Human Factors in Developing Trustworthy Service-Based Systems Stephen S. Yau Information Assurance Center, and School of Computing, Informatics, and Decision

Human Factors in Developing Trustworthy Service-Based Systems

Stephen S. YauInformation Assurance Center, and

School of Computing, Informatics, and Decision Systems Eng. Arizona State University

Tempe, Arizona USA

[email protected]

Evolution of Computing Paradigms

Main Frame Computer

Personal Computer

Object Oriented

Computing

Grid Computing

Service Computing

Cloud Computing

1950s Early 1970s

Early 1980s

Mid 1990s

2001 2005

Internet and LAN(Distributed Computing)

Virtualization Technology

Wireless Technology

2SERE 2012, S. S. Yau

Outline

Trustworthy Service-Based Systems (SBS)

Challenges of Developing Trustworthy SBS

Human Factors in Developing Trustworthy SBS

Current State of Art An Example Future Research


Service-Based Systems (SBS) Based on service-oriented architecture (SOA) Adopted in various application domains

E-business Health care information systems Homeland security Various collaborations …

Services in SBS provide standard interfaces for accessing capabilities offered by various providers

Services in SBS compose to form workflows (business processes) to provide functionality not provided by any individual service


Service-Based Systems (SBS) (cont.)

Many advantages of SBS, among them Interoperation of heterogeneous

systems Rapid composition of applications

(workflows) from distributed services Adaptation to dynamic application

requirements of users or environments (functional and QoS)


Trustworthy SBS Trustworthy SBS are needed for various

critical applications due to Over public or private networks, as well as mobile

networks – more open to attacks Interactions involving unknown entities Dynamic and pervasive environments Large-scale and cross-domain service

collaborations Distributed control Dynamic QoS expectations for multiple workflows


Trustworthy SBS (cont.)

Major aspects Human

Users and collaborators Service and infrastructure providers Insiders and outsiders

Devices, software, and system architectures Dynamic security policies and enforcement Dynamic user requirements and environments Effective techniques Cost, usability and efficiency


Trustworthy SBS (cont.)

Various system technologies are needed for developing trustworthy SBS Security Trust management Situation awareness Runtime adaptation

QoS monitoring and analysis QoS requirement trade-off Resource allocation

SERceE 2012, S. S. Yau 8

Interactions among services may have unforeseen consequences in trust, security, QoS, and risk Possible problems:

Untrusted/malicious services Intermediate results generated during service

interactions may reveal sensitive information Trustworthiness of service providers,

infrastructure providers and users

9

Challenges of Developing Trustworthy SBS

SERE 2012, S. S. Yau

Challenges of Developing Trustworthy SBS (cont.)

SBS has multiple QoS requirements from multiple users for various applications

Runtime tradeoffs among expected QoS

requirements Example: Mechanisms providing security

protection are often computationally intensive and

require certain sacrifice in other QoS (e.g. service

delay and throughput) with available resources

Cost, usability and efficiency10SERE 2012, S. S. Yau


Environments of SBS often dynamically change Make assessing trust and risk difficult Need situation awareness due to dynamic trust and risk Need adaptive enforcement of security policies

Information needed for making decisions regarding trustworthiness usually distributed on multiple services and organizations. Need cooperative decision making (e.g. delegation, policy

composition with multiple organizations, collaborative QoS management, risk assessment, trust evaluation)

Need efficient enforcement of distributed security policies Need protection against various entities



Service selection and composition How to select most appropriate services and

compose them to satisfy both functional and QoS requirements of various users, while ensure overall system trustworthiness and security?

Need meaningful and quantitative metrics for trustworthiness, security and various attributes of overall SBS

How to rank service ranking to identify the “best” services satisfying their requirements


What are Human Factors? In general, a human factor is a physical,

psychological or cognitive property of an individual or an individual in a community, which is specific to humans and influences on technological systems as well as their applications.

Examples: Influences, interests, relationships (collaboration/competition), opinions (positive/negative/neutral, support/against), knowledge (expertise), reputation, wisdom, physical and psychological factors (stress, fatigue, fear, happy).


Why Should Human Factors Be Considered in Developing TSBS?

Cyber systems become more powerful, and their applications become more diverse and pervasive

Human factors are increasingly influential on the quality and efficiency of generating the results because SBS getting more embedded Applications increasingly involving multi-party

collaborations and often more pervasive Applications must address multiple quality aspects

expected by users, such as security, privacy, trustworthiness and performance

14y SERE 2012, S. S. Yau

Three Levels of Human Factors Level 1. Direct Human-and-Human Relations

Collaboration of among users of cyber systems Level 2. Indirect Human-and-Human Relations

Service-users choose the services from service providers through service directories

First-time collaborations among the users based on past data

Level 3. Human in Communities Influence among users of cyber systems Knowledge sharing among the users of cyber systems Matching service-users’ interest with services


Direct Human-and-Human Relations

Challenges: How to quantify human factors in terms of

the determinants, such as workload on the human, fatigue, learnability, attention, vigilance, human relations, human performance, human reliability, stress, individual differences, aging, safety, and results of decision making.

How human factors affect humans themselves?


Indirect Human and Human Relations

Example: In a SBS, the providers upload their

services/applications. The users search the service/application directory for the SBS and select the services/applications they need. Besides the quality of the services/applications, each user is concerned with the trustworthiness of the services.

Challenge: How can a user choose a trustworthy service?

Related human factors: human relationships, stress, feedback, etc


Human in Communities

Challenges: How do the human factors from one person affect

other persons in the community? How do the human factors from other persons in a

community affect one person in the community? How do the human factors from one person in a

community spread in the SBS used by the community?

…


Human in Communities (cont.)

Example: In a SBS, it is not easy for users to find their match

services/applications. It is also difficult for the service providers to introduce their services/applications to possible interested users using the SBS.

Challenge: How to incorporate human factors to match all service-users with all services/applications automatically and efficiently?

Related human factors: influence, interests, opinions, human relationships …


Current State of Art

Incorporating human factors in developing cyber systems and applications Research has been mainly conducted by

researchers in psychology and sociology, and few computer scientists and engineers.

Primarily focus on human-machine interactions, human-computer interactions, situation awareness, and human errors


Current State of Art (cont.)

Automated service composition based on various formal specifications

Process calculi; BPEL4WS QoS-aware service composition in SBS

Optimizing QoS attributes of services using the genetic algorithm during the service compositions (G. Canfora, et al., University of Sannio, Italy).

QoS provisioning for composed services, based on the Service Level Agreement (SLA) contracts of individual services (X. Gu, et al., University of Illinois, Urbana)

Developing QoS-aware middleware for web service composition to maximize users’ satisfaction expressed in utility functions over QoS attributes. (L. Zeng, et al., IBM)


Tradeoffs among security and multiple QoS in SBS Development of a framework for quantifying the strength of

system security (M. Satyanarayanan, et al., Carnegie Mellon University)

An adaptive model for tradeoff between service performance and security in service-based environments (S. Yau, et al., Arizona State University)

A comprehensive QoS model for service-based systems, (I. Jureta, et al., University of Namur, Belgium)



Adaptive resource allocation in SBS A multi-layered resource management framework for

dynamic resource management in enterprise systems. (P. Lardieri, et al., Lockheed Martin Corporation)

Decentralized online resource allocation for dynamic web service applications (J. Stoesser, et al., Universitat Karlsruhe, Germany).

A regression based analytical model for dynamic resource provisioning of multi-tier applications (Q. Zhang, et al, HP Labs)

Adaptive resource allocation for SBS (S. Yau, et al., Arizona State University)



Design of SBS for QoS Monitoring and adaptation A development methodology for adaptive service-based

software systems (S. Yau, et al, Arizona State University) Comprehensive QoS monitoring of Web services and event-

based SLA violation detection (Michlmayr, et al, Vienna University of Technology, Austria)

Testing of SBS Dynamic Reconfigurable Testing of Service-Oriented

Architecture (W. Tsai, et al, Arizona State University)



Trust estimation in SBS Flexible trust model for distributed service

infrastructure (Z. Liu, University of North Carolina at Charlotte, S. Yau, Arizona State University)

Trusted computing platforms in web services (Nagarajan, et al, Macquarie University, Australia)

Trust management for context-aware service platforms (Neisse, et al, University of Twente, the Netherlands)

Improving trust estimation in SBS (S. Yau and P. Sun, Arizona State University)


An Example: Improving Trust Estimation in SBS

Improving trust estimation in service-based systems by incorporating human factors as well as QoS profiles


Trust Estimation in SBS Trust management needs to be incorporated in

SBSs to estimate service providers’ trustworthiness so that users can decide whether to accept the services provided by the providers.

Limitations of existing trust estimation approaches: Only similarity of user profiles is considered Based on pairwise trust relationship, which normally

does not include the transitive property in the propagation of trust among service providers.


An Example: Improving Trust Estimation in SBS (cont.)

Incorporating two common human factors: competition and collaboration in interpersonal relationships, as well as QoS profiles, which have significant effects on trust estimation in SBS.


Network Model to be Used in Our Approach

V: a set of vertices (service providers)

E: two sets of directed edges (intention of one vertex to another) Competing edges (dashed lines) Collaborating edges (solid lines)

Wedge: the weighting factor of an edge, which indicates the weight of competition or collaboration. In the range of [0, 1]


Definition of a Transaction

A transaction in an SBS consists of three phases: Request phase: a user of a SBS requires services. Selection phase: the requester chooses services from

the service providers of the SBS with trust values exceeding an acceptable threshold specified by the user.

Feedback phase: the requester gives feedback to the SBS on the quality of the services used.


Definition of a Transaction (cont.)

Information recorded in a transaction: Which service providers selected by a service user QoS profiles of all the selected services.

Record all aspects of quality of each service required by user

Two types: Claimed QoS profiles: provided by the service providers

along with their services when publishing the services. Feedback QoS profiles: provided by the service user in the

feedback phase after using the services.


Definition of a QoS Profile A QoS profile of a service in a transaction is a vector

with n elements, each of which represents a quantifiable aspect of the service, important for trust estimation. Consider three aspects: Accuracy, delay and throughput

Claimed QoS profile: Provided by service providers Denoted by cProfile = <cQ1, cQ2, …, cQn>

Feedback QoS profile: Provided by service users Denoted by fProfile = <fQ1, fQ2, …, fQn>


Improving Trust Estimation in SBS

Initialization Initialize the trust values of all service providers of

the SBS based on historic transactions using QoS profiles, collaboration and competition.

Utilization Update the trust values of the service providers in

current transaction using QoS profile. Update the trust values of all the other service

providers using competition and collaboration.


Trust Values (Initialization

)

Estimated Trust Values

Log(s)

Transactions (Utilization)

Trust System Adaptor

Original Trust Values

Trust Refinement Module


Profile Evaluation

Module

QoS Profiles

Log Transactions (Initialization)

Existing Trust Estimation

Trust Dissemination

1

2

1

3

4

Service-Based System


Improving Trust Estimation in SBS (cont.)

Profile evaluation module processes the transactions to extract the QoS profiles.

Trust refinement module calculates the trust values of service providers based on QoS profiles, competition and collaboration.



Log(s)

Transactions (Utilization)

Trust System Adaptor

Original Trust Values

Trust Refinement Module


Profile Evaluation

Module

QoS Profiles

Log Transactions (Initialization)

Existing Trust Estimation

Trust Dissemination

1

2

1

3

4

Service-Based System

56

Trust Values (Initialization

)

7


Improving Trust Estimation in SBS (cont.)

The estimated trust values are dis-seminated to service providers in SBS


Effect of QoS Profile on Trust Estimation

If the feedback QoS profiles of a selected service is better than its corresponding claimed QoS profiles, then the service user can decide the service provider is more trustworthy, and consequently increase the estimated trust value of the service provider.

Otherwise, decrease the estimated trust value of the service provider.


For a given service, ith signed normalized aspect difference (snadi) is the signed normalized difference between the ith aspects of its feedback QoS profile and its claimed QoS profile. Sign of snadi is decided by the system administrator, e.g.

cQi is better than fQi, positive cQi is worse than fQi, negative

snadi is given by

Comparison of Claimed and Feedback QoS Profiles


where di_max and di_min are the maximal and minimal differences of the ith aspect among all transactions

Compare cProfile and fProfile M is the overall score of the comparison.

where w1, …, wn are user-assigned weights for various aspects in QoS profile

38

Comparison of Claimed and Feedback QoS Profiles (cont.)


Improvement of Trust Estimation Using QoS Profiles

If M is positive, then the feedback QoS profile is better than the claimed QoS profile; otherwise, the feedback QoS profile is worse than the claimed QoS profile.

Improvement of the trust estimation using QoS profiles.

T’(u) is improved estimated trust value of service provider u T(u) is original estimated trust value of service provider u γ is the parameter for adjusting the effect of M

By default, γ = 0.85.


Rule 1. Competition relationship increases the trust values of the participants in the competition group.

Competition limits free-ride The more time one spends, the more one is likely to trust the

people in this group. Rule 2. Collaboration relationship increases trust.

When two persons collaborate with each other, they tend to solve problems together and this process can help to build trust between them.

Rule 3. Transitive property of trust. Whenever one service provider’s trust value changes, the trust

values of his/her neighbors will also change accordingly. The trust value of a service provider is uniformly propagated to all

the other service providers he intends to compete or collaborate with.

40

Improvement of Trust Estimation Using QoS Profiles (cont.)


Improvement of Trust Estimation Using QoS Profiles (cont.)

Rules 1 and 2 show the positive correlation between trust and competition or collaboration.

Rule 3 defines how the trust values should be propagated among the whole network of SBS. The propagation of the trust values of service

providers is similar to PageRank (a webpage reputation estimation approach).

The more people who intend to compete or collaborate with a service provider, the more trustful the service provider is.


Expertise Needed to Incorporate Human Factors in Developing TSBS

Services and cloud computing Software and systems engineering Networking, including mobile ad hoc networks,

intelligent devices, and social networks Information assurance and security Cognitive science Psychology Business Culture …

SERE 2012, S. S. Yau 42

Future Research to Incorporate Human Factors in Developing TSBS

Develop meaningful metrics to quantify human factors and QoS aspects of SBS, including trust, security and others useful for developing TSBS

Develop a general framework with necessary techniques and tools to effectively incorporate a variety of relevant human factors in developing TSBS

Validation


Thank you


Documents

Human Factors in Developing Trustworthy Service-Based Systems Stephen S. Yau Information Assurance Center, and School of Computing, Informatics, and Decision