.

[Schilit94] Schilit, B N, Adams, N I, Want, R. Con-text-aware computing applications. Proceedings ofthe Workshop on Mobile Computing Systems andApplications, 85–90. IEEE Computer Society, SantaCruz, CA, 1994.

[Simplicity] Simplicity project website. Availablefrom: http://www.ist-simplicity.org/.

[W3N3] Notation3 (N3) A readable RDF syntax.Available from:http://www.w3.org/DesignIssues/Notation3.html

[W3OWL] W3C OWL overview. Available from:http://www.w3.org/TR/owl-features/


Introduction

The communications styles and paradigms are chang-ing and the pace is breathtaking. The past ten yearsfrom the mid 90s to the present day were all aboutthe Internet revolution and mobile phones. Thesetwo phenomena together changed our lives more thananybody could have ever imagined and many thinkthat such growth in the industry can never return.Others are not so pessimistic and see the huge poten-tial that lies in combining the Internet’s openness andthe wide range of emerging wireless technologies.The table is set up for cool applications and servicestaking advantage of rapidly growing bandwidth (inboth fixed and wireless) and the flexible packet net-working technology.

So far we have seen a promise of what the newtechnologies can deliver. Skype and Google are thebiggest success stories of this decade when it comesto Internet applications. It is tempting to say that theyare showing the way for others as well. Nevertheless,the two companies seem to have adopted very differ-ent strategies; while Google is eager to disclose itsApplication Programming Interfaces (API) as webservices [1], Skype has remained very much isolatedfrom the other players. Only time will tell whichstrategy will be more successful financially. Weargue nevertheless that Google’s way of disclosingits technology via web service APIs promotes innova-tion. Not everybody needs to re-invent the wheel, butplayers can utilise each other’s APIs to produce newapplications and services. This gives ground to a newkind of value net where the service provider needn’town and operate all the software components that theservice comprises, but can rely on third party webservice APIs. The dividends and cash flows withinsuch value nets are interesting topics of their own,but are outside the scope of this article, which is moretechnology oriented. The article describes a novelweb service named Personal Network Directory Ser-vice (PNDS). The main idea behind PNDS is to pro-vide means for user authentication, which is an ele-

mentary component of any Internet service. Theauthentication is bootstrapped using the GSM’s ShortMessage Service (SMS) after which public key cer-tificates [2] can be used for authentication andencryption of Internet communications. This resultseffectively in a so-called single sign-on architecture[3],[4] where the user is seamlessly authenticatedafter an initial user-assisted login to the system.Moreover, it is shown how the PNDS can be used toimplement a multi-user virtual packet networkreferred to as PN federation. To this end, two real-lifescenarios are presented, in order to illuminate thebackground and motivation of the technical design.Finally we compare PNDS briefly with existing sin-gle-sign on solutions and discuss further work neededto make use of the developed technology.

Personal Network

Personal Network (PN) is an emerging paradigmwhere user’s personal devices appear to form an iso-lated network regardless of their physical location[5]. Co-located devices organise themselves in so-called clusters and the clusters are inter-connectedusing virtual links (e.g. IPSEC tunnels). Virtual linkstypically exploit public networks (i.e. the Internet),whereas the intra-cluster links are either fixed (e.g.Ethernet) or wireless (e.g. Bluetooth, WLAN). Secu-rity and privacy are inherent properties of the PN andthe applications need not worry about them at all.Figure 1 gives an idea of the concept.

From a technical perspective PN is actually a VirtualPrivate Network (VPN) – only authorised devices areable to join. Compared to the traditional VPNs, thePN is perhaps much more advanced in terms of self-organisation capabilities and heterogeneity in thesupported networking environments, but its topologyis still bounded by the device ownership rather thananything else. There are basically two ways by whichthe ownership can be determined: by pre-sharedsecrets [6] or by a Public Key Infrastructure (PKI) [7]

Personal Network Directory Service

M I K K O A L U T O I N , S A M I L E H T O N E N , K I M M O A H O L A , J O R I P A A N A N E N

Mikko Alutoin is

a Research

Scientist at

Technical

Research Centre

of Finland (VTT)

Sami Lehtonen

is a Research

Scientist at

Technical

Research Centre

of Finland (VTT)

Kimmo Ahola is a

Senior Research

Scientist at

Technical

Research Centre

of Finland (VTT)

Jori Paananen is a

Senior Research

Scientist at

Technical

Research Centre

of Finland (VTT)

Personal Network (PN) is an emerging computer networking concept where security and privacy of

communications as well as user centricity are emphasised. In this article we put forward a novel web

service called Personal Network Directory Service (PNDS) that provides user authentication in the

Internet by acting as a trusted third party and certificate store. This is required to facilitate authenti-

cated inter-PN communications. In addition, we propose a method for building PN federations where

a group of PNDS users share a secure virtual packet network. We demonstrate via concrete screen-

shots how user creates a PNDS account, gets a PN certificate, and finally joins a PN federation.



based system. The former approach was thoroughlyexplored in the EU funded IST MAGNET project(FP6-IST-IP-507102) [8]. The PNDS, which is aPKI-like approach, is being studied in a continuationproject called IST MAGNET Beyond (IST-FP6-IP-027396) [9].

Personal Network Directory Service

PNDS consists of a web service API and a databasebehind it. This database stores information about PNsas well as PN federations and participants in thesefederations; who created it and who maintains it (i.e.has the ability to add or remove participants or editparticipant attributes).

Figure 2 shows the initial business model for thePNDS where only a single service provider is engag-ing in offering the service. It can be argued that thiscentralised architecture is vulnerable to single pointof failure and that it traps users to a single serviceprovider. The authors would like to denote that theultimate goal is a multi-provider environment wheredifferent PNDS service providers can jointly providethe service. However, for initial piloting the cen-tralised architecture may well be used to demonstratethe idea as well as the PN and PN federation concepts.

PN Certificates

Thus, the PNDS service provider(s) is a trusted thirdparty. It stores public keys and provides PN certifi-

Figure 1 Personal network

Figure 2 PNDS service provider acts as a trusted third party

Personal area

network

(PAN)

Foreign

PANHome

cluster

Internet

User 1

Service operator network

Home

cluster

InternetUser 2

PN directory server

PN

DS

AP

I

PN

DS

AP

I



cates for those public keys. A PN certificate bindstogether a public key and a PN name so that othersmay authenticate the user, provided that they trustthe certificate issuer and its ability to authenticate theuser to whom it has signed the PN certificate. There-fore, in order to deliver the certificate signing service,the PNDS must authenticate the user credibly. Other-wise, anyone could take over a well-established PNname and thereby steal one’s digital identity. To thisend, the user is required to create a PNDS account asshown in Figure 3.

The above data are sent via PNDS API’s new_usermethod call and the PNDS account is created in thedatabase. A random password is associated with theaccount so that the subsequent method calls from theuser can be authenticated. This PNDS password issent to the user via GSM’s Short Message Service(SMS) as illustrated in Figure 4. Use of SMS ensuresfairly reliable user authentication.

All the PNDS method calls are encrypted usingSecure Sockets Layer (SSL), so that the PNDS pass-word is never sent in clear-text through the Internet.The PNDS service provider itself can also be authen-ticated via SSL, if the PNDS client terminals have itsroot certificate. Except for the new_user method, allother PNDS methods require the user’s GSM Numberand the PNDS password as arguments. Therefore theuser needs to log in, before using the actual PNDSAPI and proceeding to fetch a PN certificate and cre-ate PN federations, for example. The main loginscreen of the PNDS client application is shown inFigure 5.

The above constitutes actually a so-called single sign-on system. The user signs on once to the PNDS clientand can be authenticated via PN certificates eversince, without having to introduce any further usernames and passwords. Figure 6 shows how the clientapplication fetches a PN certificate for a PN name.The PN certificate is written for the complete PNname which in this case is “+35840123456”. Notethat the PN name needn’t reveal the user’s GSMnumber, but PN certificates can also be written forpseudonyms, such as “knight_rider”. In this caseother users do not know who is behind the pseudo-nym, but the PNDS service provider can still storethis information for legislative purposes, for example.

PN Federations Using the PNDS

As mentioned already, the PNDS can be used tobroaden the PN concept to include also the so-calledPN Federations (PN-F) where two or more differentpersons set up a shared virtual packet network, inorder to achieve a common goal [10]. To make things

more tangible, we will first present two different sce-narios where PN federations could be used. Then wewill introduce a concept called PN federation profileand discuss its design in the light of the two scenarios.

Figure 3 Creating a PNDS account

Figure 4 The user’s PNDS password is sent via SMS

Figure 5 Logging in to the PNDS client application



Football Club Federation

A football club manager needs to keep in touch withthe players, in order to arrange the training times andprepare for the matches. He also wants to deliversome confidential material to the players and inquireinto their health status. To this end he sets up a PNfederation. The nature of the federation is long-term,since the team is quite static. Every now and then aplayer gets transferred, but then the manager can takeaction and erase the player from the federation andinvite the new player to the federation. The federationshould remain operational even if the manager is on aholiday with all of his personal devices switched off.Sometimes the manager wishes that the vice manageradministrates the federation for him.

Lecture Federation

A lecturer at the university would like a possibilityto set up a temporary ad hoc network for the durationof the lectures. This would be useful for distributingthe lecture material as well as to perform a scientificstudy among the students using a digital question-naire. Even small exams could be easily carried outusing laptops and the temporary network. To imple-ment this, the lecturer uses a PN federation. Becausethe nature of the virtual network is first and foremostad hoc, the lecturer does not want to rely on havingInternet access all the time. The federations are typi-

cally short-term and valid only for the duration of thelecture. The identities of the participants need notalways be verified, but anonymous access to thefederation might sometimes be perfectly acceptable.

PN Federation Profile

The PNDS issued certificates, discussed in the previ-ous section, provide a very good starting point forsupporting PN federations. Some additional datastructures still need to be defined so that the partici-pants of the federation can be authenticated andauthorised. For this the so-called PN federation pro-file is introduced. The PN-F profile is stored some-where in the IP network, for example in the PNDSservice provider’s database or in the lecturer’s laptop.The PN-F profile contains the following information:

• Name of the federation (and a corresponding PN-Fcertificate)

• Owner of the federation• Deputies (i.e. additional federation administrators)• Invitees (i.e. who are allowed to join the federation)• Who have joined the federation• Passphrase• Federation’s private key (corresponding to the

federation’s PN certificate)• Federation’s group key

Let us next discuss the items in the PN-F profile inthe context of the specific requirements they aredesigned to satisfy.

Authentication of a Federation

As a user seeks to join a federation, she must feelsecure in the sense that she is about to join just thefederation that she has intended. For example, nomalicious person should be able to “take over” a fed-eration and start gathering private data of its users.Another point is to be prepared for unintentional mix-up of two federations due to almost similar federationnames or a typing error from the user’s side. Fortu-nately certificates can be used to fight against thesethreats. A key to this is to realise that the federationname in the PN-F profile can be just a similar kind ofglobally unique identifier as is the PN name (see pre-vious Section). After all both PN and PN-F are virtualnetworks, so it is only logical to use the same kind ofnaming scheme. Just like any PN, also the PN-Fneeds to have a certificate from the PNDS, so that theparticipants may authenticate the federation whiletrying to join. We refer to this certificate as the PN-Fcertificate.

The federation’s name may contain the owner’s PNname, but it does not have to. It might just be that the

Figure 6 Certifying a public key using PNDS API

Figure 7 A student joining an ad hoc lecture federation

ClientServer

(e.g. https://pnds.vtt.fi)

Get_certificate

(User_id, password, public key, PN name)

ACK (PN certificate



federation owner does not wish to reveal his/herGSM-number (or not even PN pseudonym). Thus,instead of math-lecture.+35840123456 the namecan just be math-lecture, for example.

In the lecturer scenario, it could be that the CertificateRevocation Lists (CRL) [2] of the PNDS are notavailable, due to lack of Internet access. (CRLs areused to list certificates which have been revocated,because the corresponding private key has been com-promised, for example.) However, the lecturer canprepare for this by having the certificates automati-cally renewed on a daily basis, while sitting in heroffice where Internet access is available. This way thecertificate’s issuing date will be only a few hours oldand the students can be fairly sure that the federa-tion’s private key has not been compromised mean-while.

Admittance to a Federation

The owner of the federation has maximum accessrights to manipulate the PN-F profile. The owner canalso manipulate a list of additional federation admin-istrators, in order to give access rights to somedeputies, for example the vice manager of the footballclub. The owner and deputies are identified by theirPN names. These fields are primarily there to autho-rise the manipulation attempts of the PN-F profileitself, but they can also be shown to the federationparticipants, if the owner so wishes.

By default the federation is empty. Everyone, includ-ing the owner, needs to explicitly join the federation.For authenticating participants, while they first tryto join the federation, there are two partially over-lapping mechanisms: PN certificates and federationpassphrases. To join a PN-F, the user always needs tohave a PN certificate (i.e. a certified public key), butthe certificate need not necessarily be issued by thePNDS – also self-signed certificates can be used incases where the federation owner wishes to allowanonymous participation in the federation. When aPNDS-issued certificate is used however, it offers anice way to authenticate the origin of the join request.To this end, there is an optional list of invitees, inwhich the owner may record the PN names of thepeople who are authorised to participate in the federa-tion. For example, when creating a federation for thefootball club, the club’s manager may invite only themembers of the club to the federation and then storetheir PN names to the invitee list. PNDS then sendsan SMS-invitation to the invitee where he is advisedto join the federation. As the club members try to jointhe federation they will identify themselves with avalid PN certificate. After verifying the certificate,it is checked whether the PN name is included in the

invitee list or not. In the case that the name is foundin the list, the request is automatically authorised. Ifthe PN name is not in the invitee list, then the requestcan be either rejected or left pending. A pendingrequest is one that the system tries to verify from thefederation’s owner subsequently. The owner can ofcourse deny the system to generate any pendingrequests altogether.

There are however cases when the invitee list and/orpending requests are not satisfactory. Think of thelecturer at the university, for example. She would liketo set up a temporary PN-F for the duration of the lec-ture. The problem is that the number of participants isoverwhelming and listing their identities in the invi-tee list beforehand is not something that the lecturerenjoys doing. In this case the federation passphrasecomes to the rescue; the lecturer announces the feder-ation name and the passphrase to the students. Anystudent can now enter the federation with a self-signed certificate by merely typing the federationpassphrase when prompted for it. Omitting the wholepassphrase from the equation would not be such agood idea, since then the number of mix-ups wouldincrease. This is because in the university campusthere could be multiple federations to which users canjoin accidentally, if the passphrase is not checked.

Federation Participation Certificate

After the mutual authentication and the admittanceof the join request, the federation participant shallbe able to operate fully within the federation. Thismeans that all other federation participants should beable to verify that the participant has been admitted.This could be done in various ways. One optionwould be to check the PN-F profile. This would how-ever fail if the PN-F profile cannot be accessed rightthen. To remedy this, we introduce PN-F participa-tion certificates which are obtained along with thereturn message of a successful join request. SeeFigure 8, for example.

The PN-F participation certificates are usually signedby the PNDS, but in the ad hoc case also the PN-Fowner can sign them. In the latter case the chain of

Figure 8 Student fetching a PN-F participation certificate

ClientServer

(e.g. https://pnds.vtt.fi)

Join (PN-F name, PN certificate, passphrase)

ACK (PN-F participation certificate)



trust goes as follows: PNDS root certificate > PN-Fcertificate > PN-F participation certificate. In the for-mer case, the middle part can be omitted. In the foot-ball club federation, it is the PNDS which signs thePN-F participation certificates as it is handling theadmittance to the federation (according to the inviteelist). In the lecture scenario, it is the lecturer. Thename, for which the PN-F participation certificate ismade, can be, for instance, Laura@math-lecture oranonymous012@math-lecture.

Message Authentication

A federation is fundamentally a network layer con-cept. This means that the federation of a message isrevealed by the IP layer packet headers. This way theoverlapping federations interfere with each other aslittle as possible: messages of unknown federationscan be discarded in the network layer without furtherprocessing. But how is a message’s federationauthenticated? A scheme [11] where each IP packet issigned and the signature is included as an IPv6 exten-sion header could be used. This offers a nice solution,but only for IPv6 based communications. Anothersolution is to implement a layer 2.5 (between MACand IPv4) for conveying the signature. But should themessages be signed with the participant’s private keyor with the federation group key which all the partici-pants share? The latter option is somewhat morestraightforward, as it does not require the recipient toknow the sender’s public key. However the formeroption provides more security. Probably both optionsshould be supported.

Encryption of Communications

Typically the federation communications should beencrypted by leveraging the public key of the mes-sage recipient. Before the encrypted message can besent, it could be checked whether the recipient holdsa corresponding PN-F participation certificate. Forgroup communications, such a method is neverthelessnot very useful. Consider the lecture scenario wherethe students are taking an exam. The lecturer sendsthe questions for everyone using a broadcast messageover the ad hoc radio network. This saves a lot ofbandwidth compared to each student individuallyfetching the questions from the lecturer’s laptop. Ifthe lecturer wants to encrypt the questions for somereason, it is not possible to use any participant’s pub-lic key, of course. Instead, the federation group keymust be used for encrypting the broadcast messages.(A similar scenario, where encryption adds morevalue, is an ad hoc meeting in an airport lounge wherethe meeting chair distributes a confidential memo tothe other participants.)

Related Work

There exists schemes which share some of the char-acteristics of the PNDS. In Liberty Alliance frame-work [3], for instance, it has been defined how agroup of web service providers may federate theiruser accounts. The service providers seek to obtainknowledge about each other’s user account databases(with a permission from the user). This is called iden-tity federation. As the providers trust each other (viaa circle of trust), they can offer the so-called singlesign-on user experience: user needs only sign-on toone service provider after which he can be seamlesslyauthenticated to the others as well. This is useful,because today the various web accounts of a user arein pieces all over the Internet and the only connectionbetween them is the user. (Fortunately many webbrowsers, such as Mozilla Firefox, can nowadaysremember passwords for various web logins, whichhelps a bit.)

The main idea in Liberty Alliance is that a web ser-vice provider can rely on a trusted third party, calledidentity provider in this context, when it comes touser authentication and authorisation. As a user triesto access a service via HTTP, the HTTP server asksthe identity provider to authenticate the user. This isachieved using HTTP redirection. After the identityprovider has authenticated the user, it provides a cre-dential called artefact for the HTTP client. The clienttries to use the service again, but this time providesalso the artefact to the server. The server contacts theidentity provider in order to map the artefact to a useridentity. Then, after knowing the user’s identity, theserver can proceed to the authorisation phase.

The Liberty Alliance framework is clearly definedfrom the service provider perspective; it assumes aclear distinction of providers and consumers. How-ever, when it comes to PN federations such a divisiondoes not exist and therefore the architecture for PNfederations should be more peer-to-peer oriented. More-over, the Liberty Alliance framework is specificallydesigned for HTTP, whereas the PNDS is more pro-tocol agnostic. The most distinct difference is never-theless with respect to ad hoc and group communica-tions: clearly the Liberty Alliance as such does notmeet the requirements of, for example, the lecturefederation presented in this article. This is becauseidentity provider is not allowed to be offline and be-cause group communications are not dealt with at all.

Windows Live ID (previously known as .NET Pass-port or Microsoft Passport Network) [4] is a singlesign-on system for web systems. Live ID is based onauthentication server, which asks for username andpassword. The server returns a time-limited GLOB-ALAUTH-cookie. The authentication server sends to



a user an ID-tag that the authentication and the com-merce server had previously agreed upon. The com-merce server will send a LOCALAUTH-cookie to theuser as return to the ID-tag. The need to authenticateis thus handled as long as these cookies are valid.When the user logs out of Live ID, the cookies areremoved.

Other identity protocols (based on URLs) are openID[12], Yadis [13], Light-Weight Identity (LID) [14].The main differentiator between the PNDS and othermeans of identity management is that even though theauthentication is done with the help of the serviceprovider, the mechanisms (i.e. certificates) are usefulalso when the service provider is not accessible, e.g.in ad hoc networks.

The 3GPP is working on the so-called GenericAuthentication Architecture (GAA) which defineshow subscriber certificates are obtained from amobile operator’s Certificate Authority (CA) [15].The 3GPP approach is in general very similar to theone described in this paper. There are some majordifferences however. Firstly, the architecture in [15]is more decentralised; a number of CAs, referred toas PKI portals, are used to write certificates. In orderto do this, a PKI portal needs to contact the HomeSubscriber System (HSS). In the PNDS approach thesubscribers fetch the certificates from one central CAwhich also maintains the subscriber records. Anotherdifference is how CA authenticates a subscriber. In[15] the authentication is based on the private keythat is stored in the subscriber’s SIM-card. We useSMS-based authentication. In the 3GPP GAA the cer-tificates are fetched using HTTP Get (with Digestauthentication) whereas we use XML-RPC [16] overSSL and the PNDS password as such for subscriberauthentication. The CA’s root certificate is hardcoded in the XML-RPC client. Finally, the conceptof PN federations is not addressed in the 3GPP GAAin any form.

Conclusions and Further Work

This article introduced a novel web service calledPersonal Network Directory Service (PNDS) that canbe used to support federations of Personal Networks(PN). The approach is to utilise the described PNDSAPI for first retrieving certificates for one’s publickey, and then for joining to the PN federation. It wasillustrated how a chain of trust can be built in order tofacilitate ad hoc federations and how communicationswithin the federations can be authenticated and en-crypted. The support for ad hoc scenarios and groupcommunications in general is what distinguishes thePNDS from the existing single sign-on solutions.

Currently the authors are working on supplementingthe PNDS API implementation. So far methods forcreating PNDS account, getting PN certificates andcreating federations have already been implemented,as well as the counterpart client software of which thescreenshots above were taken. The next steps willinvolve implementing the Join-method. The userexperience that we have managed to create is so farvery satisfying. PNDS is however only one part ofthe architecture on which federations are built. Tomake federations happen, further work is required inthe areas of mobility management, middlebox traver-sal, and service discovery. Also applications thatleverage the capabilities of the PN federations need tobe built. When it comes to the ad hoc federations,advances in MANET protocols, such as in duplicateaddress detection and IP address autoconfiguration,are required. The majority of these issues will bedealt with in the remaining period of the MAGNETBeyond IST project where a PN pilot system is beingbuilt. The pilots shall be conducted during the firsthalf of 2008.

References

1 Google APIs. http://code.google.com/apis.html

2 IETF. Housley, R, Polk, W, Ford, W, Solo, D.Internet X.509 Public Key Infrastructure Certifi-cate and Certificate Revocation List (CRL) Pro-file. RFC 3280, April 2002.

3 The Liberty Alliance project. http://www.pro-jectliberty.org

4 Windows Live ID. http://www.live.com

5 Jacobsson, M et al. Network Architecture for Per-sonal Networks. 14th IST Mobile & WirelessCommunications Summit 2005, Dresden, Ger-many, 19-23 June 2005. European InformationSociety Technologies (IST) programme. Dresden(2005), 5 p.

6 IST MAGNET Project, Deliverable 4.3.2. Finalversion of the Network-Level Security Architec-ture Specification. http://www.ist-magnet.org/pub-lic+deliverables/phase1wp4, Feb 2005.

7 IETF. Chokhani, S, Ford, W, Sabett, R, Merrill,C, Wu, S. 2003 Internet X.509 Public Key Infra-structure Certificate Policy and CertificationPractices Framework. RFC 3647, Nov. 2003.

8 Hoebeke, J et al. Personal Networks: from con-cept to a demonstrator. Proceedings of the ISTSummit 2006, Myconos, Greece, June 2006.



9 IST Project MAGNET Beyond (IST-FP6-IP-027396). http://www.ist-magnet.org

10 Hoebeke, J et al. Personal Network federations.Proceedings of the IST Summit 2006, Myconos,Greece, June 2006.

11 Candolin, C, Lundberg, J, Kari, H. Packet levelauthentication in military networks. In: Proceed-ings of the 6th Australian Information Warfare &IT Security Conference, Geelong, Australia,November 2005.

12 Recordon, D, Fitzpatrick, B. OpenID Authentica-tion 1.1, May 2006. http://openid.net/specs/openid-authentication-1_1.txt

13 Miller, J. Yadis Specification version 1.0. March2006. http://yadis.org/papers/yadis-v1.0.pdf

14 Light-Weight Identity (LID).http://lid.netmesh.org/wiki/Main_Page

15 3rd Generation Partnership Project, TS 33.221version 6.3.0, Technical Specification Group Ser-vices and System Aspects, Generic Authentica-tion Architecture (GAA). Support for subscribercertificates (Release 6), March 2006.

16 XML-RPC Homepage. http://www.xmlrpc.com

Mikko Alutoin entered Helsinki University of Technology (HUT) in 1995 and received his MSc in Electrical

Engineering, majoring in telecommunications in March 2000. He did his Master’s thesis for Nokia Networks,

where he worked as software engineer for a few years during 1998–2001. Since 2001 he has been working

as Researcher at Technical Research Centre of Finland (VTT) and has participated in a number of inter-

national EU research projects. He is also a post-graduate student at HUT.


Sami Lehtonen started studying information technology at Lappeenranta University of Technology in 1996.

His working career began at VTT in 1999 in the Networks group. Besides working he continued his studies

and graduated as a Master of Science in technology in April 2003. He is working at VTT in the Security

team as a Research Scientist. In June 2006 he got his CISSP (Certified Information System Security

Professional) certification #94049. He was a member of the information security risk assessment group

that operated under the Finnish National Information Security Advisory Board in the Finnish Ministry of

Traffic and Communications. Later he was a member of the Committee on Information Security in Critical

Infrastructure at Finnish Communications Regulatory Authority.


Kimmo Ahola received his MSc degree (telecommunications) from the University of Jyväskylä in 1997. He

has worked ever since at Technical Research Centre of Finland (VTT) on various international (e.g. EU,

EURESCOM) and national projects concerning mainly IP systems and protocols. Currently he is Senior

Research Scientist and team manager in the Adaptive Networks team.


Jori Paananen received his MSc degree (telecommunications) from Helsinki University of Technology (HUT)

in 1983. Since then he has worked at Technical Research Centre of Finland (VTT) on research projects in the

area of mobile and IP software systems and protocols. Currently he is Senior Research Scientist in the

Adaptive Networks team.



1 Introduction

The paper is organized as follows. Section 1 is anintroduction to risk analysis in a wireless environ-ment and presents the structure and the sections ofthe paper. In Section 2 we present typical cases andscenarios. In this initial study, we present two cases;namely the multi-operator diversified radio environ-ments and the Personal Networks. The reason whywe focus on these two is that they both constructmajor categories where networking environments andtopologies can be classified. In other words, if assetidentification and harm estimation are performed forthese scenarios, then every networking architecturecan be addressed by referring to these models. InSection 3 we identify the network resources and theassets. These will be considered in the subsequentsteps in terms of failure probability and harm estima-tion. Assets for network operators are not only tangi-ble, e.g. system components, but intangible as well.Under intangible assets we should consider usersatisfaction, operator’s reputation, etc. All assetsare placed in a tree structure so that we can easily seethe impact of an outage in a network component. InSection 4 we estimate the threat likelihood and therelated harm. In regard to the threat likelihood thereis a categorization based on the frequency of occur-rence of a failure or outage; however, it is importantto asses the probability for a given duration. This ismeaningful, as a risk assessment study should referto a specific time-frame, e.g. one calendar year.In Section 5 we proceed with the risk assessmentmethodology. This is a set of equations that calculatethe total exposure, which is the final outcome of therisk assessment exercise. In Section 6 we present anexample of risk analysis. This example is neither themulti-operator diversified radio nor the Personal Net-work; however, it shows the proposed methodologyfor a UMTS environment. This is because the above-mentioned scenarios are complicated configurationsand in order to show how the methodology is struc-tured, we have described a UMTS scenario and then

we extend the study to the complicated situations. Inthis example we place the network components of aUMTS network in a tree structure and we calculatethe probability of an outage in any node, as well asthe harm. For each node of the tree both the probabil-ity and the harm are calculated if we list all kinds ofthreats, e.g. denial of service, unauthorized access,and we assess the probability and the resulting cost.Finally, in Section 7 we sum up with the conclusionsand the future work.

2 Use Case and Scenarios

There are several use cases and scenarios associatedwith the proposed procedure, however we clusteredthem into two categories; namely (a) multi-operatordiversified radio environment and (b) Personal Net-works. The reason for this split is that they have adifferent philosophy in terms of who operates thenetwork and if this is based on the concept of basestation – mobile terminal and/or ad hoc communica-tion. However, in Sections 3-6 we analyze the exam-ple of a UMTS network that is more tangible and weexplain how this can be extended for the two usecases described bellow.

2.1 Multi-Operator Diversified Radio

Environment

The development of wireless systems has evolved inan unimaginable way during the last two decades. Incellular wireless systems the so-called First Genera-tion (1G) is already antiquated. The dominant Gener-ations, which are nowadays in the limelight, are 2G,2.5G and 3G. In Europe their representatives areGSM (Global System for Mobile Communication),GPRS (General Packet Radio Service) and UMTS(Universal Mobile Telecommunications System)respectively and they belong in the terrestrial widearea cellular systems. The circuit-switched GSM pro-vides very slow data rates (9.6 – 14.4 kb/s) to satisfythe burst applications, even after the appliance of

Risk Analysis in an ‘Insecure Wireless World’

S O F O K L I S K Y R I A Z A K O S , N E E L I P R A S A D

Sofoklis

Kyriazakos

is Assistant

Professor at

Aalborg Univer-

sity, Denmark

Neeli Prasad is

Associate Pro-

fessor at Aalborg

University,

Denmark

Risk analysis is the process that each operator should go through to determine the risk exposure. This

risk is linked with the possibility of the damage that could happen either from intruders or by misuse of

the resources. The goal of risk analysis is to determine the probability of potential risks and estimate

the overall damage at an annual basis in order to define new policies and measures. In the wireless

world the major feature is the air-interface, however these systems also follow the same practices

to determine the risk probability, while emphasis should also be given to the characteristics of the

transmissions and the related vulnerabilities. In this paper we deal with the issue of risk analysis for

wireless systems regardless of the access technology.



High Speed Circuit Switched Data (HSCSD), it doesnot overcome the limit of 40 kb/s. Packet-switchednetworks, based on the access network of GSM withactual changes only in the core network (GPRS),appeared with the promise of higher bit rates (theo-retically 172 kb/s), but in practice the maximum bitrate achieved is about 45 kb/s.

The UMTS access network follows a differentapproach, in comparison to GSM and GPRS, makingthe achievement of higher data rates more feasible.UMTS typically offers data rates up to 384 kb/s,even if in theory a 2 Mb/s transfer rate is possible.Nevertheless, the actual performance of UMTS hasstill to be verified during real operation conditionswith heavy network loads.

On the other hand, there exist various wireless sys-tems such as global area systems (e.g. satellite sys-tems), wireless personal area networks (WPANs),which are formed by wireless communicationsbetween devices using technologies such as

Bluetooth [1] or IEEE 802.15 [2] and WLANs(e.g. IEEE 802.11a, IEEE 802.11b or HIPERLAN)[3][4][5]. These kinds of network provide incompara-bly high data rates. For example the 802.11b WLANprovides throughput up to 5 Mb/s, while the datarates in 802.11a can be up to over 25 Mb/s, with theperspective of reaching in the future the inconceiv-able, for today, limit of 155 Mb/s.

The co-existence of these technologies results in aheterogeneous set of wireless communications sys-tems. Its active components are based on differenttheoretical backgrounds and are optimized for differ-ent ranges, exposing a great challenge for potentialco-operation of all existing and emerging systemsin a complementary way, in the concept of a 3GB(beyond 3rd Generation) system [6]. Such a hybrid3GB system is examined in this paper, focusing onthe efficient interworking of three different cellularaccess networks (GSM, GPRS, UMTS) with WLANsunder a unified and hierarchical resource manage-ment model. At this point it has to be mentioned thatthe 3rd Generation Partnership Project (3GPP), thebody that drives the standardization work for 3rdGeneration mobile communications, has already fore-seen the need for cooperation between WLANs and3rd Generation systems and has published a feasibil-ity study regarding their interworking [7].

Enhanced IP networking technologies are used tointegrate current and future systems to a unifiedsuper-network, enabling a truly seamless mobileInternet, beyond the simple wireless access to theInternet, thus extending the scope of a monolithicsystem. The Internet Protocol version 6 (IPv6) doesnot only offer virtually unlimited address space, butalso constitutes the technical foundation for evolu-tionary networking, offering also interoperability andinterconnectivity with respect to security, mobilityand Quality of Service (QoS) [8].

The major challenge in such a heterogeneous net-working environment is to exploit the advantages ofWLAN systems focusing on their seamless integra-tion in composite radio environments. More specifi-cally, a WLAN network is easy to deploy in hot-spotareas and can offer high data rates, thus achievingincreased QoS, while the installation costs are lim-ited. However, the subscriber should be able to roambetween different access technologies seamlessly.In addition, it is a common research scenario to haveseveral operators offering access to the same geo-graphical area in a cooperative scheme. This mightinvolve vertical-vertical handovers that occur whena user of one operator that makes use of resources ofa specific RAT moves seamlessly to another operatorby accessing resources of another type of RAT.

Figure 1 A diversified radio environment consistingof several access networks of one operator

B3G

UMTS

WLAN

GSM/GPRS



2.2 Personal Networks

There are several architectures of Personal Networks.The one described in this section is based on the MAG-NET Beyond [9]. Figure 2 shows the system architec-ture from a high level point of view. In this scenario auser owns several clusters. Each of the clusters formsan ad hoc local area network. All clusters togetherform a global, always for the user accessible network.

The architecture is described in more detail in [10],while routing in the PN is explained in [11]. In orderto be able to perform a risk assessment study, the sys-tem environment should be described, because threatsdepend on it. The geographical location is important,because a network, which has its elements far apart,will probably have more threats to face than a localnetwork. Also, there probably will be more threats toa wireless network in a crowded city than in a wire-less network in a rural area. Besides the location, theactors are very important. Who and/or what is ableto interact with the system in what way?

The PN’s location can be anywhere and everywhere.Because of the ubiquitous nature of the network, itscomponents can find themselves almost everywhere,

where the user wants them to be. This means the PNcan be acted upon by almost everything and everyoneyou can think of. However, we can distinguish a fewsituations where the clusters of the PN can be located:

• at home (Home network)• at work (Corporate network)• in your car (Vehicle network)• on body (Core PAN or PPAN)• any other remote cluster

The clusters are interconnected via tunnels over wire-less and/or over a wired interconnection infrastruc-ture. This interconnecting structure is shared by lotsof other PNs from different people. The MAC proto-cols used for the sharing of this structure are veryimportant both from a Quality of Service as well assecurity perspective.

The primary actor in the PN is the owner of the net-work. Other actors can be network providers, serviceproviders, others requesting services from the PN,anyone who the user sends a message or anyone whosends a message to the user, etc. In a specific use casethe actors can be described in more detail.

Figure 2 The MAGNET Beyond architecture for Personal Networks [9]

PANPersonal PAN

Local

foreign

devices

Smart building

Home cluster




Corporate cluster


Vehicular clusterInterconnecting structure

Internet, UMTS, WLAN

Ad Hoc, etc



PN is an omni-technological concept. In principle,every existing wireless or wired technology can beused and the network must be compatible with futuretechnologies. The goal is that all these different tech-nologies can interact seamlessly.

From the network requirements, architecture andenvironment, a list of network parts, which willrequire security attention, can be made. The require-ments listed below are requirements found directly inthe system requirements presented in [9] or derivedfrom them or from the network architecture and envi-ronment.

• All wireless radio links should provide fair MACschemes. Also wireless links can be weak, becausepotentially, anyone can pickup the signal, as it isbroadcast in every direction.

• The first thing two nodes must do before they canestablish a secure link, is to communicate witheach other over an insecure link. This communi-cation must establish a secured link. The way thisinsecure communication happens is very importantfor the system security, because a piconet shouldbe joined by a hostile node.

• Only trusted nodes are allowed to have access totrusted data. This implies that different levels oftrust have to be assigned to data.

• The packets will flow through the network usinga multi-hop IP protocol. This multi-hop routingshould be done securely. Basically every routeris a potential access point for an attacker.

• All nodes should be able to automatically setupclusters and a PN. This means that this automaticsetup should check the trustworthiness of othernodes and clusters, because there is no user to checkit. However, another requirement is that the usershould have a single GUI to manage all deviceswhich are part of the PN. Should the user alwaysbe asked whether a node should be accepted, evenafter the system has found the node to be trust-worthy?

• In case of handover, the target network should betrusted.

• Between the cluster connection points, which areat the gateway nodes, of a PAN to other clustersa secure link should be established.

• Remote cluster discovery and authentication shouldbe done securely.

• The network’s topology will change constantlydue to mobility or entering and leaving nodes.This means that the roles of nodes may constantlychange, for example the function of the gatewaynode may switch to other devices. Every time achange is happening the secure links will have tobe established again. This means that every topol-ogy change is a potential moment of weakness,which attackers can take advantage of.

• End user’s sensitive data stored in personal devicesmust be protected such that no disclosure of data isallowed without permission of the user.

• PN must provide a mechanism which hides privateservices from foreign nodes.

• Because Service Discovery (SD) is one of the mainfeatures of the PN it requires extra attention. Thereshould be different trust levels for services, whichdefine which data can be used for the service. Con-text Discovery is a special case of SD and containspersonal information which should be kept confi-dent.

• Every device which has special privileges, forexample a master of a service or a gateway node,should have a power check. Before running out ofpower the device should send a message to the restof the cluster that someone has to take-over thefunctionalities. This message should be sent, suchthat there is enough time for the cluster to relocatethe functionalities to other nodes. If this is not done(thus, when the device loses power before take-over), is it then possible for another (hostile) deviceto take its place, without the rest of the networknoticing? A usual take-over should be done byauthentication of the two nodes: the old and thenew one.

• Sleep mode is a power saving mode. This impliesthat the security measures running on a device insleep mode can’t be too heavy. And how do youprevent deprivation attacks?

• Communication with a foreign PN will require agood policy and model on trust.

3 Asset Identification

An asset can be a tangible item, a grade or level ofservice, staff or information. The question that shouldbe answered is what needs to be protected. Followingasset groups can be identified:

• Networking infrastructures and equipment• Software



• Operator’s reputation• Confidentiality of information (protection of

databases, voice calls, SMSes, etc.)• Availability of Resources and Services• Integrity of information

All the above assets are then presented in a structuredtree, either directly or indirectly. Network compo-nents are presented as leafs in the tree, while reputa-tion, churn, etc, are considered in the harm estima-tion. The tree has several levels, each one describinga network layer. For instance, Level 1 might be thephysical layer that can be the air-interface and trans-mission network in a cellular system. In Level 2 weneed to breakdown the further components of the air-interface and transmission network respectively. Thesame happens to Levels 3, 4 and all hierarchical lev-els that we initially decide to split the complete sys-tem. As we move towards the end of the tree we willend up with components such as databases, e.g. HLR,SMS servers that are network assets and might bevulnerable.

4 Threat Likelihood and Harm

Estimation

Multiple threats can be associated with one asset.Only those threats that could reasonably be expectedto occur, or those that will result in identifiable con-sequences should be considered when performing theexercise of risk analysis. Source of the threat may beused in determining its probability.

Following rating convention is proposed:

• Negligible• Very low (2-3 times every five years)• Low (once every year or less)• Medium (every six months or less)• High (once every month or less)• Very High (multiple times per month or less)• Extreme (multiple times per day)

However, after the initial threat likelihood categoriza-tion, it is required to asses the probability in percent-age rates referring to a given duration.

In Figure 4 one can see the tree with the networkassets and a practical way to estimate the threat likeli-hood to have an attack in one of the network compo-nents, i.e. the leaf in Level 4. This might be a specificdatabase. If the initial classification is High, i.e. oncea month and we refer to a calendar year the probabil-ity of having such an incident is 3.29 %, assumingthat the impact of an incident has practically one dayduration. One of the major advantages of structuringthe assets in tree infrastructures is that quite often the

failure probability that depends on a shortcoming onanother component can be calculated by multiplyingfailure probabilities of the above nodes. However,this is not applicable for non depending failures.

5 Risk Assessment Methodology

The risk assessment exercise is based on the follow-ing calculation. The goal is to estimate the profit lossof the operator from a probabilistic perspective sothat the operator can calculate to what extent he caninvest on security.

Following parameters are defined:

Figure 3 Network asset tree

Figure 4 Path calculation

Level 1

Level 2

Level 3

Level 4

Level 1

Level 2

Level 3

Level 4



K : total exposure (¤)

fi : Failure in asset i

ki : exposure for failure in asset i

p(fi) : probability of failure in asset i (%)

hi : harm in asset i

(1)

ki = p(fi) * hi (2)

(3)

Therefore, the operator needs to estimate the proba-bility of having a failure in each network asset andestimate the related harm. It should be highlightedthat failure probabilities and costs are only presentedfor ‘leafs’ that have no ‘children’ (additionalbranches below).

6 Example of Risk Analysis

In this section we analyze the examples of UMTS andthe extensions to multi-operator diversified environ-ment and PAN. In the examples we form the relevanttrees and we calculate the threat likelihood and theharm cost.

6.1 Risk Analysis in UMTS Networks

Following example shows in a very simply way themethodology used to estimate the costs as a resultof a failure in a UMTS network in a period of sixmonths. Based on the initial study, i.e. analysis ofdata from the data-warehouse we have listed theapproximate probabilities of failure in each of thenodes, as well as the recovery costs and other costsrelated to the incident.

Although the failure possibility of a node can beretrieved from a statistical datawarehouse of the oper-ator it is always recommendable to perform this exer-cise by calculating the threats. In this example wewill examine one of the nodes, namely HSS and wewill estimate the cost of such harm (Table 1).

The above calculation should be carried for each ofthe elements presented in Figure 5, so that it is possi-ble to make the calculations, as these listed in Table 1.

At this point it has to be highlighted that the costsare only indicative and they might significantly varybetween operators vendors. By applying equation (3)for the above table we have:

= 348,000¤

This can be translated as the costs that result from thefailure of the network over a period of six months.Therefore, the operator should consider taking mea-sures that will minimize the above amount and thecosts of these measures should be obviously less.

6.2 Risk analysis in multi-operator

diversified radio environment

and PN-network

In the previous section we have seen a simplifiedapplication of this risk assessment methodology forUMTS environments. In the multi-operator diversi-fied radio environment scenario, the first differenceis that there should be a Level where the RAT tech-nologies are presented. These can be for instanceGSM/GPRS, WLAN, SB3G, etc. Under each of thesenodes, the complete tree should be presented in thesame way as for UMTS. Apart from the differentRATs, the existence of one more Level on top isrequired to define the operator. Under each differentoperator that has an SLA with the serving operator,there is no need to have the whole hierarchy under-

K =∑

i=5,6,8,...,13

p(fi) ∗ hi =∑

i=5,6,8,...,13

ki

K =

∑

i

ki

K =∑

i

p(fi) ∗ hi

Figure 5 Risk analysis in a UMTS network

Network

Access networkCore network

CS/CN HSS IMS PS/CN Node B RNC

MGW MSC GGSN SGSN

1

2 3

4 5 6 7 8 9

10 11 12 13



neath, just to know the probability of a total failure.To complete the methodology, apart from the calcula-tions presented in 6.1, we should consider the failureprobabilities from the other operators and otherRATs, since the traffic that cannot be served due to adamage can be shifted to another network segment.

In the case of PN and PAN this is more complicatedand the correct model that should be applied varieswith the scenario. First we have to define the Per-sonal PAN and the threats that result from vulnera-bilities in the sensors and devices belonging to thisgroup. Other nodes might be all other PANs that oneuser can have access to or be accessed by, since thesemight lead to attacks and damages. Finally, theunderlying telecommunication networks should alsobe considered, since vulnerabilities there might resultin severe attacks that should be calculated in the riskassessment.

Both scenarios will be examined in detail under theframework of the future research activities of theauthors for risk assessment.

7 Conclusions

In this paper we have proposed a simple methodologyfor risk assessment that can be applied in any networkoperator. The advantages and novelty of the methodare the classification of all network assets into a treeand the calculation of the failure probability for agiven time-frame. The harm estimation is based onseparate studies that should be carried out, while atthe end the end-result is the estimated cost that theoperator will pay due to the network vulnerabilities.This is a good indication of the investments that anoperator should make in order to minimize the failureprobability and increase network performance. Themethod is accurate since it can be validated withexisting data and well structured, as not only theassets, but also the threats are classified and weighted.

In the future, this methodology will be implementedin a simulator that will be fed with data from a datawarehouse.

References

1 The Bluetooth Special Interest Group (SIG).Specification of the bluetooth system, version 1.1.2001. Available at http://www.bluetooth.org

2 Roberts, G. IEEE 802.15 Overview of WG andTask Groups. Available athttp://grouper.ieee.org/groups/802/15/pub/Tutorials.html

3 IEEE. Supplement To IEEE Standard For Infor-mation Technology- Telecommunications AndInformation Exchange Between Systems- LocalAnd Metropolitan Area Networks- SpecificRequirements- Part 11: Wireless LAN MediumAccess Control (MAC) And Physical Layer (PHY)Specifications: Higher-speed Physical LayerExtension In The 2.4 GHz Band. 2000. (IEEEStandard 802.11b-1999) (ISBN 0-7381-1811-7)

HSS threats Probability Cost Weighted

Cost

1) Unauthorized access to data 0.20 % 680,000¤ 136.000¤

a) violation of confidentiality - - -

b) eavesdropping traffic or

control data - - -

c) masquerading - - -

d) traffic analysis - - -

e) browsing, inference, etc. - - -

2) Threats to integrity 0.10 % 196,000¤ 19,600¤

3) Denial of service 0.15 % 80,000¤ 12,000¤

a) intervention: jamming

or protocol failures - - -

b) resource exhaustion - - -

c) abuse of services - - -

4) Repudiation 0.15 % 94,000¤ 14,100¤

5) Unauthorized access

to services 0.08 % 750,000¤ 60,000¤

6) Physical damage 0.32 % 1,745,000¤ 558,400¤

Total 1 % - 800,000¤

Table 1 HSS treat estimation

i f(i) p(fi) h(i) k(i)

1 Network - - -

2 Core Network - - -

3 Access Network - - -

4 CS-CN - - -

5 HSS 1 % 800,000¤ 8,000¤

6 IMS 5 % 2,000,000¤ 100,000¤

7 PS-CN - - -

8 Node B 3 % 480,000¤ 14,400¤

9 RNC 1 % 3,000,000¤ 30,000¤

10 MGW 2 % 1,200,000¤ 24,000¤

11 MSC 3 % 3,600,000¤ 108,000¤

12 GGSN 2 % 1,600,000¤ 32,000¤

13 SGSN 2 % 1,600,000¤ 32,000¤

Table 2 Risk Assessment Table



4 IEEE. Archived – Supplement to IEEE standardfor information technology telecommunicationsand information exchange between systems –local and metropolitan area networks – specificrequirements. Part 11: wireless LAN MediumAccess Control (MAC) and Physical Layer (PHY)specifications: high-speed physical layer in the 5GHz band. 1999. (IEEE Standard 802.11a-1999)(ISBN 0-7381-1809-5)

5 ETSI. Broadband Radio Access Networks(BRAN), HIPERLAN Type 2, System Overview.ETSI TR 101.683 (VI.1.2), 2000.

6 Tselikas, N et al. Architectural framework forresource management optimisation over heteroge-neous wireless networks. In: Proc. ITCom (Infor-mation Technologies and Communications) SPIEconference, Orlando, Florida, USA, 7-11 Septem-ber 2003.

7 3GPP, TSG Services and Systems Aspsects. Fea-sibility study on 3GPP system to Wireless LocalArea Network (WLAN) interworking. 3GPP TR22.934 (V6.2.0), 2003.

8 Stallings, W. IPv6: The New Internet Protocol.IEEE Communications Magazine, July 1996,96–108.

9 All Work Packages. MAGNET System Specifica-tion. IST-MAGNET Beyond, Deliverable 1.1.1,2006.

10 Olsen, R et al. Service, Resource and ContextDiscovery system specification. IST-MAGNET,Deliverable 2.2.3, 2005.

11 Jacobsson, M et al. Refined Architectures andProtocols for PN Ad-hoc Self-configuration,Interworking, Routing and Mobility Management.IST-MAGNET, Deliverable 2.4.3, 2005.

12 Australian Communications-Electronic SecurityInstruction 33 (ASCI 33). Handbook 3 – RiskManagement, version 1.0. 2004.

13 Australian Government, Department of Defence.Defence Signals Directorate.http://www.dsd.gov.au/

Sofoklis A. Kyriazakos studied Electrical Engineering in RWTH Aachen and specialized in Wireless Systems

in the Chair of Communications Networks. He then moved to the Telecommunications Laboratory of the

National Technical University of Athens (NTUA), where he received his PhD in the area of RRM in Wireless

Networks and MBA in Techno-economic Systems. He his currently working as Assistant Professor at

Aalborg University, Denmark and his research areas are RRM in Systems Beyond 3G, Risk Analysis,

Traffic Estimation and Business Modeling. He has performed a large number of publications in journals,

book chapters, conferences and standardization bodies.


Neeli Rashmi Prasad is Associate Professor and Head of Wireless Security and Sensor Networks Lab, part

of Wireless Network including Embedded systems Group (WING), Center for TeleInfrastruktur (CTIF), Aal-

borg University, Denmark. She received her PhD from the University of Rome, Italy, in the field of “adaptive

security for wireless heterogeneous networks” in 2004 and MSc (Ir.) degree in Electrical Engineering from

Delft University of Technology, The Netherlands, in the field of “Indoor Wireless Communications using Slot-

ted ISMA Protocols” in 1997. She joined Libertel (now Vodafone NL), Maastricht, The Netherlands in 1997.

From 1998 to 2001 she worked as Systems Architect for Wireless LANs in Wireless Communications and

Networking Division of Lucent Technologies (now Agere Systems), The Netherlands. From 2001 to 2003

she was with T-Mobile Netherlands as Senior Architect for Core Network Group. From 2003 to 2004 she

was Senior Research Manager at PCOM:I3, Aalborg, Denmark. Neeli Prasad has published widely and has

supervised several Masters and PhD students.



1 Introduction

System scalability deals with the exploitation of themost appropriate access network/technology. Thenecessity for the user terminal to transmit and receivetowards and from different radio access networksrequires a certain level of reconfigurability of theradio interface in order to allow the exploitation ofdifferent access technologies (e.g. MC-SS, FM-UWB,etc.) and different standards (e.g. IEEE 802.11a,IEEE 802.15.3 or IEEE 802.15.4).

Most of the past work on multimode terminalsfocused on the integration of Wireless Local AreaNetwork (WLAN) and Cellular Wide Area Network(WAN) network terminals [1]. In addition to this kindof integration, also the different network access stan-dards defined by the IEEE 802 working groups canbe integrated. IEEE 802 standards define only thephysical layer (PHY) and Medium Access Control(MAC) layer of a given interface. For end-to-endservices and their architectural view of the network,other core network elements have to be defined. Aninteresting research topic concerns the developmentof multimodal WPAN devices capable of exploitingsimultaneously the IEEE 802.15.3-based HDR inter-face and the IEEE 802.15.4-based LDR interface [2].This is a new concept of cooperation between accessnetworks/technologies since there is not only a swapof access networks dictated by coverage issues ortransmission rate shift, but it is envisioned a sort ofcooperation aimed to enhance: QoS provisioning,power and/or bandwidth efficiency, reliability andavailability. We aim to apply this concept to themultimode WPAN devices where the PHY transmis-sion bit rate of the LDR interface ranges from a fewbits per second (b/s) to 100 kb/s and the transmissionbit rate of the HDR interface ranging from 28.87 Mb/sto 130 Mb/s.

While in such multimode terminals there exists a gapin the transmission bit rate, the features of the IEEE802.15.3 based MAC allow to decrease the systemthroughput to cover this gap. The final result is theavailability of an effectively integrated dual-modeWPAN device that provides a very large range oftransmission bit rates seamlessly and transparentlyto the user.

The aim of this paper is to define the guidelines forthe implementation of dual mode LDR/HDR WPANdevices. The definition of guidelines encompass:

• Discussion of the need to develop dual modeLDR/HDR devices by identifying several multi-mode application scenarios;

• Analysis of the interference between the LDR andHDR air interfaces of reference;

• Analysis of coexistence mechanisms for the mitiga-tion of the interference within a device;

• Proposal of a protocol architecture which effec-tively exploits the dual mode capability of thedevice.

In order to introduce and discuss such guidelines, wehave organised the paper as follows. In Section 2several scenarios for multimode LDR/HDR WPANdevices are proposed. Section 3 analyses the inter-ference between the two proposed AIs. In Section 4coexistence mechanisms are discussed, while in Sec-tion 5 the protocol architecture which enables multi-mode LDR/HDR WPAN devices is proposed.Finally, conclusions are drawn in Section 6.

Coexistence Concept for the Implementation of LDR/HDR

WPAN Multimode Devices

M A U R O D E S A N C T I S , J O H N G E R R I T S , J U L I A N P É R E Z V I L A

Mauro De

Sanctis is

Assistant

Professor at

Department of

Electronics

Engineering,

University of

Roma “Tor

Vergata”, Italy

John Gerrits is

with CSEM S.A.

in Neuchatel,

Switzerland

Julián Pérez Vila

is Research and

Development

Engineer with

Telefónica I+D,

Spain

This paper defines the guidelines for the implementation of multimodal devices with Frequency

Modulation Ultra Wide Band (FM-UWB) and Multi Carrier Spread Spectrum (MC-SS) air interfaces

(AIs) for short range Low Data Rate (LDR) and High Data Rate (HDR) connections. Several novel

scenarios have been proposed which require the simultaneous exploitation of LDR and HDR Wireless

Personal Area Network (WPAN) connections. The possibility of performance degradation when one AI

is transmitting and the other one is receiving is discussed. It has been found that MC-SS transmission

can impact the FM-UWB reception. Furthermore, an overview of collaborative and non-collaborative

coexistence mechanisms is provided. Finally, the architecture that allows the efficient exploitation of

the two AIs in one device is proposed.



2 Multimode Application Scenarios

In this section we will define the application scenar-ios where both LDR and HDR interfaces are used.The simultaneous exploitation of the two air inter-faces within one multimode device can be dictatedby user needs, scenario requirements and/or by effi-ciency improvement. Simultaneous exploitation ofair interfaces does not mean that they simultaneouslytransmit and/or receive, but it means that both inter-faces are powered on, have established a connectionand have data to transmit or receive.

User needs can lead to multimode application scenar-ios when both LDR and HDR applications are run-ning on the same multimode device. This is the casewhere the user device is connected in LDR and HDRtransmission with several other wireless devices(mouse, headphone, printer, sensors, mobile gameplayer, file repository, etc.). In this case the data flowof one single connection is independent of each other.

On the other hand, scenario requirements can leadto the need for a multimode device (i.e. translationalbridge) capable of forwarding data received from anLDR (HDR) connection to a HDR (LDR) connection.In this case the data that flow through the HDR(LDR) connection depend on the data that flowthrough the LDR (HDR) connection.

Finally, when one of the two AIs is experiencing badtransmission conditions the user satisfaction can beenhanced by the efficient exploitation of the multi-mode capability of the devices. The mentioned badconditions can be raised by: interference with otherwireless technologies (e.g. WLAN), buffer load,remaining battery energy, coverage.

There are five different scenarios that require thesimultaneous operation of the LDR and HDR airinterfaces. In the following subsections, the fivescenarios are outlined and the main features of themultimode application scenarios are listed in Table 2.

2.1 Multimode Scenario for Multiple Traffic

– Scenario no. 1

Accounting for the application requirements in termsof minimum data rate, we can identify several cate-gories of applications for HDR and LDR AIs respec-tively. Typical applications that require LDR andHDR AIs are listed in Table 1.

The simultaneous exploitation of different applica-tions can be dictated by the user needs. An exampleof this multimode scenario is shown in Figure 1 wherethe central notebook is provided with the multimodeLDR/HDR AIs and is connected with several devicesthrough LDR and HDR connections. Since each con-nection is independent of each other, this type ofscenario does not require the interoperability and/orinterworking between LDR and HDR AIs for datatransmission. However, it could require the co-operation of the two AIs for the management ofthe coexistence in terms of interference avoidance.

2.2 Multimode Scenario for Aggregate

Traffic – Scenario no. 2

This scenario is characterised by the exploitation ofmany LDR connections and one HDR connectionwhere the aggregation of the LDR connections flows.The transmission links of the multimode scenario arebidirectional; however, in the following we are going toidentify the direction of transmission of the useful data.

LDR applications HDR applications

Wireless mouse or keyboard connection Video streaming

Wireless printer connection Tele conferencing

Low quality audio/voice transmission File transfer

Data sensor transmission Web browsing

Notebook or PDA synchronization with

desktop computer Mobile gaming

Table 1 List of applications for LDR and HDR AIs

Scenario Type of scenario Number of Level of Interoperability

no. devices complexity requirements

1 Generated by user needs many low none

2 Generated by efficiency improvement many medium/high yes

and/or scenario requirements

3 Generated by efficiency improvement 2 or more medium/high yes

4 Generated by efficiency improvement 2 medium/high yes

5 Generated by efficiency improvement 2 medium/high yes

Table 2 List of features of multimode application scenarios



In the first example of this scenario (see Figure 2) themultimode terminal DEV1 collects data from severalLDR devices/sensors (S1-SN) by using the FM-UWBinterface and transmits an aggregate HDR data traffic(e.g. data-sensors’ traffic) to a single mode terminalDEV2 (e.g. a data base) by using the MC-SS airinterface.

In the second example of this scenario (see Figure 3)the direction of the transmission is the opposite withrespect to the previous example since the objectiveof the transmission is to deliver data to differentdevices; when the data to be delivered are the samefor all the devices S1-SN, this example can be con-sidered as a broadcasting scenario.

In this second case, the function of the multimodedevice DEV1 is to disaggregate the traffic sent bythe single mode device DEV2.

The aggregate traffic scenario is meaningful whendirect connections between DEV2 and S1-SN cannotbe established for different reasons:

• DEV2 is a HDR single mode device and S1-SN areLDR single mode devices;

• DEV2 and S1-SN are out of coverage;

• Privacy/security motivation does not allow to con-nect DEV2 with S1-SN.

Following these considerations, the multimode sce-nario for aggregate traffic is generated by user needsor efficiency improvement. In this scenario, the multi-mode terminal DEV1 acts as a translational bridgewhich operates above the MAC sublayer and is capa-ble of converting IEEE 802.15.4 MAC frames intoIEEE 802.15.3 MAC frames and vice versa.

2.3 Multimode Scenario for LDR

Applications – Scenario no. 3

In this scenario, depicted in Figure 4, the requirementof the connection between DEV1 and DEV2 in termsof data rate is less than 100 kb/s; since both air inter-faces can satisfy this requirement and both HDR andLDR connections can be established, an interoperableparallel link can be set up where the data can bedelivered by using the momentarily available ormomentarily most suited AI. In this case, the multi-mode terminal (DEV1) acts as a dynamic switch,routing data coming from the application layer tothe most suited interface. The most suited AI can beidentified on the basis of interference level with otherradio technologies (e.g. WLAN), error rate, bufferload, etc. The coverage area of the LDR and HDRradio technologies overlaps and hence their usage

can be combined in order to obtain the best possibleconnection according to a certain criterion [3] inorder to enhance the efficiency of the connectionbetween DEV1 and DEV2.

2.4 Multimode Scenario for HDR

Applications – Scenario no. 4

When the proper MC-SS-based interface for the runningHDR application (see Table 1) cannot be momentarilyused because of interference with other technologies

HDRlink

Mobilegamingplayer

Filerepository/database Wireless

headphones

HDRlink

LDRlink

LDRlink

Wirelessmouse

S1

S2

SN

DEV2(database)

LDRlinks

HDRlink

DEV1

.

.

.

Figure 1 Multimode scenario for multiple traffic

S1

S2

SN

DEV2(database)

LDRlinks

HDRlink

DEV1

.

.

.

Figure 2 Multimode scenario for aggregate traffic(aggregation of traffic)

Figure 3 Multimode scenario for aggregate traffic(disaggregation of traffic)



(with e.g. WLAN), remaining battery energy, bufferload or coverage issues, the possibility of using twotransmission technologies allow to combat the momen-tary issue and to improve some efficiency metrics.

The difference between this scenario, depicted inFigure 5, and the previous one is due to the data ratetransmission requirements of the running application.In scenario no. 3 the data rate transmission require-ments of the running application is lower than100 kb/s, while in scenario no. 4 the data rate trans-mission requirements of the running application ishigher than 100 kb/s. In scenario no. 3 the need toestablish an interoperable parallel link is needed in

order to improve the bandwidth efficiency of thesystem, while scenario no. 4 improves the powerefficiency of the system.

Both scenarios no. 3 and 4 are able to counteract theinterference that can be generated by other radio tech-nologies over one of the two AIs.

2.5 Multimode Scenario for HDR Asymmetric

Transmissions – Scenario no. 5

In this scenario (see Figure 6) a HDR applicationwith asymmetric bandwidth requirement (e.g. a filetransfer) is running on the multimode device (DEV1)which is connected to a multimode device (DEV2)through both LDR and HDR links. In the case of reli-able data transfer by using TCP at the transport layerthe transmission is characterised by a large amountof useful data transfer with high bandwidth require-ments on the forward direction (from DEV2 to DEV1)and a small amount of data transfer (acknowledge-ments) with low delay requirements on the reversedirection (from DEV1 to DEV2).

An efficient exploitation of the dual connection is theuse of the bandwidth efficient HDR interface for thetransmission of data and the use of the energy effi-cient LDR interface for the transmission of the ACKs.This scheme of transmission allows to avoid the in-efficient use of time slot of the HDR connection forthe transmission of small size ACKs and increase theenergy efficiency of the system by using the powerefficient LDR AI for the transmission of ACKs.

Even if the possiblity of simultaneously using LDRand HDR AIs has already been addressed in the past,scenarios no. 3, 4 and 5 propose novel efficient meth-ods of exploiting dual mode devices.

3 Interference Between FM-UWB

and MC-SS

3.1 WPAN Air Interfaces

Although the MAC layer of the IEEE 802.15.3standard for HDR WPAN and the MAC layer of theIEEE 802.15.4 standard for LDR WPAN are wellestablished, the physical layers are still rather unde-fined. Multi Band Orthogonal Frequency DivisionMultiplexing (MB-OFDM) and Direct SequenceUltra Wide Band (DS-UWB) are the main candidatesfor the physical layer of the HDR standard, whileDirect Sequence Spread Spectrum (DS-SS), ChirpSpread Spectrum (CSS) and Impulse Radio UltraWide Band (IR-UWB) are the main candidates forthe physical layer of the LDR standard.

DEV2(database)

LDRlink

HDRlink

DEV1 (LDRapplication)

DEV2(database)

LDRlink

HDRlink

DEV1 (HDRapplication)

DEV2(database)

LDR link(ACK)

HDR link(data)

DEV1 (HDRapplication)

Figure 4 Multimode scenario for LDR applications

Figure 5 Multimode scenario for HDR applications

Figure 6 Multimode scenario for HDR asymmetrictransmission



The act of defining the two air interfaces for HDRand LDR WPAN has been fairly uncoordinated,resulting in coexistence issues of the proposed airinterfaces when operating in the same environment.

We have selected two air interfaces for LDR andHDR WPAN with the aim of minimizing systemcomplexity, energy consumption and coexistenceissues. In the following we describe the selected airinterfaces [2].

LDR Air Interface

The LDR air interface uses FM-UWB techniques and isa scalable air interface technology aimed at short-range(< 10 m) LDR (< 100 kb/s) applications, which is char-acterized by a low power consumption and ease-of-implementation on an integrated circuit [4]. FM-UWBuses double FM; a low modulation index digital Fre-quency Shift Keying (FSK) is followed by high modu-lation index analogue FM to create a constant envelopeUWB signal. The estimated power consumption is upto 3.5 mW for the transmitter and 7.5 mW for the low-complexity receiver. Though Frequency Division Mul-tiple Access (FDMA) techniques at sub-carrier levelcan be exploited to accommodate multiple users inextremely simple devices, the PHY is mainly targetedto operate with the IEEE 802.15.4 MAC, in whichTime Division Multiple Access (TDMA) is applied.

HDR Air Interface

For the HDR system, an Orthogonal Frequency Divi-sion Multiplexing (OFDM) transmission based PHYwith spreading in frequency domain is developed,which operates in the 5.2 GHz bands allocated toWireless Access Systems (WASs). The maximumdata rate without MIMO is approximately 130 Mb/sand the system bandwidth is 40 MHz; however analternative solution with 20 MHz is also specified toensure compliance with regulatory bodies. Thoughspreading in frequency domain is applied, it is notused as multiple access scheme, thus resulting in aMulti Carrier Spread Spectrum (MC-SS) transmissionscheme [5]. The chosen TDMA being compliant withthe foreseen IEEE 802.15.3 MAC scheme avoidsmultiple access interference. The objective of thefrequency domain spreading is to exploit diversityand to average out interference from other systems.Further, the number of spreading codes can be variedresulting in increased flexibility and scalability.

Interference Between LDR and HDR AIs

Table 3 shows the main parameters of the selectedLDR and HDR air interfaces. The interferencebetween FM-UWB and MC-SS is an important issue.The major differences are the transmission power(34 dB of difference) and signal bandwidth. Signalattenuation as a function of distance is almost equal,

since the two operating frequencies are relativelyclose to each other. Figure 7 shows the receivedpower at a distance d for free space propagationconditions for both air interfaces.

The situation becomes rather clear after inspection ofthis Figure. The maximum received signal from theFM-UWB transmitter at 10 cm distance is the sameas the level received 5 m away from the MC-SStransmitter. It is worth noting that the interferencebetween the proposed FM-UWB and MC-SS systemscan be classified as out-of-band interference. Inter-ference from MC-SS to the FM-UWB system is verylikely to happen and therefore it is the major issue ofour interference investigations.

In the next subsection we will investigate which Sig-nal-to-Interference Ratio (SIR) can be tolerated andwhat would be required in terms of interference miti-gation for various scenarios like collocated (0.1 –10 m distance) and remote (> 10 m) systems.

Parameter FM-UWB MC-SS

Transmit power -14 dBm +20 dBm

RF centre frequency 4.5 GHz 5.25 GHz

RF signal bandwidth 500 MHz 36 MHz

RF signal envelope Constant Strongly varying

Predominant Modulation FM AM

Path loss @ 1 m 45 dB 46.5 dB

Table 3 Characteristics of FM-UWB and MC-SS air interface

0

-20

-40

-60

-80

100 101

distance [m]

PRX[dBm)

MC-SS

FM-UWB

10-1

Figure 7 Received power as a function of distance for FM-UWB andMC-SS



3.2 Influence of MC-SS on FM-UWB

MC-SS signals like all multi-carrier signals show rel-atively strong envelope variations. MC-SS signals aredemodulated by the wideband FM demodulator. Boththe AM and FM components yield a signal at thedemodulator output. The part of the demodulatedMC-SS signal falling within the sub-carrier band-width of the demodulated FM-UWB signal deter-mines the sub-carrier SIR degradation.

It was found that a SIR up to -2 dB (interferer 2 dBstronger than wanted signal) can be dealt with with-out significant BER degradation (BER < 1x10-3).Figures 8, 9 and 10 show three situations:

• FM-UWB with AWGN SNR = 10 dB• FM-UWB plus MC-SS, SIR = 0 dB• FM-UWB plus MC-SS, SIR = -6 dB

b)

5.5

frequency [GHz]a)

0 40 80 120 160 200

-0,5

-1.5

-2.5Delayed TXD

Demodulated sub-carrier

RXD

time[µs]c)

0

-20

-40

-60

-800 0.5 1,0 1.5 2.0 2.5 3,0 3.5 4.0

-60

-80

-100

-120

-1403.5 4.0 4.5 5.0

spectrum of received signal [dBm/MHz]

frequency [MHz]

spectrum of wideband demodulator output

Figure 8 AWGN case with SNR = 10 dB, a) RF inputsignal, b) after wideband FM demodulator, c) com-parison of transmitted and received data

-60

-80

-100

-120

-140

0

-20

-40

-60

-800 0.5 1,0 1.5 2.0 2.5 3,0 3.5 4.0

0 40 80 120 160 200

-0,5

-1.5

-2.5 Delayed TXD


RXD

time[µs]c)

b)

3.5 4.0 4.5 5.5

frequency [GHz]


4.5 5.0

a)

frequency [MHz]


Figure 9 MC-SS interference with SIR = 0 dB, a) RFinput signal, b) after wideband FM demodulator, c) comparison of transmitted and received data



For each case the following three sub-Figures areshown:

a) the spectrum of the received RF signal showing theFM-UWB signal plus interference;

b)part of the spectrum after the wideband FM demod-ulator containing the FSK sub-carrier and noise;

c) time domain view of the delayed transmitted dataTXD, the lowpass filtered FM demodulated sub-carrier signal, and its hard-limited version: thereceived data RXD.

These figures illustrate how the received data isaffected by the interference. For the sake of clarity,only a period of 200 µs (20 bits) is shown.

At SIR = 0 dB, the FM-UWB 100 kbit/s system eas-ily survives with low BER. However, the lowpassfiltered sub-carrier demodulator signal (dotted black)starts to show visible distortion.

At SIR = -6 dB (interferer 6 dB stronger than wantedsignal) the sub-carrier SNR has become too low fordemodulation.

3.2.1 Physical Layer Interference Mitigation

Techniques

The wideband FM demodulator is sensitive to bothfrequency and envelope variations of the signal [4].The red line in Figure 11 shows its transfer functionfor a delay time t equal to N = 15 times one quarterperiod of the centre frequency, i.e. 833 ps. Since thedelay line demodulator has no bandwidth limitation,MC-SS signals at 5.25 GHz will also be demodu-lated. By limiting the demodulator input signal band-width, ideally to 4.2 – 4.8 GHz, signals outside thatbandwidth will be attenuated and the interference willbe lowered. The black line in Figure 11 shows anexample of a simple bandpass filter implemented inthe LNA. The following three paragraphs addressfiltering that can be achieved inside the FM-UWB

Figure 10 MC-SS interference with SIR = -6 dB, a) RF input signal, b) after wideband FM demodula-tor, c) comparison of transmitted and received data

-60

-80

-100

-120

-140

0

-20

-40

-60

-800 0.5 1,0 1.5 2.0 2.5 3,0 3.5 4.0


frequency [MHz]

0 40 80 120 160 200

-0,5

-1.5

-2.5 Delayed TXD


RXD

time[µs]

3.5 4.0 4.5 5.5

frequency [GHz]


5.0

c)

b)

a)

1.5

0.5

-0.5

-1.5

FM-UWB MCSS

N=15

Without BPF

With BPF Q=8

fc [GHz]

4.0 4.4 4.8 5.2 5.6 6.0

Figure 11 Transfer function of the wideband FM demodulator, withand without LNA BPF



receiver as well as by using external filters and alsoby lowering the antenna efficiency at 5.25 GHz.

LNA Filtering

The Low Noise Amplifier (LNA) can be exploited toimplement additional bandpass filtering around theFM-UWB centre frequency of 4.5 GHz, e.g. by usinga parallel resonant load in the LNA as shown inFigure 12.

With p = jω, its transfer function in the frequencydomain is given by:

By introducing a variable named detuning ν definedas:

The resonator transfer can now be rewritten as:

This 2nd order resonator is fully characterised by itsresonant frequency ω0, quality factor Q and the maxi-mum value of its transfer function H0 which has thedimension of impedance for the circuit shown in Fig-ure 13. The impedance of the parallel resonant circuitis given by:

Its magnitude and phase are given by:

The resonant frequency and quality factor are givenby:

Figure 13 shows the magnitude of the normalisedtransfer function for various Q values. Clearly, withincreasing Q, out-of-band signals are more attenu-ated. A practical limit for the quality factor is thebandwidth of the FM-UWB signal which needs tofit within the filter’s bandwidth. Therefore, Q valueshigher than 8 are not recommended. This yields 7 dBof attenuation at 5.25 GHz, not enormous, however,it comes for free.

External Filtering

The attenuation of the MC-SS signal can be increasedby using an external filter. Figure 14 shows the mea-sured transfer function of a commercial filter avail-able from Taiyo Yuden. This filter as an insertionloss of 1 dB at 4.5 GHz and an attenuation of 23 dBat 5.25 GHz. The filter size is small and it can be eas-ily placed between antenna and FM-UWB receiver.

Filtering by the Antenna

Antennas can be designed to have notches in theirfrequency characteristic. In [6] a CPW fed planarUWB antenna of small size and having a frequencyband notch feature at the MC-SS frequency band(5.25 GHz) is presented. Figure 15 shows theantenna. The slot (the smile) in the upper part ofthe metallisation lowers radiation efficiency around5.25 GHz. The authors claim attenuation valuesbetween 5 and 10 dB at 5.25 GHz.

3.2.2 Interference in Practical Situations

With the combination of internal and external filter-ing with an antenna that has notches in its frequency

H = H0

ω0

Qp

p2 +ω0

Qp + ω2

0

=H0

1 + jQ ωω0

− jQω0

ω

ν =ω

ω0

−

ω0

ω

≈ 2ω − ω0

ω0

= 2∆ω

ω0

H =H0

1 + jQν

Z =V

I=

Rp

1 + jωRpC − jωRp

ωL

=Rp

1 + jQ ω

ω0

− jQω0

ω

=Rp

1 + jQν

|Z| =Rp√

1 + Q2ν2

ω0 =1

√LC

Q =

ω0L

Rs

=

Rp

ω0L= Rp

√C

L

l

+V-

C

jQlL

Rs

lin

Figure 12 Parallel resonant circuit

f [GHz]

5

0

-5

-10

-15

-203 3.5 4 4.5 5 5.5 6

Vout [dB]

Q = 2

Q = 4

Q = 8

Figure 13 Normalised resonator transfer function with Q as parameter



characteristic, an attenuation between 30 and 40 dBof the MC-SS signal at 5.25 GHz appears feasible.What does this imply for the practical situationbetween LDR and HDR far apart and close to eachother? The following two paragraphs provide theanswer.

Interference Between Remote AIs

With 35 dB of additional attenuation of the MC-SSsignal, the received signal at distance d from an FM-UWB and MC-SS transmitter become equal, seeFigure 7. Remembering that an SIR down to -2 dB isacceptable, this roughly means that the FM-UWBreception is not disturbed by MC-SS signals thatorigin from a transmitter further away than the FM-UWB transmitter. Interference from remote HDRdevices can thus be tolerated.

Interference Between Co-located AIs

Interference from co-located HDR devices, that maybe closer than the LDR devices, will be a problem.It is necessary to design higher layer coexistencemechanisms when the physical layer has no moremargin.

4 Coexistence Mechanisms

There are two categories of coexistence mechanisms:collaborative coexistence mechanisms where the twointerfering AIs/networks exchange information andnon-collaborative coexistence mechanisms where theexchange of information is not allowed [7]. The pos-sibility of exchanging information is quite easy whenthe two AIs are co-located in the same dual-mode ter-minal. Collaborative coexistence mechanisms enablethe use of multiple AIs at limited levels of interfer-

ence by a smart scheduling of the channel access,thereby suffering from some performance degrada-tion. These mechanisms are employed on the MACand higher layers and require control informationexchange among active peers.

Non-collaborative coexistence mechanisms involveeither only the PHY, or the PHY and MAC layers,and they do not need an exchange of control informa-tion between active peers. These methods basicallyexploit fluctuations of the channel quality in time,frequency, or space, to optimize the form of the trans-mitted signals with the aim of minimizing distortionby cochannel interference at the receiver. Their per-formance depends on the availability of informationand the capability to predict the encountered cochan-nel interference.

Various proprietary collaborative coexistence mecha-nisms exist to coordinate the radio activity in order toprevent simultaneous operation of co-located AIs.

Figure 14 Measured transfer function of Taiyo Yuden bandpass filter

Figure 15 Planar UWB antenna with notch at 5.25 GHz and its gain

3 4 521

0

-10

-20

-30

-40

-50

frequency [GHz]6

filter transfer [dB]

10

5

0

-5

-10

-15

-20

31 5 7 9 11

Without V-slot, boresight gain

With V-slot, boresight gain

Without V-slot, sidesight gain

With V-slot, sidesight gain

frequency [GHz]

gain [dB]



Most of them are designed to manage the coexistenceof Bluetooth and WLAN systems [8]. The approachesvary in detail but essentially act to interleave opera-tion in order to make the operations appear simulta-neous. The techniques address the scheduling andpriority setting of the two systems, making trade-offson transmission duty cycle, idle times, and packettype (data, beacon). The packets of one system canbe sent while the other is idle and vice versa; the endeffect is to deliver reliable communication on bothsystems with a negligible loss of throughput. This is atime division approach which depends on the duplexmethod and the multiple access scheme of the two AIs.

There are three approaches that belong to the cate-gory of collaborative mechanisms:

1)Host software approaches (driver-level and dualmode switching);

2)MAC-level approaches;

3)System level approaches covering the entire wire-less sub-system.

Host Software Approaches

The host software approach is a time-divisionapproach, essentially based on the separation of theoperational periods for each AI, and it has two possi-ble implementations:

1)Dual-mode radio switching. This approach worksby completely suspending the operation of one AIwhile the other is operational. There are two meth-ods to implement it. The first method requires turn-ing off the non-operating AI without signalling toother nodes in the network. The drawback of thismethod is that it can reduce performance. The sec-

ond method acts as a signal to other network nodesthat the operation of the AI is suspended.

2)Drivel-level switching. This approach is similar tothe previous one but the functionality of control ismanaged at the driver level and it includes user-dependent switching, discriminatory switching,successful-transmission switching, statisticalswitching and time delay switching. In thisapproach, application transmit requests deliveredto the operating system, are mediated at the driverlevel, thereby avoiding simultaneous transmissionand collisions. This approach degrades the through-put because only one AI is active at a time; as aresult, systems using driver-level transmit switch-ing can suffer from dropped packets. As with dual-mode radio switching, this approach does notswitch quickly.

MAC Level Approaches

In the MAC-level switching we can apply one of thefollowing solutions: to modify either MAC layers(802.15.4-based LDR MAC and 802.15.3-basedHDR MAC) or to develop a new and self-containedmodule that communicates with both MACs.

In both cases this approach performs approximatelythe same functionality of driver level switching, butadds a predictable latency, and it would be the suit-able solution to establish high performance coexis-tence mechanism in case of real time transmissions.This technique does not suffer from transmittingsignals into incoming receptions.

Furthermore, since MAC-level switching is per-formed in the baseband, it is able to change theoperational interface at a much faster rate than hostsoftware approaches.

System Level Approaches

This approach encompass the entire wireless sub-sys-tem: for example, a driver-level switching techniquemay generate the best user experience in a low band-width synchronisation scenario, while MAC-levelswitching will manage interference much more effec-tively for real time or voice traffic, or when a user haswireless peripherals such as speakers or a keyboard.

5 Protocol Architecture

The protocol architecture (see Figure 16) includes theUniversal Convergence Layer (UCL) and two differ-ent radio air interface stacks, one for LDR based on aIEEE 802.15.4 medium access control layer and FM-UWB physical technology, and a second stack forHDR based on IEEE 802.15.3 medium access controllayer and MC-SS physical technology.

Logical link control

Specific convergence sub-layer

MC-SS physical

802.15.3 medium access

Universal convergence sub-layer

FM-UWB physical

802.15.4 medium access

Cross Layer Functionality

Figure 16 Protocol architecture of the multimodeterminal



The UCL is the component which is going to providea common abstraction of every air interface; in thissense, the UCL is able to provide convergence ofmanagement and data operations over both air inter-faces.

The protocol architecture of the UCL is composed bytwo sublayers, the Logical Link Control and the Spe-cific Convergence Sub-Layer:

• The Logical Link Control (LLC) provides the com-mon abstraction to control and transmit data overseveral radio air interface technologies. Inside theLLC there is an internal database which containsinformation associated to every radio air interface.According to this information the LLC is able toensure the coexistence of several radio air interfacetechnologies in the PAN through multimode opti-misation mechanisms.

• The Specific Convergence Sub-Layer (SCSL) is incharge of the convergence of frames and serviceoperations between the LLC and specific radio airinterface stacks.

Through these entities and functionalities, the UCLwill

• Provide both management and data operation overseveral radio air interface technologies;

• Support upper layers with information about cur-rent configuration of MAC and PHY functionali-tites in a multimode PAN;

• Coordinate coexistence optimisation and activationof specific optimisation mechanisms for specific airinterface layers.

6 Conclusion

This paper has presented the issues arising from thesimultaneous exploitation of LDR and HDR WPANair interfaces within the same environment. Severalscenarios have been proposed which clarify the needfor multimode LDR/HDR WPAN air interfaces. Thetwo air interfaces for the LDR and HDR WPAN havebeen outlined focusing on the need for high coexis-tence levels. It has been found that, even if LNAfiltering, external filtering and antenna filtering aregood solutions for the management of interferencebetween FM-UWB and MC-SS AIs, when the two

AIs are located close to each other, MAC and higherlayers mechanisms are needed to allow coexistence.

Acknowledgement

This work is supported by the European Commissionunder the FP6 MAGNET Beyond Integrated Project.The authors wish to acknowledge the collaboration ofthe partners involved in the work package 3 of thisproject.

References

1 Bantas, S et al. Architecture considerations andintegrated-passives-based design for a dual-modeGPRS-WLAN SiGe RF transceiver. IEEE 58thVehicular Technology Conference, 4, 2237–2241,October 2003.

2 De Sanctis, M et al. Coexistence Concept for theImplementation of the FM-UWB and MC-SS RASolutions. IST-027396 MAGNET Beyond projectdeliverable D3.3.1, June 2006.

3 Ferreira, L, Serrador, A, M. Correia, L M, Svaet,S. Concepts of Simultaneous Use in the Conver-gence of Wireless Systems. COST 273,TD(04)102, June 2004.

4 Gerrits, J F M et al. Principles and Limitations ofUltra Wideband FM Communications Systems.EURASIP Journal on Applied Signal Processing,Special Issue on UWB-State of the Art, 2005 (3),382–396, March 2005.

5 Schoo, K et al. MC-SS for Personal Area Net-works – A Combined PHY and MAC Approach.14th IST Mobile & Wireless CommunicationsSummit 2005, Dresden (GE), 19–23 June 2005.

6 Kim, Y, Kwon, D H. CPW-fed planar ultra wide-band antenna having a frequency band notchfunction. Electronics Letters, 40 (7), April 2004.

7 IEEE. Coexistence of Wireless Personal AreaNetworks with Other Wireless Devices Operatingin Unlicensed Frequency Bands. IEEE, August2003. (IEEE Standard 802.15.2)

8 Wi-Fi™ (802.11b) and Bluetooth™: An Exami-nation of Coexistence Approaches. MobilianCorporation, 2001. (White paper)



Mauro De Sanctis is Assistant Professor at the Department of Electronics Engineering, University of Roma

“Tor Vergata”, Italy. He received the “Laurea” degree in Telecommunications Engineering in 2002 and the

PhD degree in Telecommunications and Microelectronics Engineering in 2006 from the same university.

He was involved in the MAGNET (My personal Adaptive Global NET) European FP6 integrated project and

in the SatNEx European network of excellence; he is currently involved in the MAGNET Beyond European

FP6 integrated project as WP3/Task3 leader. In 2006 he worked as post-doctoral research fellow for the

ESA/ARIADNA extended study named “The Flower Constellation Set and its Possible Applications”. He

is/was involved in several Italian national research projects: ICONA (Integration of COmmunication and

NAvigation services) from January 2006 to December 2007, SHINES (Satellite and HAP Integrated

NEtworks and Services) from January 2003 to December 2004, CABIS (CDMA for Broadband mobile

terrestrial-satellite Integrated Systems) from January 2001 to December 2002. He was involved in the

DAVID satellite mission (DAta and Video Interactive Distribution) and is currently involved in the WAVE

satellite mission (W-band Analysis and VErification) of the ASI (Italian Space Agency). In the autumn of

2004, he joined the CTIF (Center for TeleInFrastructure), a research centre focusing on modern tele-

communcations technologies located at the University of Aalborg, Denmark. He is currently serving as

Associate Editor of the IEEEAerospace and Electronic Systems Magaqzine. His main areas of interest are:

satellite networks and constellations (in particular Flower Constellations), stratospheric platforms, resource

managment of short range wireless systems.


John F.M. Gerrits received the MScEE degree from Delft University of Technology, the Netherlands, in 1987

with final thesis on the design of integrated high-performance harmonic oscillator circuits. In 1988 he joined

the Philips T&M division in Enschede, the Netherlands, where he designed integrated oscillator and data-

acquisition systems for oscilloscope applications. In 1991 he joined CSEM in Neuchâtel, Switzerland, where

he has been involved in both system and circuit design of a single-chip low-power VHF radio receiver for

hearing aid applications and of a single-chip UHF transceiver for ISM applications. His current work involves

system and circuit design of UWB radio systems, RF and EM simulation techniques and measurement

methodology. He is currently working towards his PhD on the fundamental aspects and practical realiza-

tions of FM UWB communication systems at Delft University of Technology. He is editor and co-author of

the book Low-power design techniques and CAD tools for analog and RF integrated circuits published by

Kluwer in 2001. He is the winner of the 2006 European Conference on Wireless Technologies Prize and

holds three European and one US patent.


Julián Pérez Vila received his high engineering degree in Computer Science, Software Engineering special-

ity, from the “Universidad Pontificia de Salamanca” in Madrid, 1997-1998. Following his engineering studies

he worked as an analyst-programmer of Meta4 tools for internal systems of the company El Corte Ingles,

adapting payroll, human resource and accounting modules. In 2000 he began work in Telefonica I+D in the

“3g consultancy group”, editing and contributing several 3g reports on the demand of Telefónica Móviles,

increasing knowledge regarding multiple technologies:

• UMTS-IMS, GPRS and GSM architectures

• Ipv6, OSA and Quality of Service architectures

• Multimode terminals and data-roaming aspects

• Interoperability between Mobile Networks and Wireless LAN

• Universal user profiles, presence servers and SIP technology

After this consultancy period he contributed research and development activities related to knowledge

management and began his participation in the IST European MAGNET project, where he is coordinating

the research and development contributions to the areas of context management in personal networks

and multimode radio air interface technologies.



Introduction

What is excellent service? Delivery of exactly whatyou want and need at the right time, in a mannerappropriate to your immediate situation. In this paper,we consider some challenges and issues associatedwith realizing this with a conceptual system we call4G-Smart Wireless Living (4G-SMILING or 4GS),which aims to provide the highest level of automatedservice to mobile users. 4GS would pro-actively pro-vide a “service cocktail,” intended to optimize a per-son’s daily activities, eliminate mundane, stressing,and time-consuming work, and proactively provideservices and products to improve convenience andquality of life, 24/7. We think that the most success-ful 4G system design will be service-driven, and thatthe user’s internet-based agent will constantly opti-mize the mode of delivery of these services.

Researchers seeking to provide the highest levels ofservice and convenience to mobile users are found intwo fields: wireless communications and pervasive orubiquitous computing. However, they take differentapproaches, have different business models, and theymake very different assumptions about requirementsin wireless access technology. We claim that 4GSresides in the intersection of these two fields, and wesuggest a strategy for achieving this intersection thatfocuses on access optimization. We do not providedetailed solutions to the problem in this paper.Rather, we identify likely elements of 4GS, proposea high-level architecture, review some of the currentand recent related projects, and identify importantchallenges to realizing 4GS.

Personalization and Context

Awareness

Part of knowing wants and needs implies familiarityor personalization. This requires an integration andorganization of a user’s preferences, perhaps initially

determined through an interview, but later automati-cally adapted over time, also known as the userprofile. The other part of knowing wants and needsis sensing, which includes traditional forms of userinput, as well as non-traditional forms, such as speechrecognition and on-body and off-body sensor net-works. Proper exploitation of the user profile allowsthe user to simplify his requests. As a very simpleexample, a frequent patron of a particular restaurantcan just ask for “the usual.” An even higher level ofservice and convenience is when the user doesn’teven have to ask. Continuing with the restaurant anal-ogy, the wait staff recognizes the patron and immedi-ately presents the desired meal without requiring asingle word from the user.

The second restaurant example shows that sensingis important to a high level of service. Indeed, thesensed location, user activity and mood can indicatethe relevant parts of the user profiles [10], as shownin Figure 1. For example, a user’s profile for “home”and “eating” may invoke a certain type of music play-ing and a certain filter on which calls the user is will-ing to accept. Applications that exploit sensed infor-

The Unpredictable Future

– Personalized Services and Applications Architecture

M A R Y A N N I N G R A M , R A M J E E P R A S A D , K I M S K A U E

Mary Ann

Ingram is a

Professor at

Georgia Institute

of Technology,

Atlanta, USA

Kim Skaue is

CEO at C3

faculty at

Aalborg Univer-

sity, Denmark

Ramjee Prasad

is Director of

Center for Tele-

infrastruktur

(CTIF) at

Aalborg Univer-

sity, Denmark

In this paper, we discuss our ideas for a concept we call 4G-Smart Wireless Living (4G-SMILING or

4GS). These ideas are based on the view that a focus on service is essential for the development

of 4th generation wireless services. “Essential,” that is, for a lucrative and efficient utilization of the

communications infrastructure. In this concept paper, we imagine the highest levels of service and

convenience, and what technologies, especially wireless access technologies, will be needed to

realize this vision. We observe that the requirements assumed for wireless access technology in the

two fields most related to 4GS, wireless communications and ubiquitous or pervasive computing, are

very different, and we propose a strategy to harmonize these. We also discuss current related projects

and identify some challenges to achieving the 4GS vision

home

public

work

location

activity psych/physical state

eating sleeping happy tired

Figure 1 Example subsets of the user profile



mation are context aware. Proper utilization of theuser profile and sensing technology should minimizeclumsy user data input, as shown in Figure 2.

4G-SMILING (4GS) would offer a “service cock-tail”, which is an extremely personalized and context-aware mix of business and pleasure applications,which automatically adapts as the user moves andacts, 24/7. As the user moves through differentlyequipped or enabled environments, e.g. home, train,automobile, public street, office, stores, or restau-rants, through different activities, e.g. relaxing, work-ing, exercising, cooking, shopping, or visiting, andthrough different moods and emotions, 4GS con-stantly senses the changes, and presents the desiredservices to the user in the most efficient way.

Push, Pull, and Modalities

Computer applications are classified as push, pull ora combination of both [1]. Pull applications are con-ventional and are initiated by the end user. The usermust make choices at every step and must pursue theinformation that he or she needs. The application hasno or very little memory of past interactions with aparticular user. Constant user presence is necessary,and pull applications tend to be short-running (on theorder of minutes). There is no personalization. Anexample pull application is finding the schedule fora train and making reservations online.

A push application is the opposite of a pull applica-tion. A push application is initiated by the applica-tion. It “engages a user at the right time by proac-tively pushing an interactive session to the user” [1].Constant user presence is not necessary, and pushapplications can run for days or more. “Right time”engagement requires context awareness and personal-ization or exploitation of the user profile. Althoughthere will always be a need to retain some pull appli-cations, the highest levels of service will come frompush applications.

Push applications have a variety of ways to presentinformation to the end user; these are called presenta-tion modalities [1]. Conventional examples include

websites, cell phones, email, short message service.Push applications of the future will have many moreavenues to reach the user. Non-conventional exam-ples include on-body actuators that stimulate thesenses directly, creating sounds, tactile sensations,aromas, and visual displays [11]. Other non-conven-tional examples are off-body displays and actuators.The distinction between on- and off-body presenta-tion modalities is particularly important with regardto wireless communications requirements, as off-body modalities may be directly connected to thewired infrastructure. This subject will be discussedin more detail in a later section.

One challenge that follows from a plurality of presen-tation modalities is how to map the content to eachmodality within the set. Currently, content producersgenerally provide the content in several differentforms, which is unscalable and expensive [1]. Ideally,a producer need to provide only one version of thecontent, and then an automatic process maps thecontent onto the modalities [1]. The mapping shouldoptimize a quality function, which could depend onmany parameters, for example, reliability of the con-nection, presentation quality, power consumption onbattery-powered devices and cost.

We may also consider extraction modalities, or waysto collect information from the end user. While theconventional ones are normally associated with pullapplications (again, websites, cell phones, email,short message service), non-conventional ones, suchas sensor networks, can provide the sensing technol-ogy needed for context awareness. A sensor networkcan include on- and off-body sensors. On-body exam-ples are biometric sensors (e.g. blood pressure), whileoff-body examples include cameras and microphoneswith voice recognition. As with presentation modali-ties, the on-body modalities will impact the wirelesscommunications requirements.

In the long term (5-10 years), deployed off-bodymodality packages may evolve into several standardclasses, generic to many applications. However, inthe nearer term (2-3 years), we do not expect them tobe very standardized, and they will be sold in formsthat are customized to different applications. Forexample, we can identify and imagine a number ofoff-body modality packages that could be madeavailable to consumers and enterprises in Table 1. Forexample, in “Fail-Safe Cooking,” the user can selectone or more recipes. Flashing lights show and audiotells the user where to find the necessary items. Videodemonstrates a procedure in a step-by-step way.Embedded processors and sensors in the stove oroven alert the user exactly when a cooking step iscomplete and turns off the heat. The challenge for

Figure 2 Proper sensing and exploitation of the userprofile should minimize clumsy user data entry



4GS is to detect what off-body modality packagesare available to the user in whatever is his currentenvironment, and to optimally adapt the contentmapping process.

Databases, Security and Privacy

A key technology associated with the user profile andits use in providing context-aware information to theuser is database management. The user profile is adatabase and the various types of content that may beprovided to the user are in other databases. Databasemanagement is concerned with the design of efficientprotocols that query or modify databases. The usualconcerns are how to efficiently tag information sothat it can be located by machines with minimumdelay. Network “bottlenecks” and disc access timeare discussed. Communication costs, in terms ofpacket loss and delay, are rarely taken into account,even though the costs of wireless communication canbe significant. Consideration of communication costswould be a necessary part of 4GS design. Determin-ing control, access, and security of the user profileare other major challenges for 4GS, but the rewardis a highly personalized and convenient service.

A fully developed user profile and highly perceptivesensing mechanisms will be very attractive to thosewho seek to harm or exploit the user. The conse-quence of achieving the high level of service andconvenience is that the service will have exceptionalaccess to the user. Adam Greenfield wrote, “By com-

parison with the World Wide Web, ubiquitous com-puting is vastly more insinuative. By intention anddesign, it asserts itself in every moment and throughevery aperture contemporary life affords it. It iseveryware” [19]. Therefore, protection of the profileand sensor data is paramount. This level of access issimilar to a surgeon’s access to a patient. The patienttrusts the surgeon, and by relation, all the surgeon’ssupport staff and systems. The same level of confi-dentiality that is accorded to patient data must beaccorded to the user profile and sensor data.

The surgeon analogy may also be considered fromthe point of view of advertising. The surgeon and hissupport staff are certainly the targets of much adver-tising, about drugs, medical devices and services –there is a lot of money involved. However, the patientis not exposed to all that advertising. Rather, thesurgeon and staff make nearly all the decisions aboutwhat to use and how to use it. Typically, the patienttrusts that the medical team is well-trained, under-stands his needs clearly, and is acting in the bestinterests of the patient. Therefore, the medical staffwill expose the patient to only a limited range oftreatment options, often only if the patient asks. It isanticipated that the highest level of user service willhave similar characteristics. The service will employa strong and narrow filter on advertising, based on theuser profile and context information. Lower-cost ver-sions of 4GS will admit more advertising.

Title Purpose Example Components

“Basic Home Information System” Entry level web-based personal Moderately sized audio/video panels

information system (added to an for each room, connected by WLAN,

existing home) controlled by a server in the home,

connected to the internet, with speech

rec/synth

“Aging in Place” [2] Enable elderly people to stay Floor sensors (fall detector); distributed

in their homes video camera network (detects habit

change)

“Fail-Safe Cooking” Simplify cooking Thermal sensors, video displays, indicator

lights, interface to stove and oven

“Game Place” Video game entertainment Video display, controllers

“Battery Saver Kiosk” Offload computation and long-range A user device clone that plugs into the

communication from on-body to wall power socket

off-body, when in close proximity

“CitiDisplay” Provides information and entertainment Large video display; RFID detector;

to users in a public environment voice recognition

“Home Moods” Creates personalized, dynamic Distributed audio/video, RFID detector,

environment in the home biometric sensors, aroma generator

Table 1 Off-body modality packages that may be offered in the near term (2-3 years)



Mobility and Power Management

Database management, personalization, contextawareness, and confidentiality, are all topics thatcould be considered for exclusively wired networks.However, the convenience aspect of the best informa-tion service demands wireless communication toenable mobility.

Mobility generally implies battery power, and conve-nience means reasonably long battery lifetimes. Ahigh level information service adds more functionsto the current voice, video, email, and messaging thatexisting 3G phones have. The additional functionswill require additional energy. An example of this isGPS, which is available on certain phones today andis very useful for context awareness. However, GPSis known to consume a high amount of energy, andtherefore it is usually disabled on the phones. Unfor-tunately, battery technology moves more slowly thanthe technologies that depend on batteries. Thus forthe services we envision, battery life will shortenunless special measures are taken.

One approach is to provide the user with frequentand convenient opportunities to recharge his or herpersonal devices. For example, chairs could havechargers built into the armrests.

However, another consideration is wireless link avail-ability. Single wireless connections are inherentlyunreliable, and particular air interfaces, such as 3G,WLAN, or Bluetooth, may not be available in all theplaces that a user goes. Multi-band software definedradio (MB-SDR) is an attractive solution to this prob-lem. An SDR can be programmed to work with anyair interface designed for the band of the SDR [3].Adding multi-band capability makes the radioextremely flexible. This adds reliability because theradio can use the best quality air interface at anyinstant. This “intelligence” for finding the best airinterface is called cognitive radio [3]. However, it isgenerally agreed that a barrier to deploying SDR isits projected short battery life.

One challenge is how to determine the power andbandwidth requirements for personal devices aswe move toward a world with ambient intelligence(AmI) and pervasive computing. “Ambient Intelli-gence represents a vision of the future where we shallbe surrounded by electronic environments, sensitiveand responsive to people” [4]. Pervasive or ubiqui-tous computing is the idea that all sorts of common-place items in our surroundings will be able to com-pute.

The current trend is to pack more and more onto thehandheld personal device – more computing capabil-

ity, more memory, more radios (e.g. 3G, WLAN,Bluetooth, GPS), and more application software resi-dent on the device. At the same time, the wirelessbandwidth requirements are going up.

However, if we believe the AmI folks, then most ofthis will not be needed most of the time. Imaginewalking into a room that is equipped with networkedspeakers, microphone, camera, other sensing devicesand a visual display. This network detects your pres-ence and authenticates you – perhaps by the RFIDchip you wear or have imbedded under your skin, orby biometric sensors. You don’t even have to “log-in”. Conceivably, all your applications and servicesrun on servers that are located somewhere on theinternet, not necessarily near your immediate loca-tion. You may have a seat, use the wireless mousethat is conveniently on the chair arm, and proceed toscroll, point and click (if you prefer interaction theold fashioned way), or you can have a verbal dialogwith the system. What are your battery requirementsin this scenario? RFID batteries can last eight yearsor more. The bandwidth requirement between youron-body wireless device(s) could be extremely lowbecause all they have to do is identify you and possi-bly send some biometric information. The off-bodynetwork, on the other hand, all has access to wiredpower, has a much larger display, and may have amuch more reliable connection than the wirelessconnection to the personal device. The idea of havingnearly all the presentation and extraction modalitiesoff-body, and reliably present in our environments,seems the ultimate in convenience.

There is clearly a huge gulf between today’s trend inwireless access and the AmI future. AmI assumes agreat deal of infrastructure, although modular, low-cost, easy-to-install devices don’t seem so futuristic(for example, consider the “Basic Home InformationSystem” in Table 1). We propose the following strat-egy for achieving convergence in these two fields:that personal communication devices or terminals beconsidered as optional modalities when wired modal-ities also exist, and that an internet-based agent ormiddleware provider make the decisions about whichmodality to use at any particular time, to best servethe needs of the user.

We note that this is a generalization of the “alwaysbest [wireless] connection” idea from cognitive radio.The generalization is that the best connection may bewired and the best modality may be off-body. Thegeneralization moves the control away from the per-sonal wireless device or wireless service provider toan internet-based agent. This movement has profoundimplications on business models for services.



Convergence of these fields does not mean that wire-less providers and equipment manufacturers shouldcease their efforts to provide faster and better wire-less access to the user, because there will always beplaces with no or only limited off-body modalities.Also, there will be times when the user chooses not touse an off-body modality for privacy reasons. On theother hand, most of us spend a great deal of time inrelatively private spaces, which in the near future canbe economically equipped with off-body, pervasivecomputing-type devices. The great majority of theseoff-body devices will be connected to the internet andwired to power. Also the success of ring tones and ofmulti-player video gaming shows that people like toexpress themselves in public places and entertainthemselves on a shared display. As ubiquitous com-puting becomes a reality, we conjecture that the per-centage of time that personal communication devicesrequire high-bandwidth connections will decline.

User Acceptance

The most sophisticated technology will not be suc-cessful if users are afraid of it or view it as too diffi-cult to use. The biggest challenges to user acceptanceare expected to be

• Hardware and software support• Simple and intuitive user interface• Security, privacy and protection from SPAM• Pulling information to initially build the user

profile.

The envisioned superior service will be a fantastictechnology, assuming it all works. However, theservice will have unprecedented complexity. Withpotentially both on- and off-body presentation andextraction modalities, and the maintenance of theuser profile and the subscriptions, the level needed ofpersonal hardware and software support will exceedwhat corporate employees have grown to expect fromtheir information technology (IT) support department.

Assuming the system works, then the user interfacemust be simple enough for a non-technical personto use easily. Of course, the interface should includevoice recognition and voice synthesis. However, withsensing technology and context awareness, otherforms of communication may be possible, for exam-ple, off-body video sensors could interpret minimalhand motions in the air instead of mouse movementsand clicks. One research group thinks that the detec-tion and synthesis of emotion will be the key to thesuccess of human-machine interaction [12].

The quality of the user profile is central to the useful-ness and convenience of 4G-SMILING. Once the

user profile is initialized, then conceivably, it can beautomatically updated through a combination of con-text awareness, mood sensing, and occasional userinputs. We imagine that mood sensing could be per-formed by an “interpreter service” that would takeas its inputs relevant context information (location,activity, time-of-day) and available sensed data aboutthe user’s psychological state. If it is determined thatthe user doesn’t like something in a certain context,as shown in Figure 3, then the user profile can beupdated to reflect this new information.

A possibility for improving the user profile is “inten-tional perturbation”. We imagine that most users willnot want static profiles. Users need exposure to newthings, but how fast, what and when? Of course, 4GSmust include web browsing and opportunities for theuser to directly modify his or her profile. In addition,however, we imagine that a variety of ways could beused to systematically perturb a user’s profile, andthen use the interpreter service to bring the profile toa new optimized state. Example bases of systematicperturbation may include (1) what your friends aredoing, (2) purchaser communities (“others whobought this product also bought…”), (3) fashionshows and celebrity profiles, (4) systematic randomperturbations, and (5) advertising.

Potential 4G-SMILING Architecture

Figure 4 shows a concept for the 4G-SMILING archi-tecture. Enclosed within the bold rectilinear outline isthe “Service Manager” (SM), which incorporates theuser profile (UP), the UP editor and filter, the veryimportant “Push Computer,” various agreements thatthe user has with web-based services, access services,and content providers, and the “Optimal ContentMapper.” Any connections from the SM to entitiesoutside the SM are assumed to be through the inter-net. Enclosed within the dashed curve are local enti-ties that we are assuming would be most efficientlylocated in relatively close physical proximity to the

“I do notlike it”

Interpreter

On-bodysensors

Off-bodysensors

Figure 3 Example of the interpreter



user. These include the various wireless access pointsor base stations, a local directory or “Locality ProfileService,” (LPS), and one or more Local Off-BodySensor Data Services. This means that the SM couldbe on a server anywhere connected to the internet.The off-body modality services, such as the servercontrolling the “Fail-Safe Cooking” modality set,would also be local to the user. Lower-data rate ser-vices, such as the Context Detection Service, are notrequired to be near the user.

We observe that the Optimal Content Mapper (OCM)is both within the SM and considered a local entity.This is because the OCM will be handling high-datarate content, such as high-resolution video. If the restof the SM is on a distant server, it would not makesense for the video to first be shipped to that distantserver before being sent to the user. The OCM wouldneed to reside on a server local to the user, possiblythe same one that handles the Off-Body ModalityServices.

The LPS need not necessarily be close to the user, butto have a scalable system (i.e. to avoid too much traf-fic at a Regional or National Profile Server), localinformation should be distributed in its associatedlocality.

The Push Computer determines the content or ser-vices that will be presented to the user. It bases thisdetermination on (1) requests from the user or frompush applications, (2) the subset of the UP that hasbeen identified by the Context-Based UP Filter, (3)the modality availabilities and agreements, and (4)the content and service availabilities. It then requestsor commands that the content or service be deliveredto the user via the server that holds the OCM.

Queries and modifications can go both ways betweenthe Push Computer and the Content and ApplicationProviders. The most useful applications will querythe Push Computer about the User Profile. For exam-ple, a “travel assistant” application may detect that a

Content and application providers

UserProfile (UP)

Context detectionservice

Off-bodywirelessmodality

list

Context-based UP

editor

Wirelessaccess

Off-body

Locality profile service

Modalityagreements

Optimalcontentmapper

Pushcomputer

Mood interpreterservice

Media &services

Localoff-body

sensor dataservice

Localphysicalbusiness

list

Off-bodymodalityservices

WiMAXcellularWLAN

bluetoothP2P

Service manager

Local Entities

Q

Q

Q

Q

Q

Q

Content and service agreements

Infotainment

Queries, modifications andcontext (low rate data) High-rate data

Personal finance

Business apps

Health

Q

Context-based UP

filter

Off-bodywired

modalitylist

Figure 4 Possible 4G-SMILING architecture



flight has been cancelled, and will need to consult theUser Profile to determine such things as friends andassociates at the destination who should be notified,and user preferences about hotels so reservations canbe changed. On the other hand, the Push Computercan query the applications for possible context infor-mation. An example of this is as follows. The popu-larity of gaming today, especially internet gaming,suggests that in the future, “virtual” context (i.e.within a game) may be as important as physical con-text (just ask any parent who has ever tried to inter-rupt a teenager playing an internet game). Finally,application history may cause a modification to theUser Profile, while the Push Computer can cause achange of state in an application.

Storing the User Profile on a server, rather than onthe user’s personal device, may seem risky, since itwill contain sensitive and personal data. However, wepropose that it should not be on the user’s personaldevice because of the nature of push applications. Asstated earlier, they are long-term running programs,which will perform better with an always-on connec-tion so they can interact constantly with other data-bases and applications. Having it on the internetmeans the user is not tied to a single personal wire-less device, and enables the user to maintain and ben-efit from the services even if he or she temporarilyloses the use of a personal wireless device. Also, stor-ing the UP on the internet is more consistent with thenotions of pervasive or ubiquitous computing, whichrequire only RFID to identify users. Finally, we notethat much sensitive and personal data (e.g. medicalrecords) is already stored on databases not directlyunder users’ control. The security of this UP data isof utmost importance and constitutes one of thebiggest challenges of 4GS.

Some Related Fields and Projects

The 4GS vision is very interdisciplinary and includeselements in the fields of pervasive or ubiquitous com-puting, ambient intelligence (AmI), context aware-ness, cooperation between heterogeneous radioaccess networks, content distribution, sensor net-working, location-based services, music and imageinformation retrieval, emotion sensing, wirelessnetwork security, semantic web development [15],dynamic partitioning of hardware and software, andbusiness model development.

There are also many existing projects that are rele-vant. Some are focused on a subset of relevant topics,or are focused on particular applications. Some con-sider different approaches. The following is a discus-sion of a few, mostly European, existing projectsrelated to the 4G-SMILING (4GS) vision. They

include Magnet Beyond [5], Puma [2], HUMAINE[12], MobiLife [9], Omnipresent [6], e-Sense [7],Daidalos [8], CRUISE [13] and SPICE [10]. This listonly partially captures the activities going on todayin this subject area.

The predecessor to Magnet Beyond (MB) is Magnet[9], which is about how to create a Personal Network(PN) that is a secure, trusted and seamless connectionbetween a user’s personal area network (PAN) andremote networks that supply the user’s applicationsand services, such as the office network and the homenetwork. MB extends the PN to federations of PNsthat are associated with other people with whom theuser would like to share information [5,14,6]. Thesefederations are defined by the security and privacyboundary that encloses the federated PNs [5,14,6].MB is also concerned with context awareness. Secu-rity is adapted to the service or application beingused and the user profile. Service discovery withinthe PNs and federated PNs is made to be contextaware; example contexts are location (e.g. home,office, hospital, public) and role (parent, spouse,child, employee, etc.). Device power management isalso a consideration in MB. Reconfigurable hardwareand adaptive mapping of algorithms onto differenttypes of computing platforms (ASICs, FPGAs, etc.)are methods for saving battery life that are currentlybeing investigated. In MB, peer networking is consid-ered for the devices with the PN, and the short-rangeair interfaces are optimized for the PAN applications.

The focus of Magnet and MB is the provision of thesecure connections within and between PNs. In con-trast, the focus of 4G-SMILING is on the ultimateservice application. 4GS depends on the connectionsto the extent that they impact the quality of the ser-vice that is delivered. 4GS starts with the application,and asks how the underlying network should be opti-mized to provide the best possible service to the user.Topics of relevance to 4GS that are outside the cur-rent scope of MB include:

• User acceptance• Automatic perception of user reaction• PN resilience and dependability• Adaptation of content to modality• Access rights to database, user anonymity, trust

establishment• Global service discovery outside of federated PNs• Exploiting non-PN peer-to-peer networks for scala-

bility and low power• Relating content profiles and personal profiles to

achieve personalized content• Distribution and control of profile information• Business models.



The Puma System, developed at the IBM T.J. WatsonResearch Center, appears to be the “first generic sys-tem for accessing Web applications from arbitrarycollaboration modalities” [1]. The “collaborationmodalities” are conventional ones associated withbusiness applications (IM, email, phone) [1]. Pumais a context-aware, push application, which employsa “Pusher,” which is similar to the Service Manager(SM) in Figure 4. The Pusher includes the “PushEngine,” which is similar to the Push Computer ofFigure 4, and a “Modality Resolver” and “ModalityBots,” which together are like the OCM in Figure 4.Puma provides the “view aspects” of an applicationin a modality-independent format.

The 4GS vision represents an extension of Puma toinclude a much broader set of content and applica-tions, and a larger set of modalities, including off-body modalities, with a mapping that minimizespower consumption on the on-body devices. 4GS alsoincludes systematic profile perturbation and moodinterpretation.

Like Puma, Daidalos considers a middleware app-roach to managing multiple modalities, however itrestricts to on-body wireless modalities. Daidalosassumes that the telecom operator “takes the role ofthe service provider and offers the most crucial ser-vices to its customers,” [8] and focuses on operator/provider challenges when a piece of middlewareautomatically chooses the “best” wireless interface inresponse to user mobility and needs [8]. It considerscontext awareness and personalization, with emphasison quality of service (QoS), security, and how thebusiness models change from vertical to horizontalwhen the user has a virtual identity that is separatefrom a specific air interface [8]. In contrast to Daida-los, the 4GS vision does not assume that the telecomoperators are the chief providers of services. This isbecause there will probably be multiple modalitiessimultaneously available to the user, likely to be pro-vided by different telecom operators, and the user,or at least the user’s Service Manager, will choosewhich combination of modalities (and therefore whichtelecom operators) to exploit. Also, 4GS extendsDaidalos to include off-body modalities, mood inter-pretation, and systematic profile perturbation.

Like our 4GS vision, the goal of MobiLife [9,16] isto exploit off-body modalities, however it is in starkcontrast to 4GS because apparently the content andservices flow through the user on-body device beforebeing transmitted to the off-body devices. In otherwords, the user device acts as an intermediate serverin the data path. While the MobiLife approach maybe more scalable, it is not battery-power efficient and,if a wired connection to the modality is available, the

MobiLife approach might provide a much less reli-able connection (two wireless hops in place of awired connection).

The e-Sense project is mainly about how to integratesensor networks and Beyond 3G infrastructure [7].The emphasis is on providing the internetwork con-nections to enable context awareness. The 4GS visionwould extend e-Sense to support more modalities,including wired modalities, and is more applicationoriented.

The SPICE project addresses the challenge of how todeliver context-aware and personalized services as auser moves from one administrative domain to an-other [10,17]. It therefore has a lot in common withthe 4GS vision. One SPICE objective is the develop-ment of semantics for the user profile data, so that thedata can be efficiently used by different applicationsand services. Another is development of models andclassification for context data. A Provisioning Frame-work and a Content Broker are proposed, which aresimilar to the Service Manager and Context DetectionService, respectively, in Figure 4. A chief differencebetween SPICE and 4GS is that in SPICE all themodalities are on the user’s mobile device, whereasin 4GS, the emphasis is on always off-loading com-putation, communication, and/or modalities awayfrom the user’s personal devices to those resourcesin the user’s environment, when they are availableand when it is advantageous to do so.

HUMAINE is a European Network of Excellenceproject that is about developing emotion-sensitivemultimodal interfaces so that humans and machinesmay interact in a way that is natural to the humans[18]. HUMAINE and similar work is a critical partof the 4GS vision; it will be needed for all the mod-alities. For example, many of the services might bedelivered by a familiar face in the on- or off-bodyvisual displays. A more explicit indication of theseconcepts in Figure 4 is the direct connection betweenthe Mood Interpreter Service and the User Profile.

CRUISE (CReating Ubiquitous Intelligent SensingEnvironments) is another European Network ofExcellence project with the objective of coordinationand network among researchers of wireless sensornetworks [13]. This includes integration and sharingof testbeds and research tools. CRUISE topics arenecessary for the context-awareness aspects of 4GS.

Conclusions

In this paper, we have explored and imagined someof the features of the ultimate mobile service, whichwe refer to as the “service cocktail” or 4G-SMILING



(4GS). Two of the most important are that it wouldbe highly context-aware and personalized. We haveobserved that two fields, Ambient Intelligence (AmI)and Mobile Communications, are both dedicated toproviding the mobile user with the ultimate in ser-vice, yet they make very different assumptions aboutthe requirements for mobile wireless access. Wesuggest that when the user is in an AmI or similarlyenabled environment, as many functions as possibleand reasonable be off-loaded to the wired environ-ment, to provide a more reliable connection, to savepower, and to provide a better experience for the user.Some of the many challenges remaining to realize4GS include detect what off-body modality packagesare currently available to the user, mapping the con-tent to each modality, security and privacy of the userprofile, and how to update and intentionally perturbthe user profile.

References

1 Chandramouli, B et al. Pushing the Envelope ofPervasive Access. In: IEEE International Confer-ence on Pervasive Services (ICPS), July 2005.

2 Abowd, G A et al. The Aware Home: DevelopingTechnologies for Successful Aging. Workshopheld in conjunction with American Association ofArtificial Intelligence (AAAI) Conference 2002,Alberta, Canada, July 2002.

3 Mitola, J III. Cognitive Radio Architecture.Wiley, 2006.

4 EUSAI – Second European Symposium onAmbient Intelligence. URL: http://www.eusai.net/

5 IST. MAGNET Beyond (IST-FP6-IP-027396).URL: http://www.magnet.aau.dk

6 Calin, D, McGee, A R, Chadrashekar, U, Prasad,R. MAGNET: An approach for secure personalNetworking in Beyond 3G Wireless Networks.Bell Labs Technical Journal, 11 (1), 79–98 2006.

7 e-sense – Capturing Ambient Intelligence forMobile Communications through Wireless SensorNetworks. URL: http://www.ist-esense.org/

8 Aguiar, R, Farshchian, B A, Sarma, A, Einsiedler,H. Daidalos: the global architecture and its instan-tiations. Proc. of the 15th IST Mobile Summit,Myconos, Greece, 8 June 2006.

9 IST MobiLife. URL: http://www.ist-mobilife.org/

10 Zhdanova, A V et al. Context acquisition, repre-sentation and employment in mobile service plat-forms. Proc. of the 15th IST Mobile Summit,Myconos, Greece, 8 June 2006.

11 Kernchen, R, Mrohs, B. Context-aware multi-modal output selection for the device and modal-ity function. 6th International Workshop onApplications and Services in Wireless Networks,Berlin, 29-31 May 2006.

12 Schröder, M, Cowie, R. Developing a consistentview on emotion-oriented computing. In: Renals,S, Bengio, S (eds.) Machine Learning for Multi-modal Interaction. Springer LNCS 3869,194–205, 2006.

13 URL: http://www.telecom.ece.ntua.gr/cruise/

14 Prasad, R, Olsen, R L. The unpredictable future:Personal networks paving towards 4G. Telektro-nikk, 102 (1), 150–160, 2006.

15 Hepp, M. Semantic web and semantic web ser-vices, father and son or indivisible twins? IEEEInternet Computing, March-April, 85–88, 2006.

16 Kernchen, R et al. Context-awareness in MobiL-ife. Proc. of the 15th IST Mobile Summit,Mykonos, Greece, June 2006.

17 European IST-FP6 project SPICE. URL:http://www.ist-spice.org/index.html

18 Humaine. URL: http://emotion-research.net/

19 Greenfield, A. All watched over by machines ofloving grace: Some ethical guidelines for userexperience in ubiquitous-computing settings.12/01/2004. URL:http://www.boxesandarrows.com/view/all_watched_over_by_machines_of_loving_grace_some_ethical_guidelines_for_user_experience_in_ubiquitous_computing_settings_1



Mary Ann Ingram received the BEE and PhD degrees from the Georgia Institute of Technology (Georgia

Tech) in 1983 and 1989, respectively. From 1983 to 1986, she was a Research Engineer with the Georgia

Tech Research Institute in Atlanta, performing studies on radar electronic countermeasure (ECM) systems.

In 1986, she became a graduate research assistant with the School of Electrical and Computer Engineering

at Georgia Tech, where in 1989, she became a Faculty Member and is currently Professor. Her early

research areas were optical communications and radar systems. In 1997, she established the Smart

Antenna Research Laboratory (SARL), which emphasizes the application of multiple antennas to wireless

communication systems. The SARL performs system analysis and design, channel measurement, and proto-

typing relating to a wide range of wireless applications, including wireless local area network (WLAN) and

satellite communications, with focus on the lower layers of communication networks.


For a presentation of Ramjee Prasad, please turn to page 3.

Kim Skaue was appointed CEO at C3 (Convergence, Connectivity and Communication) faculty at the

University of Aalborg in 2006. The C3 faculty is a 4G innovation center. Kim Skaue has extensive

experience in the telecom industry serving as Vice President Strategy and Business Development at

Sonofon (Denmark) and Vice President in Telenor Nordic (Norway) with direct responsibility for building

a cross Nordic mobile operation. He was the strategic architect behind the very successfull MultiPlan

(vPBX) concept in Sonofon, a concept which today accounts for half of Sonofon’s revenues. Besides his top

management positions Kim Skaue is a board member in several companies and serves as external lecturer

at Aarhus Business School teaching Innovation and Business Development. Kim Skaue holds an MSc

electronic engineering – MBA in International Management and is a Top Governance graduate.



1 Historical Background

The International Telecommunication Union (ITU) isan intergovernmental organisation and a specialisedagency of the United Nations (UN) for telecommuni-cations consisting of 191 Member States and over600 Sector Members. ITU’s main objective is to pro-mote connectivity and interoperability between itsmembers and to foster the development of all kindsof telecommunications worldwide.

The organisation was established in Paris on 17 May1865 under the name of the International TelegraphUnion to undertake the task of harmonising intercon-nection of the national telegraph networks. Norway

was one of the “founding fathers” of the Union andNorway has taken active part in the work of theUnion since its inception.

ITU offers a unique partnership between MemberStates and the private sector by allowing the privatesector to become so-called Sector Members in ITU’sthree sectors. With the liberalisation of the telecom-munication sector in Norway, Telenor has chosen totake part in the activities of the ITU by being a SectorMember of ITU’s three sectors – the Radiocommuni-cation Sector (BR), the Telecommunication Standard-ization Sector (TSB) and the TelecommunicationDevelopment Sector (BDT). Norway is a MemberState of the ITU and the Ministry of Transport andCommunications is responsible for Norway’s mem-bership in the ITU, whereas the Norwegian Postand Telecommunications Authority (NPT) has beenappointed Norway’s “ITU Administration” and isresponsible for the day-to-day management of ITUrelated matters on behalf of Norway.

2 General about ITU

Plenipotentiary Conference

The Plenipotentiary Conference (PPC) is ITU’s toppolicy-making body. The PP is held every four yearsand sets the Union’s general policies, adopts four-year strategic and financial plans and elects the seniormanagement of the organisation, the Council (theUnion’s governing body) and the Radio RegulationsBoard (RRB). The PP is the key event where ITU’sstrategy for the next four years is decided, determin-ing the Union’s ability to influence the developmentand growth of information and communication tech-nologies (ICT) worldwide in the light of changes inthe industry and the needs of its membership. As anintergovernmental conference only Member Stateshave the right to send delegations to the Plenipoten-tiary Conference. Each Member State has one vote.A number of international organisations and SectorMembers may attend the PP as observers.

3 The Plenipotentiary Conference

2006 in Antalya, Turkey

The 17th ITU Plenipotentiary Conference took placefrom 6 till 24 November 2006 in Antalya, Turkey,

ITU Plenipotentiary Conference 2006 – PP-06,

Antalya, 6–24 November 2006

An Overview of Main Results of the Conference

A N N E L I S E L I L L E B Ø

Anne Lise Lillebø

is Director,

Telenor ASA,

Group

Regulatory

The article gives an account of the main results of PP-06 seen from a Telenor and a Sector Member

perspective.

Sungate Port Royal Hotel, Antalya, Turkey – Venue of PP-06



hosted by the Telecommunication Authority ofTurkey and was opened by the Prime Minister ofTurkey, His Excellency, Recep Tayyip Erdogan. Theopening ceremony was also addressed by ITU Secre-tary-General, Yoshio Utsumi, and Turkey’s Ministerof Transport, Binali Yildirim. Almost 2000 delegatesparticipated in the PP-06 from over 173 countriesrepresenting both government and the private sectoras well as regional and international organisations.

In addition to the national delegations, there wereobservers from the United Nations and their spe-cialised agencies, regional telecommunication organi-sations, intergovernmental organisations operatingsatellite systems and Sector Members. ETNO – theEuropean Telecommunications Network Operator’sAssociation – was represented as an observer.

This is the first time that the PP lasted for three weeks;previous PPs have had four weeks to finalise thework, and the time schedule was very tight. Electionsof new top management attracted major attention asfour of the five elected officials were outgoing andcould not stand for reelection in their present electedpost. Although elections started on Thursday in thefirst week of the conference, it took another week tofinalise all the elections, and this clearly hampered theprogress of the essential work which was handled inthe Committees 5 and 6 and the Working Group of thePlenary. There is a growing tendency that the PPsbecome more and more politicised with issues suchas elections taking too much time and attention.

Other issues high on the agenda of the PP-06 werequestions relating to developing countries and coun-tries with economies in transition, ITU’s role regard-ing Internet, ITU’s role in the follow up of the WorldSummit on the Information Society (WSIS), bridgingthe digital divide, the role of civil society in the workof the ITU, combating spam, enhancing cyber secu-rity and the finances of the Union.

3.1 The Tasks of the Plenipotentiary

Conference

The tasks of the conference are laid down in Article 8of ITU’s Constitution

• To decide on the strategic direction of ITU anddevelop new policies that will shape the worldwidedevelopment of telecommunications and informa-tion and communication technologies (ICT);

• Adopt the ITU Strategic Plan which outlines theUnion’s orientations, goals and priorities for theperiod 2008–2011;

• Decide on the Union’s Financial Plan that willprovide the resources needed to meet the goalsand deliverables set out in the Strategic Plan;

• Elect the Union’s five highest-ranking officials:Secretary-General, the Deputy Secretary-Generalas well as Directors of the three Sectors of theUnion: the Radiocommunication Bureau (BR), theTelecommunication Standardization Bureau (TSB)and the Telecommunication Development Bureau(BDT). The conference will also elect the membersof the Council and of the Radio Regulations Board;

• Consider and adopt proposals for amendments toITU’s Constitution and Convention put forward byMember States;

• Treat proposals submitted by Member States to thePP-06.

3.2 Election of Chairman and

Vice-Chairmen of the Conference

The Plenipotentiary Conference 2006 was very ablychaired by Dr. Tanju Cataltepe, Turkey, and KnutSmaaland, Norwegian Post and TelecommunicationsAuthority, was elected as one of the five vice-chair-men. The vice-chairmen are elected on the basis of

Sungate Port Royal Hotel. Taurus Mountains behind



Chairman and Vice-Chairmen of the Conference:

• Dr T. Cataltepe (Turkey) chairman

Vice-Chairmen:

• K. Smaaland (Norway)

• L. Reiman (Russian Federation)

• P. Mvouomo (Republic of Congo)

• H. Chono (Japan)

• M.J. Mulla (Saudi Arabia)

Committee Chairman

Committee 1: Steering Committee Dr. T. Cataltepe, Turkey

Composed of the chairman and the vice-chairmen of the conference and the chairmen and

vice-chairmen of the other committees and of the Working Group of the Plenary

Committee 2: Credentials Committee E. Ndukwe, Nigeria

Committee 3: Budget Control Committee R. Gonzales Bustamante, Mexico

Committee 4: Editorial Committee M.-T. Alajouanine. France

Committee 5: Policy and Legal Matters K. Arasteh, Islamic Republic of Iran

Committee 6: Administration and Management F. Riehl, Switzerland

Working Group of the Plenary on the World Summit on the Information Society R.N. Agarwal, India

Dr Tanju Cataltepe, Turkey, Chairman of PP-06

Knut Smaaland, Norwegian Post and Telecommuni-cation Authority, Vice-Chairman of PP-06, Region B– Western Europe

a fair geographical distribution among ITU’s fiveadministrative regions; viz. Region A – the Americas,Region B – Western Europe, Region C – EasternEurope, Region D – Africa and Region E – Asia/Australasia.

3.3 Structure of the Conference

– Committees and Chairmen

and Vice-Chairmen

The Plenipotentiary Conference has four statutorycommittees: the Steering Committee, the CredentialsCommittee, the Editorial Committee and the BudgetControl Committee. In addition the PP-06 agreed toset up two substantive committees and one workinggroup of the plenary:

• Committee 5 – Policy and Legal MattersTo consider policy matters of the Union, to exam-ine proposals for amending the Constitution andthe Convention and to consider any other questionsof a legal nature raised during the conference;

• Committee 6 – Administration and ManagementTo consider the draft strategic plan, to prepare draftfinancial policies and a draft financial plan for2008 – 2011;

• Working Group of the Plenary on the WorldSummit on the Information Society (WSIS)To consider issues related to the outcome of theWSIS and the Internet.



the right to speak. The representatives from Telenorwere accepted by the Ministry of Transport and Com-munications to be members of the Norwegian delega-tion as delegates.

5 Norwegian and Nordic

Preparations to the PP-06

5.1 Norwegian Preparations

A preparatory meeting for the PP-06 was arranged bythe Norwegian Post and Telecommunications Author-ity and the Ministry of Transport and Communica-tions in July 2006 where all interested parties inNorway were invited to attend. NPT made a broad pre-sentation of the CEPT European Common proposals(ECPs) and other relevant issues for Norway. Telenorparticipated in the meeting.

5.2 NITU

On the Nordic level, preparations for the PP-06 havebeen carried out in the NITU group (Nordic ITUcooperation) which normally meets once a year.

Members of the Norwegian Delegation

Member Function Affiliation

Eva Hildrum Head Ministry of Transport and

Secretary General Communications

Ottar Ostnes Head Ministry of Transport and

Director General Communications

Jens C. Koch Deputy Head Ministry of Transport and

Director Communications

Willy Jensen Deputy Head Norwegian Post and

Director General Telecommunications Authority

Knut Smaaland Delegate Norwegian Post and

Special Adviser Telecommunications Authority

Kjersti Hamborgstrøm Delegate Telenor Satellite Broadcasting

Manager

Anne Lise Lillebø Delegate Telenor ASA

Director

Tom Dahl-Hansen Delegate Norwegian Post and

Adviser Telecommunications Authority

Marie-Thérèse Alajouanine, France, Chairman of theEditorial Committee (Committee 4)

Eva Hildrum, Secretary General of the Ministry ofTransport and Communications, Head of NorwegianDelegation

Ottar Ostnes, Director General, Ministry ofTransport and Communications, Head ofNorwegian Delegation

4 Norwegian Delegation to the

PP-06

The Norwegian delegation was headed by Eva Hil-drum, Secretary General of the Ministry of Transportand Communications, and Ottar Ostnes, DirectorGeneral, Ministry of Transport and Communications.Willy Jensen, Director General, Norwegian Post andTelecommunications Authority, and Jens C. Kock,Director, Ministry of Transport and Communications,were Deputy Heads of the Norwegian delegation.The delegation also included representatives from theNorwegian Post and Telecommunications Authorityand Telenor ASA. Telenor was represented by KjerstiT. Hamborgstrøm, Telenor Satellite Broadcasting,and Anne Lise Lillebø, Telenor ASA.

Since the Plenipotentiary Conference is an intergov-ernmental conference, it is only open for direct partic-ipation from ITU Member States. Sector Membersare only allowed to participate as observers without



NITU is an informal group consisting of Nordic tele-com regulators with representatives from the privatesector invited to attend. For the last four-year period,NITU has been chaired by Norway in its capacity asITU Council member in Region B (Western Europe).Telenor has taken part in these Nordic preparationstogether with representatives from the Ministry ofTransport and Communications and the NorwegianPost and Telecommunications Authority. This timethere were no Nordic contributions to the PP-06, asmost of the preparations are done in a European con-text, and the Nordic countries have all been active inthis work.

The Nordic countries (Denmark, Finland, Iceland,Norway and Sweden) have agreed on a rotationscheme regarding Nordic candidacies to the ITUCouncil. After a period of four years, Norway hasdecided to step down and the Nordic candidate forthe Council in the next four-year period 2007 – 2010is Sweden.

6 European Preparations for the

PP-06

6.1 CEPT

Europe’s regional organisation for post and telecom-munications authorities CEPT – the European Com-mittee for Post and Telecommunications with 47member countries – has been in charge of the Euro-pean preparations for the PP-06. The ElectronicCommunications Committee (ECC) of CEPT has aWorking Group ITU dedicated to general ITU mat-ters including the preparations for Council and thePP. Observers from APT, CITEL and ETNO havebeen invited to participate as observers in the meet-ings of the CEPT WG ITU, and this has proved veryuseful for the dissemination and exchange of informa-tion regarding the European positions for the PP-06.

The CEPT WG ITU set up a special Project Team –CEPT WG ITU PT PP-06 to prepare the PP-06. ThePT PP-06 was chaired by Knut Smaaland, NorwegianPost and Telecommunications Authority. The aim ofthe comprehensive preparatory work is to developEuropean Common Proposals – ECPs – to be submit-ted to the PP-06 as European proposals. The PT PP-06 prepared comprehensive briefing material for thePP-06 and succeeded in agreeing on 28 ECPs whichwere all submitted as formal European proposals tothe PP-06.

6.2 ETNO

ETNO - the European Telecommunications NetworkOperators’ Association – has a Working Group ITUdedicated to overall ITU matters. The ETNO WG

ITU has been responsible for preparing the PP-06 onbehalf of its members. The ETNO WG ITU is chairedby Dominique Würges, France Telecom, and hasconcentrated on issues of special interest to the ITUSector Members such as the retention of the ratiobetween Member States contributions and the SectorMember contributions, no widening of the scope ofthe Union, enhanced efficiency of work of the Union,possible abolition of the International Telecommuni-cation Regulations (ITRs), and ITU’s role in a postWSIS environment. The ETNO WG ITU issued anETNO Reflection Document on strategic develop-ment of ITU which was also submitted to CEPT PTPP-06 for consideration.

Anne Lise Lillebø, Telenor ASA and KjerstiHamborgstrøm, Telenor Satellite Broadcasting –Delegates in the Norwegian Delegation

Mr Dominique Würges, France Telecom;ETNO observer



At the PP-06 in Antalya, ETNO was formally repre-sented by Michael Bartholomew, ETNO Director,Thierry Dieu, Communications Manager andDominique Würges, France Telecom, Chairmanof the ETNO WG ITU.

6.3 Cooperation between CEPT and ETNO

The CEPT WG ITU and ETNO have agreed on amutual exchange of observers, and ETNO observerswere invited to attend all the meetings of the CEPTWG ITU and the PT PP-06. The cooperation betweenEuropean telecom regulators and authorities andEuropean telecom operators has taken place in avery good and transparent environment and has beenhighly appreciated by the private sector. Europe isone of the few regions that associates so closely itsprivate sector to the preparation of the PlenipotentiaryConference and such cooperation is key to ensuringthat the ECPs (European Common Proposals) arediscussed among various stakeholders in Europe andthat positions reflect market needs.

6.4 The European Commission

The European Commission (EC) participates as coun-cillor in the CEPT WG ITU and the PT PP-06 and theEuropean Union has recognised CEPT as the compe-tent body to carry out European preparations formajor ITU conferences.

CEPT and the European Commission (EC) organiseda joint workshop in Brussels on 19 October 2006 toinform embassies and diplomatic delegations in Brus-sels on CEPT and European preparations and policyviews for the PP-06. ETNO was also invited to thisworkshop and Michael Bartholomew, the ETNODirector, made a presentation outlining ETNO viewson PP-06 issues of interest to the private sector. Thismeeting was a good occasion for exchanging views onmajor issues prior to the conference and it provided theopportunity to explain the positions taken by CEPT.

7 Major Results from PP-06

7.1 General

As at previous conferences there was a lack of will tomake decisions and a tendency to postpone decisionson important issues to the subsequent PP. Due to lackof consensus no financial plan was adopted and anumber of issues such as the management and func-tioning of the Union, and the management of thebudget will be studied by future Council WorkingGroups. On a positive note both ETNO and its mem-bers were fairly satisfied with the results regarding nochange in scope of ITU’s mandate, issues regardingInternet, the status quo of the system determiningSector Members’ financial contributions and nochange for the ITRs.

7.2 Elections of ITU’s Top Management

The election of the Union’s five highest-ranking offi-cials, the members of the Radio Regulations Board(RRB) and the Council is one of the major tasks ofthe PP, and PP-06 saw a record high number of can-didates for the elected posts. The incumbent Secre-tary-General, the Deputy Secretary-General, theDirector of the Telecommunication StandardisationBureau (TSB) and the Director of the Telecommuni-cation Development Bureau (BDT) were not eligiblefor reelection, and there were a total of seven candi-dates for the Secretary-General, four candidates forthe Deputy Secretary-General, four for the TSBDirector and four candidates for the BDT Director.Only the director of the Radiocommunication Bureau(BR) was eligible for reelection, and he was not chal-lenged by any contenders.

The PP-06 agreed to start elections on Thursday inthe first week of the conference, and it took a whole

Members of the Norwegian Delegation: Eva Hildrum, Head, Ministryof Transport and Communications; Jens C. Kock, Deputy Head,Ministry of Transport and Communications; Knut Smaaland,Delegate, Norwegian Post and Telecommunication Authority

Members of the Norwegian Delegation: Willy Jensen, Deputy Head,Norwegian Post and Telecommunications Authority; Ottar Ostnes,Head, Ministry of Transport and Communications; and Jens C. Kock,Deputy Head, Ministry of Transport and Communications



week to finalise elections of the top management, theRRB and the Council. Elections attract much atten-tion, and despite good efforts from the chairmen ofCommittees 5 and 6, very little progress was madeduring this period. This is also due to the fact thatmany Member States with candidatures refrain fromexpressing views on contentious issues before theelections are finalised. From a Sector Member pointof view this situation is detrimental to the progress ofwork and considerably reduces the efficiency of theconference. Out of three weeks, the main part of thejob was performed in the last one and a half weeks!

Despite efforts to coordinate European candidatures,there were two European candidates for the post ofSecretary-General, one European candidate for theDeputy Secretary-General and two European candi-dates for the TSB Director. This was a very difficultsituation for Europe and in our opinion, it stronglyweakened the possibilities of having Europeanselected to any of the four posts that were to be filled.

7.2.1 Secretary-General and

Deputy Secretary-General

Hamadoun Touré from Mali was elected as Secre-tary-General in the third round of voting with amajority 95 votes out of 155 Member States presentand voting. The required majority for election was78 votes. Mr Touré has already served eight yearsas the Director of the BDT. The CEPT candidate,Mathias Kurth from Germany, received 60 votes,whereas the other European candidate, Marc Furrerfrom Swizerland, withdrew his candidature after thesecond round.

Houlin Zhao, People’s Republic of China, waselected as Deputy Secretary-General by a landslidemajority of 93 votes in the first round of voting. TheEuropean candidate, Carlos Sanchez from Spain,received 34 votes. Mr Zhao is outgoing Director ofTSB where he has served for two terms, totallingeight years.

7.2.2 Bureau Directors

BR – Radiocommunication Bureau

Valery Timofeev from the Russian Federation wasreelected as Director of the RadiocommunicationBureau in the first round. There was no candidatechallenging Mr Timofeev and he received almostunanimous support from the Plenipotentiary.

TSB – Telecommunication Standardization Bureau

Malcolm Johnson of UK was elected as Directorof the Telecommunication Standardization Bureau(TSB) in the third round. The other European candi-date, Mr Bigi from Italy, withdrew after the firstround. This election turned out to be a thriller with aclose race among the remaining candidates. Mr John-son was elected with 83 votes against 79 votes for theJapanese candidate Mr Inue. From a Sector Memberpoint of view, we believe that Mr Johnson is wellpositioned to take up the challenging task of manag-ing the TSB. For many years Mr Johnson has been anactive proponent for changing TSB’s way of workingwith the aim of speeding up activities and increasingthe role of the private sector. Mr Johnson’s experi-ence of reform work of ITU will serve as a good basisfor his challenging tasks in the TSB.

BDT – Telecommunication Development Bureau

Sami Al-Basheer of Saudi Arabia was elected asDirector of the Telecommunication Development

ITU’s new top management. Left to right: Sami Al-Basheer, Director of Telecommunication DevelopmentBureau; Houlin Zhao, Deputy Secretary-General; Hamadoun I. Touré, Secretary-General; Valery Timofeev,Director of the Radiocommunication Bureau; and Malcolm Johnson, Director of the Telecommunication Stan-dardization Bureau

Ph

oto

: IT

U/J

ea

n-M

arc

Fe

rré

ITU’s top management team 2007 – 2010

Post Elected candidate

Secretary-General Hamadoun Touré, Mali

Deputy Secretary-General Houlin Zhao, People’s Republic of China

Director, Radiocommunication Valery Timofeev, Russian Federation

Bureau (BR)

Director, Telecommunication Malcolm Johnson, United Kingdom

Standardization Bureau (TSB)

Director, Telecommunication Sami Al-Basheer, Saudi Arabia

Development Bureau (BDT)



Sector (BDT) in the third round of voting with a totalof 91 votes while his contender Patrick Masambu,Uganda, received 70 votes.

ITU’s new management team will serve for the four-year period 2006 – 2010 and PP-06 decided that thenew officials will take up their duties on 1 January2007.

7.2.3 Elections to the Council

At present there are 46 seats at the ITU Councilreflecting Article 4 in the Convention that the numbershall not exceed 25 per cent of the total number ofMember States. The seats are distributed among thefive administrative regions of the Union as follows:

• Region A – Americas (8 seats)• Region B – Western Europe (8 seats)• Region C – Eastern Europe (5 seats)• Region D – Africa (13 seats)• Region E – Asia/Australasia (12 seats)

In Region B – Western Europe – there were nine can-didates for eight seats. Norway has been a member ofCouncil in Region B for the past four-year period, butbased on the Nordic rotation scheme, Norway haddecided to step down and Sweden was this time theNordic candidate for Council in Region B. Sweden waselected with a comfortable number of votes and will

serve on the Council for the next four-year period up tothe next Plenipotentiary in 2010. Once more UnitedKingdom failed to be elected as member of the Council.

Telenor is pleased to see that the Nordic countries –through Sweden – managed to keep a seat on the ITUCouncil. If needed, matters of interest to our com-pany within the remit of Council can be taken up inthe Nordic cooperation group NITU, and providedthat there is agreement among the Nordic MemberStates, this can be brought to Council by Sweden.One example of such matters is the Council Decisionto open a trial of offering free electronic access toITU-T Recommendations in 2007.

In Region C – Eastern Europe – there were eightcandidates for five seats. In the elections held on 16November, Ukraine and Poland received 94 voteseach, and a second ballot took place on 17 Novemberto determine the fifth member of the Council forRegion C. In the second ballot Ukraine received 77votes and Poland 69 votes.

From Telenor’s point of view, we highly appreciatethe fact that some of the countries where Telenor hasmobile operations have been elected to the Council.In Region C, this applies to the Russian Federationand Ukraine and in Region E – Asia/Australasia, toThailand, Malaysia and Pakistan.

7.2.4 Elections to the Radio Regulations

Board (RRB)

The Radio Regulations Board has a total of 12members based on a geographical distributionamong ITU’s administrative regions:

• Region A – Americas, 2 seats,• Region B – Western Europe, 2 seats• Region C – Eastern Europe, 2 seats• Region D – Africa, 3 seats• Region E – Asia and Australasia, 3 seats

Since Telenor has important mobile operations inPakistan and Malaysia we note with pleasure that thecandidates from these two countries were elected asmembers of the RRB in region E.The results of the elections to Council for Region B –

Western Europe (8 seats)

Country Votes

France 140

Spain 134

Switzerland 133

Germany 132

Sweden 132

Italy 123

Portugal 121

Turkey 120

The results of the elections to Council for Region C –

Eastern Europe (5 seats)

Country Votes

Russian Federation 135

Romania 102

Bulgaria 99

Czech Republic 96

Ukraine 77

Valery Timofeev, Russian Federation, re-elected BRDirector, with the author



7.3 The European Common Proposals –

ECP – Results at PP-06

CEPT Member States had submitted a total of 28European Common Proposals (ECPs) to the PP-06.CEPT members were pleased to see that a number ofECPs were accepted by the conference. In the follow-ing focus will be on ECPs of interest to Sector Mem-bers who are members of ETNO:

ECP no 5: Sector Members observers at Council

In this ECP CEPT seeks to clarify the status of SectorMembers being observers and proposes to delete thetext stating that Sector Members “may be representedat meetings of the Council, ...” as it is difficult tosee how a few Sector Members can represent sucha diversified group as Sector Members in the threesectors of the ITU. PP-06 agreed to this deletion, andreplaced “represented” by “attend” and the updatedtext reads as follows: “Sector Members may attend,as observers, in meetings of the Council, ...”. Thenew text has been added to No 60B, Article 4 in theConvention. ETNO Sector Members agree with thisamendment.

ECP no 8: International Telecommunication

Regulations – ITRs

The current ITRs were adopted by ITU’s WorldAdministrative Telegraph and Telephone Conference(WATTC) in Melbourne in 1988. The agreed treatyrepresented a very delicate compromise between thecountries that had just started on market liberalisationand the developing countries. The most important ofthe articles of the ITRs from 1988 concern “Chargingand accounting” (of international telecommunica-tions), the “International network” and “Internationalservices”.

The need for a review of the ITRs was considered bythe Plenipotentiary Conferences in Minneapolis in1998 and in Marrakesh in 2002. The PP in Marrakeshresolved that Council should set up a Council Work-ing Group (CWG) to undertake a review of the ITRsand report back to PP-06. However, this CWG didnot reach any consensus and presented three differentviews on a possible way forward: to leave the ITRsunchanged, to amend the ITRs including adding newprovisions, or to terminate the ITRs and transfer cer-tain provisions to the CS, CV and ITU-T Recommen-dations.

Most European countries consider that the ITRs intheir present form are substantially out of date. TheITRs no longer reflect the major commercial andoperational changes which have taken place sincethen, especially the widespread liberalisation oftelecommunications services and the competitiveinternational telecommunications environment.

The European proposal to PP-06 was based on theview that the current ITRs no longer serve the pur-pose for which they were designed. The ECP on theITRs proposes the following:

• ITU-T should identify the operational issues in theITRs and if possible, develop ITU-T Recommenda-tions for approval by the World Telecommunica-tion Standardization Assembly in 2008 (WTSA-08);

• Further to identify which parts of the ITRs thatmerit treaty level status and those which are redun-dant (already covered in the CS/CV or obsolete);

• WTSA-08 (World Telecommunication Standard-ization Assembly) should make recommendationsto the PP-10 on any further action including thepossible termination of the ITRs;

• A decision on a possible WCIT (World Conferenceon International Telecommunications) should bepostponed until the PP-10.

A future Plenipotentiary Conference should be giventhe explicit authority to terminate the ITRs. InEurope’s view the holding of a WCIT is very costlyand would put additional strain on the tight budgetof the Union.

As a Sector Member and telecommunications opera-tor Telenor fully supported the CEPT view in thismatter. ETNO members are in favour of authorising afuture Plenipotentiary Conference to revise and possi-bly terminate the ITRs and to transfer relevant provi-sions of the ITRs to the CS/CV. As an alternativeETNO members could also accept to keep status quo.However, ETNO members are opposed to the idea of

Marianne Treschow, Director-General, NationalPost and Telecom Agency, Deputy Head of SwedishDelegation



a World Conference on International Telecommuni-cations which is believed to open up the possibility ofexpanding the scope of the Union and introduce newregulation for Internet and mobile telecommunication.

This issue turned out to be one of the most conten-tious questions of the PP-06. Many Member Statesfrom Asia were in favour of launching yet anotherstudy of the provisions of the ITRs and wanted topostpone any decision regarding the instrument to thePP-10. While a number of Member States from theAmericas wanted to retain the ITRs as is, there was agreat majority of Member States from the Arab coun-tries, Africa and certain developing countries thatwanted to retain the ITRs and to amend the instru-ment to cover new areas.

The results of the discussions are embedded in Reso-lution COM5/4 “Review of the International Tele-

communication Regulations”: The resolution con-cludes as follows:

• ITU-T should undertake a review of the existingITRs in cooperation with the other sectors, withITU-T as the focal point. The World Telecommuni-cation Policy Forum (WTPF) scheduled for 2009 istasked to deal with Internet related issues, conver-gence and NGN. The WTPF will prepare reportsand possibly opinions for consideration by ITUmembers.

• A World Conference on International Telecommu-nication (WCIT) will be convened in Geneva in2012.

• The Council is instructed to adopt the agenda ofthis WCIT by 2011 and urges the sectors to carryout studies within their field of competence toprepare for the WCIT.

ITU’s legal adviser informed that according to hisinterpretation of the CS/CV it is not appropriate for aPP to have the authority to terminate the ITRs. Thiscan only be carried out by the legal entity which cre-ated the treaty. However, a PP can transform itselfinto a WCIT during or after a PP if the MemberStates agree to this.

CEPT members found that the compromise solutionin Res COM5/4 was acceptable as a possible WCITwas postponed for another six years. The next PP in2010 can postpone it further, if so agreed. The out-come of this issue will be a challenge to the ITUMember States especially as regards the financialconsequences for the Union. Considerable prepara-tions are required in order to hold a WCIT, and inour opinion, it is questionable whether the result willbenefit the future work of ITU. From a Sector Mem-ber’s point of view it is positive that there will be nochange regarding the ITRs in the next six years. Thewhole question will come up for debate at the PP-10,and no-one can predict the development within thetelecommunications field in this timeframe. It is tobe hoped that the world has changed and that moreMember States will understand that the ITRs are notan adequate way of regulating international telecom-munications and that such countries will be able tosupport Europe and other countries in their effortsto abrogate the ITRs.

ECP no 11: TELECOM

The TELECOM exhibitions operate in a very compet-itive environment, and Europe proposes to remove thedistinction between regional and worldwide TELE-COM exhibitions and forums as the distinction is nolonger considered relevant in the present economic

7Nordic team: Anders Frederich, Deputy Head, Swedish Delegation;Kjersti Hamborgstrøm, Delegate, Norwegian Delegation; and JørnJensby, Delegate, Danish Delegation

The results of the election of the

Radio Regulations Board

RRB members 2007 – 2010:

Region A: Americas (2 seats)

Julie Napier Zoller, United States

Robert W. Jones, Canada

Region B: Western Europe (2 seats)

Mindaugas Zilinskas, Lithuania

Martine Limoudin, France

Region C: Eastern Europe (2 seats)

Baiysh Nurmatov, Kyrgyzstan

Wladyslaw Moron, Poland

Region D: Africa (3 seats)

Hassan Lebbadi, Morocco

Shola Taylor, Nigeria

Aboubakar Zourmba, Cameroon

Region E: Asia and Australasia (3 seats)

Ali Ebadi, Malaysia

Shahzada Alam Malik, Pakistan

P.K. Garg, India



context. In the ECP it is proposed that ITU shouldonly organise world TELECOMs which could rotatein the ITU Regions. Europe also sees a need for moretransparency concerning the TELECOM Board andproposes that the Secretary-General’s proposal for thecomposition of the Board should be approved by theITU Council. In the ECP it is also stressed that theTELECOM events should be financially successful.

ETNO supported the CEPT proposal and underlinedthe commercial role of TELECOM exhibitions andthat they should be organised according to marketdemand and not be driven by geographical or politi-cal considerations.

It turned out that this was a very sensitive issue, espe-cially for developing countries and countries witheconomies in transition. It was argued that regionalevents would bring the potential benefits of telecom-munications closer to the people of all continents byfocusing on regional problems and proposing solu-tions adapted to local environment. The Europeanproposal to merge world and regional TELECOMinto a single global event was not supported.

The PP-06 concluded in Resolution 11 “World andregional telecommunication/information and commu-nication technology exhibitions and forums” that ITUshould continue to organise world and regional TELE-COM exhibitions and forums on a regular basis,taking due account of the need to ensure the financialsuccess of such exhibitions. The principle of rotationis retained and now it also applies to world eventswhen several countries have submitted competitiveoffers. The resolution stresses the importance of hav-ing a transparent process open to all interested partiesin the selection of the venue for world ITU TELE-COM events. When selecting the venue the financialviability of the event should be considered as well asthe results of market and feasibility studies. The com-position of the Board is to be approved by Council.

It is noted that the final resolution supports the con-tinuation of both world and regional events. Consid-eration of financial viability and the transparency inthe selection of the venue of world events including arotation system for deciding on the venue of regionalITU TELECOM events are welcomed by the privatesector

ECP no 12: The future role of ITU in implementing the

outcomes of the World Summit on the Information

Society (WSIS)

The first session of the WSIS was held in 2003 andresulted in a WSIS Geneva Declaration of principlesand the Geneva Plan of Action which focuses on ITUcore functions of importance to the information soci-

ety such as assistance in bridging the digital divide,fostering regional and international cooperation,spectrum management and development of standards.

The second phase of WSIS was held in Tunis in 2005and adopted the Tunis Commitment and the TunisAgenda for the Information Society where specificaction lines were identified. There will be a multi-stakeholder implementation of the Geneva Plan ofAction and the Tunis Agenda, and ITU was identifiedas possible moderator/facilitator for action line C2(information and communication infrastructure) andC5 (building confidence and security in the use ofICTs) of the Tunis Agenda and as a potential partnerfor a number of other action lines (C1, C3, C4, C6,C7 and C11). Attention is also drawn to the WTDC-06 (World Telecommunication Development Confer-ence) in Doha which adopted Resolution 30 on “Therole of the ITU-D in implementing the outcomes ofthe WSIS”.

The draft Resolution text proposed by Europe opensup the possibility of a structured and focusedapproach to implement the action lines identified aspriorities and calls for the deployment of ITU exper-tise. It is pointed out that the implementation of theWSIS outcomes can be done within the current man-date of the ITU. The proposal also emphasises theneed for partnerships with actors outside the ITUtowards delivering WSIS outcomes.

The PP-06 agreed on Resolution GT-PLEN 6 on“ITU’s role in implementing the outcomes of theWorld Summit on the Information Society” whichreiterates ITU’s commitment to take part in the multi-stakeholder follow up of WSIS and acknowledgesITU’s special role as moderator/facilitator for imple-menting action lines C2 and C5 and requests theCouncil to oversee ITU’s implementation of theWSIS outcome. PP-06 requests the Council to main-tain the Council Working Group on WSIS – CWGWSIS – to facilitate guidance on the ITU implemen-tation of relevant WSIS outcomes.

The agreed resolution captures Europe’s views onthe role that ITU may be able to play post-WSIS andrecognises ITU’s existing role. ETNO fully concurswith the European position that ITU’s role in thefollow up of WSIS can be done within its presentmandate and does not see any need for widening thescope of the Union or change its name. ETNO alsosupports the creation of the Internet GovernanceForum (IGF) agreed by the second session of theWSIS. IGF should be seen as an excellent opportu-nity for effective multi-stakeholder involvement inpolicy shaping, recognising the shared responsibilityof all stakeholders.



ECP no 13: “Clarification of the responsibilities of the

SG in the management of the Union and the discon-

tinuation of the Coordination Committee”

The proposal calls for greater clarity with respect tothe roles of the Secretary-General, the Deputy Secre-tary-General and the Directors of the three Bureaux.The CS and CV should clearly state that the Secre-tary-General is responsible for the overall manage-ment of the Union’s resources and is accountable forthis to the Council. According to the current Article 6of the Convention, decisions can only be taken byconsensus between the elected officials. This can infact hamper decisive management and can result inCouncil being forced to take management decisionswhich should properly be for the officials of theUnion. With such a clarification of the roles, Europeproposes to discontinue the Coordination Committee.

Although this ECP does not directly affect SectorMembers, ETNO supported the general thrust ofthe proposal: to streamline management proceduresamong the ITU elected officials and to avoid unnec-essary bureaucracy.

There was no consensus on the CEPT proposal, butelements of this ECP are included in another Resolu-tion (Res COM5/5 ) which is treated in the followingparagraph regarding ECP No 14.

ECP No 14: Appointment rather than election of the

Directors of the Bureaux

This question is a recurrent issue and has been on theslate for a number of years. In contrast to most UNagencies that have two elected officials, ITU has atotal of five elected officials. The existing arrange-ment of five elected officials politicises the manage-ment of the organisation and creates a lack of clarityas to who is accountable for the decisions made.Europe finds that this federal structure adds to thecomplexity in the management of the Union andcompromises its efficiency.

Instead Europe favours a system of two elected offi-cials: the Secretary-General and the Deputy Secretary-General. It is proposed that the Bureaux Directorsshould be appointed according to usual UN practice.Europe proposed to establish a Council Working Groupto consider the change to an appointment process.

ETNO members have for many years argued that theBureaux Directors should be appointed based on aprofessional job description. Although the presentCEPT proposal goes in the right direction, ETNOwould have preferred that the present PP-06 shoulddecide on the matter instead of leaving the issue to aCouncil Working Group. This means that a decisioncan first be taken by the PP in 2010 and any imple-mentation of a possible decision would take placefor the PP in 2014.

Unfortunately there was very little support for thisEuropean proposal at the PP-06. Member States fromthe Americas, Africa and Asia-Pacific agreed that thestatus quo with five elected officials functioned in agood manner and that there was no need to changethe system. From a Sector Member’s point of viewthis is surprising when account is taken of discussionsin Council on how to manage the Union and all thetime spent on arranging elections of elected officialsat the PP-06. Elections of the five officials were ter-minated on Thursday in the second week. As atten-tion was concentrated on the elections, work in thecommittees suffered from lack of commitment andimportant decisions were deferred till after the elec-tions. Some Administrations were clearly reluctantto discuss contentious issues as this might have animpact on the pending elections. With fewer electionsto perform the work of the PP could be done in amore efficient manner and the duration of the PPcould even be reduced.

The PP-06 agreed to cover issues regarding responsi-bilities and accountability of the elected officials, theCoordination Committee, the elections of the electedofficials and reporting in the Union in ResolutionCOM5/5 “Study on the management and functioningof the Union”. The Resolution resolves that theCouncil should conduct a study aiming at overallimprovement of the efficiency of ITU management,addressing the following issues:

• Reporting structure in the Union• Role, accountability, number and tenure/term of

office of elected officials• Functioning of the Coordination Committee• Election procedures• Responsibility, accountability and transparency of

the advisory groups.

Ian Hutchings, Head, New Zealand delegation, and KjerstiHamborgstrøm, Delegate, Norwegian Delegation



Telenor notes that this study is only open for the par-ticipation of Member States. In our view this limita-tion of participation is not very constructive and actu-ally prevents the Sector Members from discussing thefunctioning of the advisory groups in the three sectorswhere they can participate directly in their capacity asSector Members.

ECP No 16 “Follow-up of Resolution 110 (Marrakesh,

2002) (Review of the contribution of Sector Members

towards defraying the expenses of the International

Telecommunication Union) – Reduction of time of

denunciation of Sector Members

Following Resolution 110 from PP-02, a CouncilWorking Group reported its findings to the Council 05:

• The period of time recommended from the datewhen the notification of the denunciation of aSector Member has been received by the Secretary-General is proposed to be reduced from one year tosix months. The same provision will also apply forAssociates. This is supported by Europe. In addi-tion, European countries propose to modify CV480to allow Sector Members to decrease their class ofcontribution in case of exceptional circumstances.

The aim of the proposal is to improve the financialoversight of Sector Members and Associates to thebenefit of the financial stability of ITU.

The PP-06 agreed to these proposals and CV240 andCV480 have been modified accordingly. The confer-ence also adopted Resolution COM6/3 “Improvementof management and follow-up of the defrayal of ITUexpenses by Sector Members and Associates”.Resolves 2 in this Resolution lays down that “in caseof a merger between Sector Members or Associatesof the same Sector, duly notified to the Secretary-General, No. 240 of the Convention shall not applyand shall thus not have the effect of requiring theSector Member or the Associate resulting from themerger to pay more than one contribution for its par-ticipation in the work of the Sector concerned.”

As a Sector Member Telenor welcomes these modifi-cations which are better adapted to the rapid pace ofthe market and the financial realities faced by private-sector entities.

ECP No 17 “The terms of office of the elected officials”

In line with the current provisions elected officialsare re-eligible to the same post only once, but there isno limit for an elected official presenting his candida-ture for one of the other four elected posts when hehas finalised his two possible terms. In theory, thesame person can remain in elected posts for 40 years!This prevents the possibility of having new people

with fresh ideas in the management of the Union.Europe proposes changes to Article 2, No 13 of theConvention to cover this point.

Although elections are the responsibility of the Mem-ber States, the ETNO Sector Members welcomed thisproposal. In our opinion, all members of ITU willbenefit from a competent and efficient management,and all organisations need a degree of renewal in theirmanagement teams.

This proposal turned out to be very controversial, butafter extensive discussion the European proposal wassupported and the PP-06 agreed to modify Article 2,No 13 of the Convention as follows: “Elected Offi-cials: …., and they shall be eligible for re-electiononce only for the same post. Re-election shall meanthat it is possible for only a second term, regardlessof whether it is consecutive or not”.

ECP No 20: “Resolution 130 of Marrakesh on the role

of ITU in information and communication network

security” – Cybersecurity and combating spam

The Resolution from Marrakesh addresses ITU’s rolein information and communication network security.Since 2002 new threats have emerged with the globaldevelopment of Internet such as spam, protection ofpersonal data, privacy, intellectual property protec-tion, fight against illicit content etc. that couldthreaten the security and stability of telecommunica-tions networks. The World Telecommunication Stan-dardization Assembly (WTSA) in Florianopolis in2004 adopted two resolutions on the issue of spam:combating spam and countering spam by technicalmeans. The second session of the WSIS in Tunis in2005 also underlined the importance of security inICTs in its Agenda of Tunis and established ActionLine C5 “Building confidence and security in the useof ICTs” where ITU is identified as the moderator/

Wearing the same jackets from PP-98: Houlin Zhao, Deputy Secretary-General Elect, People’s Republic of China, with the author



facilitator for this WSIS Action Line. The WorldTelecommunication Development Conference(WTCD) in Doha 2006 also identified cybersecurityas a priority activity of the BDT.

Europe proposed to extend the scope of Resolution130 from Marrakesh to cover all aspects concerningpossible threats to the stability and security of net-works.

The PP-06 agreed to update Resolution 130 anddecided on a new title: “Strengthening the role ofITU in building confidence and security in the useof information and communication technologies”.The updated resolution instructs the Director of theTelecommunication Development Bureau ( BDT) todevelop, consistent with the results of WTDC-06 andthe subsequent meeting pursuant to Resolution 45(Doha, 2006) of that conference, the projects forenhancing cooperation on cybersecurity and com-bating spam responding to the needs of developingcountries, in close collaboration with the relevantpartners. Members are invited to develop the neces-sary relevant legislation and reference is made toregional initiatives such as the Council of Europe’sConvention on Cybercrime. Finally the Directors of

the three Bureaux are encouraged to pursue the workof security of ICTs in the respective Study Groups,relevant projects and to continue collaboration withrelevant organisations.

It is expected that increased focus will be put onissues regarding security in ICTs and the role thatITU can play on the world scene.

ECP 21: “Updating and fusion of Resolution 102 of

Marrakesh on the Management of Internet Domain

names and Addresses and Resolution 133 of Marrakesh

on the Role of administrations of Member States in the

management of internationalized (multilingual)

domain names”

In the wake of the WSIS, it is necessary for ITU toreassess how the Union can contribute to the realisa-tion of the Tunis Agenda based on its expertise andrecognised mandate. CEPT proposed to update Res102 of Marrakesh dealing with IP addresses anddomain names and to integrate Res 133 of Marrakeshabout implementation of IDNs (Internet Domainnames) into Res 102 to have only one resolution deal-ing with Internet resources as a whole. The proposednew resolution on Internet resources invites ITU tocontribute constructively to the work on Internet gov-ernance. ITU should follow the work of the InternetGovernance Forum (IGF) established by the secondphase of the WSIS in Tunis in 2005. ITU is requestedto improve its cooperation with relevant organisationsand to participate in the development of globally-applicable principles on public policy issues relatedto the coordination and management of Internetresources.

There was no consensus on merging the two resolu-tions, but the resolutions were updated to reflect therecent development within this area after the twophases of the WSIS.

The PP-06 agreed on an updated text of Res 102reflecting a number of CEPT’s points. The updatedresolution 102 is called “ITU’s role with regard tointernational public policy issues pertaining to theInternet and the management of Internet resources,including domain names and addresses”. The Resolu-tion refers to the role given to ITU in this area: thatITU is dealing with technical and policy issuesrelated to IP-based networks including the Internetand evolution to NGN, ITU’s development ofENUM, internationalised domain name and countrycode top-level domain (ccTLD) and emphasises thatITU should start a process of enhanced cooperationwith all stakeholders in accordance with the TunisAgenda. The Secretary-General is instructed toorganise consultations on these issues among the ITUmembers and other stakeholders and submit propos-

Eva Hildrum, Head of Norwegian Delegation



als to the Council 2007 through the Council WorkingGroup on WSIS.

Likewise, Resolution 133 “Role of administrationsof Member States in the management of internation-alised (multilingual) domain names” was updated.The Resolution refers to the Geneva Action Plan andthe Tunis Agenda of the WSIS to advance the processfor the introduction of multilingualism in a numberof areas including domain names and instructs theSecretary-General to take an active part in all inter-national discussions and initiatives on the deploymentand management of internationalised Internet domainnames, in cooperation with relevant organisationssuch as WIPO and UNESCO.

The two updated resolutions are somewhat enlarged,and seen from a Sector Member perspective, it isimportant that responsibilities taken on by ITU inthese resolutions fall within its present mandate.

ECP No 23: “Cross-references between the CS/CV and

the Radio Regulations”

In this proposal it is pointed out that the definitions ofthe broadcasting service and the mobile service areboth in the Constitution and the Convention and inthe Radio Regulations whereas all other service defi-nitions are in Article 1 of the Radio Regulations. Bydeleting the definitions in the CS/CV, future WRCswill have the possibility of updating the definitionsif necessary taking into account the development ofconvergence between different telecommunicationsservices and without being bound by provisions inthe CS/CV. Many European Sector Members were infavour of this proposal, as it would remove an obsta-cle to a possible change in the Radio Regulations ifa future WRC should agree on changes in the defini-tions. However, the proposal to delete the definitionsof broadcasting and mobile service from the Constitu-tion and Convention was not very well received anddid not obtain enough support to go forward.

ECP 25 Gender mainstreaming in ITU

CEPT countries proposed to amend Resolution 70from Marrakesh in order to reflect Resolution 55adopted by the World Telecommunication Develop-ment Conference (WTDC) in Doha, 2006, promotinggender equality towards all-inclusive informationsocieties.

PP-06 agreed to revise Resolution 70 “Gender main-streaming in ITU and promotion of gender equalitytowards all-inclusive information societies” endors-ing Resolution 55 (Doha, 2006) on promoting genderequality towards all-inclusive information societies;to continue the work being done at ITU, and particu-larly in BDT, to promote gender equality in ICTs by

recommending measures at the international, regionaland national level on policies and programmes thatimprove socio-economic conditions for women, par-ticularly in developing countries; and to incorporatethe gender perspective in the implementation of theITU strategic plan and financial plan for 2008-2011as well as in the operational plans of the Bureaux andthe General Secretariat.

7.4 Budget and the Financial Plan

One of the main tasks of the Plenipotentiary Confer-ence is to establish a basis for the budget of the Unionand determine related financial limits. This impliesestablishing the total number of contributory units forthe period up to the next Plenipotentiary Conferenceon the basis of the classes on contribution announcedby Member States. For the past four years the finan-cial situation of the Union has been extremelystrained, and the Secretary-General has been forcedto implement a number of cost saving schemes tobalance income and expenditure. These measuresinclude a system of results-based budgeting, opera-tional planning and time-tracking, the introductionof cost recovery, and reforms to the secretariat andthe business model of TELECOM.

7.4.1 Decision 5 “Income and expenditure for

the Union for the period 2008 to 2011”

The definitive choice of class of contribution by Mem-ber States is made before the end of the conference.Sector Members may choose their class of contributionwithin a period of three months after the Conference.The Plenipotentiary Conference shall adopt the defi-nitive Financial Plan including the upper limit of thevalue of the contributory unit which will serve as basisfor establishing the budgets to be adopted by the Coun-cil during the financial period concerned.

Kirsten Bak, Head, Danish Delegation



It is noted with dismay that the PP-06 did not manageto adopt the financial plan for the next four-yearperiod due to lack of consensus.

However, PP-06 adopted Revised Decision 5“Income and expenditure for the Union for 2008 to2011” with a total expenditure of CHF 664 million,a negative balance of CHF 39 million.

The upper limit of the amount of contributory unitof Member States for the years 2009 – 2011 shall beCHF 330,000; for the years 2008 – 2009 the contribu-tory unit of Member States shall not exceed CHF318,000 and a cap of CHF 85 million for the years2008 – 2011 was set for expenditure on interpreta-tion, translation and text processing regarding the sixofficial languages of the Union. This means that theupper limit of the amount of the Member State con-tributory unit remains the same as for the previousplenipotentiary period.

The Decision contains 18 options for reducing expen-diture, including

• Limitation of the number of Study Group meetingsand their duration;

• Limitation of the number of days for the advisorygroups to three days per year maximum. Additionalmeetings may be held on a cost recovery basis; i.e.costs are financed by the requesting Sectors;

• Cost savings through better management of theITU regional presence;

• Reduction in the cost of documentation of confer-ences and meetings;

• Consideration of savings in languages (translation,interpretation) for Study Groups meetings and pub-lications;

• Reduce the number of meetings of the RRB fromfour to three per year.

7.4.2 Ratio Between Member States and Sector

Members’ Contributions

Two options of the quadrennial Draft Financial Planfor the period 2008 – 2011 were proposed. The onlydifference between the two is the ratio for determin-ing the contributory unit payable by Sector Members.Option 1 is based on the current ratio of one fifth ofthe contributory unit paid by Member States. Option2 reflects a revision of the ratio to one fourth of thecontributory unit paid by Member States. Option 1indicates a shortfall of income of CHF 33.3 million

whereas Option 2 indicates a shortfall of income ofCHF 12 milllion.

A number of Member States supported a change in theratio between Member States and Sector Members’contributions from one fifth to one fourth, represent-ing an increase of 24 % of the value of the SectorMember unit of contribution. It was argued that thecontributory unit of Sector Members did not cover thedirect costs of their participation in the Sectors, andthat Sector Members should pay more in line withtheir alleged increased rights over the last few years.

ETNO members were strongly against this unilateralincrease of Sector Members’ contributions only andcautioned that instead of generating more income tothe sectors, such an increase could lead to SectorMembers leaving the sectors or that they might preferto go into national delegations instead. Attention wasdrawn to the vast contribution made by Sector Mem-bers in the work of the sectors which should be takeninto account when assessing the contribution of Sec-tor Members. It would be a very negative sign to theoutside world if ITU should decide to only increasethe contributions from the private sector in an effortto balance its budget.

After a lengthy and very heated debate, the ratio willnot be changed at this juncture.

7.5 The Strategic Plan 2008 – 2011

The Strategic Plan for the Union for 2008 – 2011 wasadopted in Resolution 71.

The resolution points to the many developments thathave taken place in the telecommunication and theinformation and communication technology (ICT)environment that have significant implications forITU as a whole. A total of seven goals are listed inthe strategic plan:

Goal 1: Maintaining and extending internationalcooperation among all Member States and with rele-vant regional organizations for the improvementand rational use of information and communicationinfrastructure of all kinds, taking the appropriateleading role in United Nations system initiatives onICTs, as called for by the relevant WSIS outcomes.

Goal 2: Assisting in bridging the national and inter-national digital divides in ICTs, by facilitating inter-operability, interconnection and global connectivityof networks and services, and by playing a leadingrole, within its mandate, in the multistakeholder pro-cess for the follow-up and implementation of the rele-vant WSIS goals and objectives.



Goal 3: Widening the Union’s membership, extend-ing participation and facilitating cooperation of anincreasing number of administrations and organiza-tions, as well as new actors, such as relevant WSISstakeholders.

Goal 4: Developing tools, based on contributionsfrom members, to promote end-user confidence, andto safeguard the efficiency, security, integrity andinteroperability of networks.1)

Goal 5: Continuing to improve the efficiency andeffectiveness of ITU’s structures and services andtheir relevance to the requirements of membershipand the wider global community.

Goal 6: Disseminating information and know-howto provide the membership and the wider community,particularly developing countries, with capabilities toleverage the benefits of, inter alia, private-sector par-ticipation, competition, globalisation, network secu-rity and efficiency and technological change in theirICT sector, and enhancing the capacity of ITU Mem-ber States, in particular developing countries, forinnovation in ICTs.

Goal 7: Promoting the development of an enablingenvironment that assists governments in fosteringsupportive, transparent, pro-competitive, harmonizedand predictable policies, as well as legal and regula-tory frameworks that provide appropriate incentivesfor investment in, and development of, the informa-tion society.

7.7 Radio Related Matters

7.7.1 Implementation of Additional Corrective

Measures Relating to Cost Recovery for

Satellite Network Filings – Decision COM6/1

As a satellite operator Telenor has taken particularinterest in this issue which has been discussed for anumber of years in the Council and at various PPs.With the recent decisions made by Council 05, it isclear that certain invoices for Satellite Network Fil-ings had been issued on an incorrect basis. However,since the accounts were already drawn up, only thePP has the authority to cancel invoices in ITU.

In Decision COM6/1 “Implementation of additionalcorrective measure relating to cost recovery on satel-lite network filings” the PP-06 decided to implementthe corrective measures set forth in the Council-05Decisions 531, 532 and 534 and in the RRB decision

(41st meeting, Geneva, 4-8 September 2006) inrespect of invoices issued for the 2002-2003 period.

However, there was no agreement on the issueof cancelling unpaid invoices, and the issue wasdeferred to Council together with the delegation ofcompetence to write off the debt in this area.

7.7.2 Resolution 86 (Rev Marrakesh) Processing

Charges for Satellite Network Filings and

Administrative Procedures

PP-06 discussed Resolution 86 at length and decidedthat WRC-07 should consider the matter and reportto the next Plenipotentiary Conference as to whetherR86 should be deleted or updated.

7.7.3 Periodicity of WRCs and RAs

The periodicity of WRC and RA was discussed andextended from two to three to three to four years.CS Article 13 “Radiocommunication Conferences andRadiocommunication Assemblies” has been changedaccordingly (MOD 90, 91). The change reflects thecurrent situation where it is not feasible to finalise thecomprehensive preparations for a WRC within a timelimit of two to three years. Another reason for prolong-ing the periods between the WRCs is also the very dif-ficult financial situation of the Union. It is only possi-bly to sustain one major conference of the year withoutseriously affecting the balance of the budget.

7.8 Advisory Groups

A proposal from the United States to align theprovisions in the Convention regarding the advisorygroups in the three sectors caused considerabledebate. Although the proposal might seem to be asimple alignment, it could have great impact on theway the TSAG – the Telecommunication Standard-ization Advisory Group – is working. The three advi-

1) Information and communication network efficiency and security cover threats including, inter alia, spam, cybercrime, viruses, wormsand denial-of-service attacks.

Marianne Treschow, Deputy Head of SwedishDelegation, and Jørn Jensby, Delegate, DanishDelegation



sory groups are intrinsically different in nature andthe present provisions regarding the functioning ofthese groups are adapted to their different nature andindividual needs. The US proposal implied that thewording “... and act through the Director” should beapplied to all three advisory groups. This would meana change of substance to the work of the TSAG.

Telenor finds that TSAG functions fairly well todayand the advisory group is authorised to act on behalfof the WTSA in between two WTSAs in order to beable to speed up the work of the TelecommunicationStandardization Sector. From a Sector Member’spoint of view, there is no need to introduce a clausethat in our opinion will contribute to slowing downthe work of the TSAG.

There was substantial support for the US proposal,although many European Member States were againstany change to the provisions regarding TSAG. As acompromise, PP-06 agreed to align the text regardingthe Telecommunication Development AdvisoryGroup, but leave the provisions regarding TSAGunchanged. CEPT and ETNO members are pleasedwith the outcome of the discussions, but it is expectedthat this issue will be raised at another juncture.

7.9 Revised Resolution 122 The Evolving

Role of the WTSA

The PP-06 in Antalya updated this resolution totake into account i.a. the results of the WSIS in theGeneva Declaration of Principles which recognisesITU’s core competencies in the fields of informationand communication technologies and underscoresITU’s role as a unique, worldwide venue for govern-ment and industry to work together to foster thedevelopment and use of interoperable and non-dis-criminatory standards based on openness, and whichare both demand-driven and sensitive to the needs ofthe use. The WTSA is requested to encourage closecooperation and coordination with relevant standardsdeveloping organisations in both developed anddeveloping countries. The TSB Director is askedto consider organising a worldwide standardisationround table and coordination meeting possibly inconjunction with WTSA, for one day immediatelyprior to the assembly.

Telenor agrees to this updating which captures recentdevelopments regarding convergence and the WSISoutcome. The resolution reiterates the core tasks ofthe WTSA and links it to the strategic plan of theUnion for the next four-year period.

7.10 Use of Terminology in ITU Resolutions

Taking into account the outcome of the WSIS, manyMember States see a need for updating the terminol-

ogy relating to telecommunications in the basicinstruments of the Union, including a possible changeof name. ETNO did not see any need to change thescope and the name of ITU, and believed that thetasks bestowed on ITU regarding the follow up ofWSIS could be performed within its present mandate.There was no consensus for changing the mandateand the name, but the PP-06 adopted Resolution GT-PLEN/8 “Review of terminology used in the Consti-tution and Convention of the International Telecom-munication Union” which instructs the Council toestablish a working group open to Member States tostudy i.a. the use of the term “telecommunications” inthe CS and CV, to identify options for integrating anynew terminology in the CS and CV where appropriateand to consider what terms should be used in the CSand CV to reflect the impact of information and com-munication technologies (ICTs) and ICT applicationsin ITU activities. The Council WG will submit theirfinal report to PP-10. From a Sector Member’s pointof view, it is interesting to see that this group is lim-ited to participation from Member States.

7.11 Definitions and Terminology Relating

to Building Confidence

Res COM5/7 “Study of definitions and terminologyrelating to building confidence and security in the useof information and communication technologies”instructs Council to set up a Council Working Groupopen to all Member States and Sector Members tostudy the issue of terminology related to buildingconfidence and security in the use of ICTs, and toexamine and develop definitions in this regard. It isnoted that this CWG will be open to the Sector Mem-bers whereas the CWG on definition of telecoms islimited to Member States!!

7.12 World Telecommunication Policy

Forum

This Forum was considered by the PP-06 to be anexcellent instrument for discussing matters related toinformation on telecommunication policy and regula-tory matters on global and cross-sectoral issues. Withreference to ITU’s role in the follow up of WSIS andthe PP-06 Resolution on the review of the Inter-national Telecommunication Regulations, it wasdecided in Decision GT-PLEN/A “Fourth WorldTelecommunication Policy Forum” to convene thefourth WTPF in Geneva in the first quarter of 2009.As indicated in the Decision it is expected that themain topic for this Policy Forum will be the discus-sion on the future of the ITRs.

7.13 Strengthening the Regional Presence

A number of proposals called for an increased com-mitment from ITU in regional activities of the mem-bership. The Council has already adopted a number



of resolutions aiming at introducing measures tostrengthen the regional presence, and similar resolu-tions were adopted by the World TelecommunicationDevelopment Conference in Doha in 2006. This isnow reflected in the ITU Strategic Plan for 2008 –2011 which recommends strengthening communica-tion channels among BDT, the Member States andthe Sector Members and Associates and betweenBDT – both headquarters and the regional offices –and the General Secretariat and the R and T sectors.

The PP-06 adopted Resolution 25 “Strengthening theregional presence” where it is resolved to carry out areview of ITU’s regional presence. At the same timethe regional presence should be strengthened and abroadening of the information dissemination func-tions of the regional presence is required. Theregional offices should be expanded and strengthenedand they should be provided with greater autonomyin terms both of decision-making and of addressingcrucial needs of Member States in the region. TheSecretary-General is instructed to include an evalua-tion of the effectiveness of ITU’s regional presence inthe United Nations Joint Inspection Unit (JIU) workprogramme. The Director of the BDT is instructed i.a.to develop specific operational and financial plans forthe regional presence and to take the necessary mea-sures to ensure the effective incorporation of BR andTSB activities in the regional offices.

As a Sector Member with mobile operations in anumber of Asian countries, Telenor appreciates theinitiative to make the regional presence more efficientand more adapted to the needs of the region in co-operation with all three sectors of the ITU.

7.14 Next-Generation Network Deployment

in Developing Countries (NGN)

In the Geneva Declaration of Principles adopted byWSIS, it is emphasised that a well-developed infor-mation and communication network infrastructureand application which are easily accessible andaffordable and the increased use of broadband andother innovative technologies can accelerate thesocial and economic progress of countries and thewell-being of all individuals, communities and peo-ples. Account is taken of the fact that many develop-ing countries face the task of conducting a smoothmigration from existing networks to NGNs and thatNGNs can facilitate the delivery of a wide range ofadvanced ICT-based services for building the infor-mation society.

PP-06 adopted Resolution GT-PLEN/3 “Next-genera-tion network deployment in developing countries”where the Directors of the three Bureaux are in-structed to coordinate studies and programmes under

the Next-Generation Network Global Standards Ini-tiative (NGN-GSI) of ITU-T and of the Global Net-work Planning Initiatives (GNi) of the ITU-D, and tocoordinate ongoing work being carried out by studygroups as defined by WTDC (Doha, 2006).

This Resolution puts NGN on the agenda of all coun-tries and underlines the importance of deployingNGN and the information society world-wide, in-cluding the developing countries and countries witheconomies in transition.

7.15 Telecommunications in Emergency

and Disaster Situations for Mitigation

and Relief

In light of a number of natural disasters experiencedduring the past few years such as the tsunami in Asiaand a number of earthquakes, there is an increasedawareness of the importance of telecommunicationsin disaster relief. Resolution GT-PLEN/2 “The use oftelecommunications/information and communicationtechnologies for monitoring and management inemergency and disaster situations for early warning,prevention, mitigation and relief” underlines that aninternational standard for communication on alert andwarning information can assist in the provision ofeffective and appropriate humanitarian assistance andin mitigating the consequences of disasters, in partic-ular in developing countries.

7.16 The Global Symposium for Regulators

(GSR)

There was overall agreement that telecommunicationsregulators should continue to have a specific platform

Ottar Ostnes, Head, Norwegian Delegation



for sharing and exchanging matters concerning regu-latory issues in the form of the Global Symposiumfor Regulators. Some Member States wanted to in-clude the GSR in the basic instruments of the Union,but the compromise was to have a resolution – ResGT-PLEN/4 “The Global Symposium for Regulators”– where it is resolved that the GSR shall be estab-lished as a regular activity within the ITU-D pro-gramme and that the GRS should be arranged on anannual basis rotating in differing regions of the world.

As a member of the D-Sector, Telenor finds that theGRSs are very important events for regulators fromall over the world. There is considerable need forexchange of information and knowledge about how aregulator should function in a modern telecommuni-cations market, and ITU can play an important role indisseminating information and providing knowledgesharing.

7.17 Participation of Civil Society in the

Activities of the Union

After the WSIS, it became clear that ITU did notcater for the participation of the civil society in itswork. The WSIS concluded that all stakeholders havean important role to play in the development of theinformation society, and a number of Member Statesargued that ITU should adapt itself to the informationsociety and to allow participation also from the civilsociety.

A compromise agreement was reached in Res GT-PLEN/7 “Study on the participation of all relevantstakeholders in the activities on the Union related tothe World Summit on the Information Society” inwhich ITU is asked to conduct a study on the partici-pation of all relevant stakeholders in the activities ofthe Union related to WSIS. The Council is instructedto set up a Council Working Group open to MemberStates – or to task an existing working group – tostudy a number of aspects related to this issue suchas:

• Conduct open consultations on the inclusion of rele-vant stakeholders in the activities of ITU related tothe WSIS, including additional tasks to be per-formed by ITU as a result of theWSIS outputs;

• Develop a set of criteria for defining which stake-holders are relevant to participate in ITU activitiesrelated to WSIS;

• Identify efforts needed to ensure effective partici-pation of all relevant stakeholders from developingcountries.

As a Sector Member it is not clear how this willevolve and the outcome will be studied when avail-able.

7.18 Use of Six Official Languages of the

Union

Res COM6/5 “Use of the six official languages of theUnion on an equal footing” calls for the full imple-mentation of the use of the six official languages(Arabic, Chinese, English, French, Russian and Span-ish) on an equal footing and instructs the Council toreview interim measures for interpretation and trans-lation. The Council Working Group on Languagesshould continue to monitor progress and report to theCouncil on the implementation of this resolution.

Based on the fact that ITU is an intergovernmentalorganisation and part of the UN family, it is under-standable that the question of languages is of crucialimportance for a number of Member States. How-ever, when considering the predicted shortfall in thefinancial plan and the cap set for expenditure tointerpretation and translation, there is an imbalancebetween the wish of the PP-06 and the actual meansmade available for these activities. For the sake ofefficiency, it would have been preferred to conductmuch of the work in the Study Groups in the sectorsin one langue only.

8 Future Council Working Groups

Set up as a Consequence of the

PP-06 Decisions

A number of issues will be studied further, either inCouncil Working Groups (CWGs) or in a process tobe determined by Council.

The following Council Working Groups (CWGs)were established by the Council Extraordinary Ses-sion held in Antalya on 24 November 2006 immedi-ately after the PP-06:

• Management and budget group (MBG) (ResolutionCOM6/6). J Mendès, Portugal, was provisionallynominated as chairman. The CWG will consist ofMember States members of Council, the Secretary-General and the Directors of the Bureaux.

• Participation of all relevant stakeholders in theactivities of the Union related to the World Summiton the Information Society (Resolution GT-PLEN/7).A Cristiani, Argentine, was provisionally nomi-nated as chairman. The CWG is open to MemberStates only. Both Member States and Sector Mem-bers are invited to submit written contributions tothe group.



• Definitions on security of ICT (Resolution COM5/7).N Kisrawi, Syria, was provisionally nominated aschairman. The CWG is open both to MemberStates and Sector Members.

Council Working Group to be created at the Septem-ber 2007 Session of Council:

• Terminology in the Constitution/Convention ofITU (Telecommunications) (Resolution GT-PLEN/8). The CWG will only be open to MemberStates.

Studies to be developed by Council (process to bedetermined):

• Number of Member States members of Council(Resolution PLEN/1), open to Member States;

• Study on the management and functioning of theUnion (number of elected officials etc.) (Resolu-tion COM5/5) open to Member States.

Two old CWGs will continue:

• The role of ITU in the follow up of WSIS (Resolu-tion GT-PLEN/6);

• Use of languages (Resolution COM6/5).

9 Future Conferences,

Assemblies and Forums of the

Union (2008 – 2011)

• World Telecommunication Standardization Assem-bly (WTSA), between May and November 2008;

• World Telecommunication Policy Forum (WTPF),first quarter 2009;

• World Telecommunication Development Confer-ence (WTDC), March 2010;

• Plenipotentiary Conference (PP-10), October/November 2010;

• RA and WRC, February/March 2011.

Mexico has invited to host the PlenipotentiaryConference in 2010.

10 Entry into Force

The amendments to the Constitution and the Conven-tion contained in the present instrument shall, as awhole and in the form of one single instrument, enterinto force on 11 January 2008 between Member

States being at that time parties to the Constitutionand the Convention of the International Telecommu-nication Union (Geneva, 1992), and having depositedbefore that date their instrument of ratification,acceptance or approval of, or accession to the presentamending instrument.

Acronyms and Abbreviations

APT - Asia Pacific TelecommunityBDT - Telecommunication Development

BureauBR - Radiocommunication BureauCcTLD - Country code Top Level DomainCEPT - Conférence européenne des Administra-

tions des Postes et TélécommunicationsCHF - Swiss FrancsCITEL - Inter-American Telecommunication

CommissionCS - (ITU) ConstitutionCV - (ITU) ConventionCWG - Council Working GroupEC - European CommissionENUM - Mapping parts or all of the ITU-T

Recommendation E.164 internationalpublic telecommunication numberingplan into the Internet Domain NameSystem (“DNS”)

ETNO - European Telecommunication NetworkOperators’ Association

GNi - Global Network Planning InitiativesGSB - Global Symposium for RegulatorsICT - Information and Communication

TechnologyIDNs - Internet Domain NamesIGF - Internet Governance ForumITRs - International Telecommunication

RegulationsITU - International Telecommunication UnionITU-D - Telecommunication Development SectorITU-R - Radiocommunication SectorITU-T - Telecommunication Standardization

SectorJIO - UN’s Joint Inspection UnitMS - Member StatesNGN - Next Generation NetworkNGN-GSI - Next Generation Network Global

Standards InitiativeNPT - Norwegian Post and Telecommunica-

tions AuthorityNITU - Nordic ITU CooperationPP - Plenipotentiary ConferenceRA - Radiocommunication AssemblyRAG - Radiocommunication Advisory GroupRR - Radio RegulationsRRB - Radio Regulations BoardSG - Study Group



SM - Sector MembersTDAG - Telecommunication Development

Advisory GroupTSAG - Telecommunication Standardization

Advisory GroupUN - United NationsUNESCO - UN’s Educational, scientific and Cultural

OrganisationWATTC - World Administrative Telegraph and

Telephone ConferenceWCIT - World Conference on International

Telecommunications

WG Plen - Working Group of the PlenaryWIPO - World Intellectual Property OrganisationWRC - World Radiocommunication ConferenceWSIS - World Summit on the Information

SocietyWTDC - World Telecommunication Development

ConferenceWTPF - World Telecommunication Policy ForumWTSA - World Telecommunication

Standardization Assembly

Anne Lise Lillebø is Director, Telenor ASA, Group Regulatory. Her main responsibilities include spectrum

management and policy matters related to international telecommuncations organisations. She has been a

member of the Norwegian delegation to IUT’s Plenipotentiary Conferences in 1989, 1992, 1994, 1998,

2002 and 2006. She holds a Master of Arts degree from the University of Oslo.



Documents

Personal Networks - Telenor