28th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2011)

April 26-28, 2011, National Telecommunication Institute, Egypt

A Novel Trust-Based Cross-Layer Model for Wireless Sensor Networks

ABSTRACT

Hosam A. Rahhali, Ihab A. AlP, Samir 1. Shaheen3 lFaculty of Engineering, Cairo University, Giza, Egypt, hrahhal@ieee.org

2Faculty of Engineering, Helwan University, Helwan, Egypt, ihabaly@ieee.org 3Faculty of Engineering, Cairo University, Giza, Egypt, sshaheen@eng.cu.edu.eg

Wireless Sensor Networks (WSNs) are vulnerable to attacks (selfish or malicious i.e. misbehaving nodes) due to the nature of the wireless media , restricted resource and the natural co-operations of sensors. Therefore, the security issue is very critical in WSN. The decision making in a WSN is essential for carrying out certain tasks as it aids sensors establish collaborations. In order to assist this process, trust management models could play a relevant role. Up to our knowledge, there is no one used the cross-layer concept in computing and updating the trust values. So, this paper presents new model for trust in WSN, called A Trust-Based Cross-Layer Model, which use cross-layer concept (ACKs from data link layer and TCP layer) to design trust-based model for sensor networks that guarantee the trust route from source to sink and isolate the malicious node. The simulation results and analysis show that our model is scalable and its display high performance even if the percent of malicious nodes is high.

Keywords: wireless sensor networks, trust, malicious, cross-layer.

I. INTRODUCTION

Recently, the rapid advances in processor, memory, and radio technology have enabled the development of distributed networks of small, inexpensive nodes that are capable of sensing, computing, and wireless communication. Wireless sensor networks (WSNs) are collections of a large amount of small devices equipped with integrated sensing and wireless communication capabilities. They are considered as a special case of ad hoc networks with reduced or no mobility [1]. Sensor networks share some important characteristics with ad hoc networks, for example they share the need for self-organization, wireless multi-hop operation, and time-variability in topology [2]. The sensor networks are expected to find widespread use in a variety of applications. Examples of such applications include environmental monitoring - which involves monitoring air, soil, vegetation and water, condition based maintenance, habitat monitoring (determining the plant and animal species population and behavior), seismic detection, military surveillance, inventory tracking, smart spaces etc. [3,4]. However, as wireless sensor networks (WSNs) grow to be more popular and widely used, security becomes a very serious concern. Users do not want to reveal their data to unauthorized people as the disclosed information could be used for malicious purposes. This concern is even more relevant to wireless environments where anyone can overhear a message sent over the radio. Therefore, even a very useful and convenient system might not be appealing to the users if it is not secure. Therefore, a smart trust management scheme is needed to identify trustworthiness of sensor nodes in order to distinguish between malicious nodes and good nodes, and to strengthen reliable nodes and weaken suspicious nodes. Trust is currently a hot issue in computer networks which can solve security problems caused by malicious members, so building trust relationships among sensor nodes has been recognized as a novel approach to improve security in WSNs. In this paper, we propose new trust model for Wireless Sensor Networks based on cross-layer idea called A Trust-Based Cross-Layer Model (TCLM) for Wireless Sensor Networks. In our model we use the direct and indirect observation of the nodes to compute the trust values and the acknowledgements (ACKs) from data link layer and TCP layer are considered to update these values. The rest of the paper is organized as follows: related work is summarized in section 2. The description of our proposed model is given in section 3. Simulation results are shown in section 4, followed by conclusions in section 5.

II. RELATED WORK

The operation of Mobile Ad Hoc networks (MANET) and Sensor Networks depends on the cooperation of participating nodes. Since nodes have limited resources, in particular battery power, cooperation may come at a significant expense. In recent time the issue of security in wireless Sensor networks has been addressed in several

28th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2011)

April 26-28, 2011, National Telecommunication Institute, Egypt

works. The reputation and trust systems have been proved useful mechanism to address the threat of compromised or faulted entities. In this section, we examine the literature relevant to this research. The authors in [5] have proposed CORE (COllaborative REputation) to enforce node cooperation in Mobile Ad hoc Networks, this approach is formed and updated along time through direct observations and through information provided by other members of the community, the nodes exchange only positive reputation information. CONFIDANT(Cooperation Of Nodes- Fairness In Dynamic Ad-hoc NeTworks) [6] is a distributed, symmetric reputation model based on direct observation and on second-hand information from other nodes and is updated according to a Bayesian estimation. S. Ganeriwal et al. have proposed a RFSN(Reputation-based Framework for Sensor Networks) [7] which is the fIrst reputation and trust-based model designed and developed exclusively for sensor networks, it uses watchdog mechanism to build trust rating. But the watchdog cannot record all the behavior due to its own fault or network error, so there is some uncertainty events in the trust system. M. Pushpalatha et al. [8] have proposed a trust- based-energy- aware routing model in MANET. During route discovery, node with more trust and maximum energy capacity is selected as a router based on a parameter called 'Reliability'. Route request from the source is accepted by a node only if its reliability is high. Otherwise, the route request is discarded. Bayesian approach based model [9] enables recovery and prevents the sudden exploitation of good reputation built over time by introducing re-evaluation and reputation fading. In [lO],the authors have proposed a Hermes scheme which is conceptual framework for trust establishment with respect to reliable packet delivery in the presence of potentially malicious nodes. This scheme use the fIrst-hand and second -hand information to update the trust values. The authors in [11], have proposed a distributed trust-based framework and a mechanism for the election of trustworthy cluster heads. Each node stores a trust table for all surrounding nodes and these values are reported to the cluster head only and upon request. Xiao et al. [12], developed a mechanism called SensorRank for rating sensors in terms of correlation by exploring Markov Chains in the network. A network voting (TrustVoting) was proposed to determine faulty sensor readings, if the sensor is faulty then the node will not participate in the voting. H. Chen et al. [13], have proposed a reputation-based trust which borrows tools from probability, statistics and mathematics analysis and they suggested a new term certainty used in trust system and they argued that the positive or negative outcomes for a certain event is not enough information to make a decision in WSNs. The authors build up a reputation space and trust space in WSNs, and defme a transformation from reputation space to trust space. In [14], the authors have proposed a trust-based LEACH protocol to provide secure routing, which is an integration of a trust management module with a trust-based routing module. The trust management module is responsible for building trust relationships among sensor nodes with novel methodologies to provide efficient monitoring, trust exchange and evaluation. The trust-based routing module is a modified version of original protocol with the same head-election algorithm and working phases, while having trust-based decision-making. D. Hui-hui et al. [15], have proposed a multi-angle trust mechanism for nodes in Wireless Sensor Networks which adding the sensing data and the node's energy in the factors of trust assessment in addition to communication , and new trust models to calculate the trust values of communication trust, the sensed data and the node's energy. In [16], the authors have proposed a new protocol TTSN (Task-based Trust Framework in Sensor Networks) to construct a trust framework model in wireless sensor networks. The sensor node has different trust rating for different task. The proposed model use watchdog scheme to observe the behavior in different events of these nodes and broadcast their trust ratings. The authors in [17], have proposed an approach called BTRM-WSN which is a bio-inspired trust and reputation model for WSNs aimed to achieve to most trustworthy path leading to the most reputable node in a WSN offering a certain service. It is based on the bio-inspired algorithm of ant colony system. Most of existing methods are designed for MANETs and suppose that each node have the full route from source to destination (i.e. designed for source routing ), while for the hop by hop routing they discuss only one­hop neighbours not along the full route, that is not suitable for WSNs. In addition to that, no one used the cross­layer concept in computing and updating the trust values. In our work we are going to make use of data available at both DLL and Transport layer to make the network-layer decision more accurate, which improve the performance of the trust model. These reasons are the motivation behind this research.

III. THE PROPOSED MODEL

In this section we describe the problem, assumption, and our proposed model.

A. PROBLEM DEFINITION

Given a sink(s) and a set of static homogeneous sensors S that are deployed randomly in a two­dimensional field. Sensors are homogeneous in terms of their initial energy, sensing range, and communication range. some nodes are sources of transmission and some nodes act as relay nodes. the problem can be summarized by these questions: which nodes can we trust in the network?How can we use cross-layer concept to specify an

28th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2011)

April 26-28, 2011, National Telecommunication Institute, Egypt

accurate trust- threshold to distinguish between legitimate nodes and malicious nodes? How can we detect and isolate the malicious nodes in wireless sensor network? How do we assure that a node is transferring the data packets correctly? Then, malicious-free route between the source and destination could be selected.

B. ASSUMPTIONS Assumption that may lead to solve the problem:

1) the sensing area of each sensor s E S is a circle with radius (rs )centered at the location of the sensor. 2) all the sensors have similar communication range (cs).

3) some nodes are malicious and attempt to obstruct network performance. 4) a malicious node will always participate in route setup operation. 5) all nodes in sensor network cooperate in order to create the routes and forward packets.

6) all nodes in sensor network operate in promiscuous mode which allows a node to intercept and read each network packet that arrives it rather than just packets addressed to it. While When a data packet is transmitted in non-promiscuous mode, all the neighbor nodes "listen to" the data to determine if the network address included in the data packet is theirs. If it isn't, the data packet is passed onto the next node until the node with the correct network address is reached. That node then receives and reads the data.

C. THE PROPOSED MODEL (TCLM)

We adapt the model in [18], to use it in our work as shown in figure(I).

Rptwppn npjphhnr nnrlp� i R, i

----------------�----------------r �

Watching data Trust (t) Worthiness (W) Sending & receiving between Treatment Ratio (r) Mapping t & r �w(t,r)=f(t,r)

neighbors

t E [0,1] W E[O,l]

Routing decision Average V Viewpoint (V)

V=f (W)

\... ,/ V

Between any pair of nodes i & k

Multi- trust Level

+--Fig. 1: TCLM model.

Now, we explain the function of each box in this model:

-Watching data (like watchdog algorithm): observant neighbors nodes in transmission/reception range, Due to the broadcast nature of the wireless medium, a given node can collect information about the packet forwarding behavior of its neighbor nodes by snooping all received frames at the MAC layer and recording packet delivery statistics.

-The packet statistics can be used to compute a pair of values associated with each neighbor node, which we call trust (t) and Treatment Ratio(r). We define trust as a value in the range [0, 1], which indicates the degree to which a given node believes that its neighbor node behaves normally or can be "trusted" to forward the packets it receives for forwarding. Let L denote the cumulative number of packets forwarded correctly and let N denote the cumulative number of packets sent for forwarding by the node up to the current time. Then the trust value, t, signed to a node is defined as follows based on Bayesian statistics[I9] and Beta distribution[20]:

28th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2011)

April 26-28, 2011, National Telecommunication Institute, Egypt

t=L/N (1)

-We define the Treatment Ratio(r), associated with a given trust value t, as a value in the range [0,1] which indicates a measure of statistical dispersion in the trust value. The pair of values (t, r) characterizes the degree of belief that the neighbor is reliable with respect to packet delivery and the statistical confidence in this belief. The Treatment Ratio, value, r, associated with the trust value t is defined as follows based on Bayesian statistics and Beta distribution[IO]:

r = 1 _ -,-�1_2L---'.( N------:

- L,..-)

(N+1)N 2 (2)

-The Worthiness (W) associated with a pair(t, c) is mapping the trust value, t, and Treatment Ratio value, r, into a

Single value to facilitate trust-based decision and is defined as [10]:

(3)

Where ,c, is a parameter that determines the relative importance of the trust value t vs. the Treatment Ratio value r. The "default" value of worthiness is defined as:

Wdej =W(O.5, O) (4)

The value W def can be interpreted as an initial threshold for worthiness. If the worthiness of a node exceeds

Wdef then the node is considered trustworthy Otherwise, the node is viewed as untrustworthy.

- We generalize the notion of worthiness to the concept of Viewpoint (V) , which incorporates worthiness Values information from non-neighbors nodes, and defined as:

W'k 1, if i and k are nieghbors,

V'k 1, max: PEPik

if P. k *cf>, 1,

Wdej othenvise,

Where , Pi k ' denote to the Set of paths from node i to node k. ,

- Average V (v) , presents the average of V over Watching windows.

(5)

Now, we can apply viewpoint metric to execute the routing decision or to create Multi-Trust level method depend on the sensitive of executing task in order to balance the traffic and the power in the sensor networks. To compute the t, r, W, V values: we define two counters, L-counter , N-counter ,and two timers, the first one,

tgi1 , The timeout Interval of this timer is set to a value greater than the maximum round-trip time (RTT)between

two neighbour nodes in the network. The second timer, t¥c�, The timeout Interval of this timer is set to a value depend on the network diameter. the L-counter and N-counter are updated as follow: when any node (i) send packet (p), to node G) , the N-counter(i) is incremented by one and the timers are initiated. If the acknowledgements from data link layer (ACKDLL ), and TCP layer, ( ACKTCP ) , are received by node (i) before the timers expires, the L-counter(i) will incremented by one, else , the L-counter(i) not update. Also the two counters for all nodes along the full route will updated in same way. The flow chart of the proposed method is shown in Figure (2).

28th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2011)

April 26-28, 2011, National Telecommunication Institute, Egypt

start

N y

Compute penalty value

Compute credit value

The link exclude

No action

N

N

N

IV. SIMULATION RESULTS

Fig. 2: Flow Chart ofTCLM.

Compute overall trust value

We test our model against a number of random WSNs, each of them has a random topology, so it could happen that our model was tested against networks where benevolent sensors were very near to the clients (maybe one hop forward) or quite the contrary, that is, WSNs where benevolent sensors were quite far from the clients. In this section, we perform two scenarios to evaluate the performance of the TCLM scheme using simulation experiments. In each scenario we calculate few performance metrics:

1) Accuracy : how many times the proposed model is able to select the right benevolent node to interact with. In other words, we would like to know the selection percentage of trustworthy nodes or Accuracy.

2) standard deviation(SD) of the selection percentage of trustworthy nodes (accuracy).

3) Average Path Length leading to trustworthy destinations. 4) Average Energy Consumption : Energy consumption is a critical issue when dealing with wireless

sensor networks, since these ones are commonly composed by resource-constrained devices with limited features in terms of processing, memory and communicating capabilities. Therefore, we measure the average energy consumption needed by the proposed approach. As pointed out by [21,

28th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2011)

April 26-28, 2011, National Telecommunication Institute, Egypt

22], In the most commonly used model, the energy consumption for transmitting a message at distance d is:

E(d) = da +C (6)

where a E [2, 4] represents the media attenuation factor and C is a constant denoting the power used to process the radio Signal. Following these authors' direction, we have chosen a value of a = 4 and C =105. Additionally,

2 the sensors belonging to our generated networks are spread along a (100 * 100) m area, and each of them has a radio range of 10 meters.

A. Scenario 1

One of the advantages of our trust model is its scalability, the size of sensor networks can vary from a handful of nodes until thousands of them, developing a scalable model is a critical issue. In this scenario we study the impact of density on our model, i.e. scalability. For that we run the simulator under the same conditions, i.e. the number of networks( equal to 25) , the number of executions over each network (equal to 25) i.e. each client applied for a service 25 times, , the percent of client(always equal to 15%) , The 85% left were, therefore, sensors acting as servers, relay(equal to 25% of servers) and malicious(equal to 40% of servers ) nodes , the rest servers are benevolent sensors and radio range(equal to 10 m), but we vary the number of nodes (15 to 1000) in each network. Figure 3 shows the variation in the accuracy of selection trustworthy nodes vs. the number of sensors in the network. As has been shown, the accuracy still above 98% till when the size of the network in 1000 sensor.

100

>-. 99.5 g .... ::l 99 () ()

<r: ;:f( 0 98.5

98

0 200 400 600 800 1000 1200

sensors number

Fig. 3:Accuracy vs. density.

When we say the accuracy is 80% , for instance, could be reached because the model always found a trustworthy server the 80% of the times, or just because it found it the 100% of the times in half the tested wireless sensor networks and the 60%, in the other half, for example. That is the reason why we decided to measure and show the standard deviation related to the accuracy. Figure 4 shows the variation in the standard deviation of selection trustworthy nodes vs. the number of sensors in the network. As has been shown, the standard deviation remains quite low and less than 2.5 % till if the number of sensors is 1000 sensor.

2.5 2

Cl 1.5 r:/J ;:f( 1

/ 0

0.5 0

0 200 400 600 800 1000

sensors number

Fig. 4:%SD vs. density.

28th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2011)

April 26-28, 2011, National Telecommunication Institute, Egypt

The proposed model is aimed to find the closest benevolent sensors to the client requesting the service. Due to the restrictions related to wireless sensor networks, the resources consumption saving is a critical issue. Therefore, a

shorter path leading to the final trustworthy sensors implies less involved sensors, consequently, less global utilization of resources such as energy or bandwidth. The variation in the Average Path length vs. the number of sensors in the network is presented in Figure 5.

2.35 ..c:

� til 2.3 s:: .., ...l 2.25 ..c: c;; 2.2 0-.., 2.15 bJ) � .., 2.1 ;>

« o 200 400 600 800 1000

sensors number

Fig. 5: Average Path Length vs. density.

As it is seen from the figure5, the average path length decrease as the number of the nodes in the network increases because the number of neighbors for each sensor will increase , consequently, the probability a sensor which has the required service is closet to client will be increased. As it can be observed, any trustworthy server is never reached (on average terms) at more than 2.35 hops. Finally, the last experiment consisted of measuring the average energy consumption in the networks vs. the density, and it can be observed in Table!.

T bl 1 A a e : verage E nergy onsumpllOn c f Sensors Number Average Energy Consumption

15 8.5*10All 25 3.8*10A12 50 3.3*10A14

100 2.7*10A16 200 6.6*10A18 300 7.8*10A19 400 7.1*10A20 500 S.2*10A21

1000 9.4*10A23

As has been shown in table 1, the bigger the number of sensors is, the higher is the energy consumption. The increase of energy needed as the size of the network grows is explained because it implies an increase in the density of the network too. So the average number of links increases rapidly and, therefore, a higher number of messages are sent. From the result of the first scenario we can state that the proposed model is scalable.

B. Scenario 2

In this scenario we study the impact of malicious node on our model, i.e. we evaluate the proposed model when the percent of the malicious nodes increased. For that we run the simulator under the same conditions, i.e. the number of networks(equal to 50) , the number of executions over each network (equal to 50) i.e. each client applied for a service 50 times, the percent of client(always equal to 15%) , The 85% left were, therefore, sensors acting as servers, relay(equal to 9% of servers), and we vary the percent of malicious nodes (10% to 90%).nodes , the rest servers are benevolent sensors and radio range(equal to 10 m), and radio range(equal to 10 m). we repeated those experiments over WSNs composed of 50, 100, 300 and 500 sensors (with the same percentages of clients, servers and malicious servers). Figure 6 shows the variation in the accuracy of selection trustworthy nodes vs. the percent of malicious sensors in the network.

28th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2011)

April 26-28, 2011, National Telecommunication Institute, Egypt

100 >. 80 <.) � 60 <.) �40 � " 20

a

a 20 40 60 % malicious nodes

80 100

Fig. 6:Accuracy vs. % malicious sensors.

-+-network size=50

_ network size= I 00

--'-network size=300

�network size=500

As has been shown, the proposed model achieved accuracy value above 97% when the percentage of malicious sensors is less than or equal to 80%,in every case. The worst case when the percentage of malicious sensors is 90 and the number of sensor in the network is 500 because the most sensors in the network are malicious. The variation in the standard deviation of selection trustworthy nodes vs. the percent of malicious sensors in the network is presented in Figure 7. As has been shown, the standard deviation remains quite low and less than or equal 5 % , when the percentage of malicious sensors is less than or equal to 80%.

15 c -+-network size=50 0 .� _ network size=100 '�1O 0 --.- network size=300 "0 .... «l -g 5 «l

� network size=500

en � "

0

0 20 40 60 80 100 % malicious nodes

Fig. 7:%SD vs. % malicious sensors.

Figure 8 shows the variation in the average path length vs. the percent of malicious sensors in the network.

8 ..c to c 6 d) --l ..c «i 4 0.. d) OIl «l 2 .... d) ;> --<

a

a 20 40 60 % malicious nodes

80 100

-+-network size=50

_ network size= I 00

--'-network size =300

� network size=500

Fig. 8: Average Path Length vs. % malicious sensors.

As it is seen from the figure8, the average path length increase as the percent of malicious increase because the number of malicious sensors in the network will been increased ,as a result the route to server or destination who has the service will be longer. Consequently the average path length will increase. As it can be observed, any trustworthy server is never reached (on average terms) at more than 4.21 hops, when the percentage of malicious sensors is less than or equal to 80%. While when percent of malicious be 90% and the number of sensors in the network is 500, the average path length will be about 7 hops which consider the worst case. Finally in this

28th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2011)

April 26-28, 2011, National Telecommunication Institute, Egypt

scenario, the last experiment consisted of measuring the average energy consumption in the networks vs. the percent of malicious sensors in the network., and it can be observed in Table2. Two direct conclusions can be deducted from the table2. The first one, the bigger the number of sensors is, the higher is the energy consumption. And the other one, the greater the percentage of malicious sensors is, the higher is the power consumption as well. Regarding the percentage of malicious sensors, in a network where this kind of sensors are in majority it is more difficult to find a benevolent one and, consequently, more messages need to be sent as well, for that, more energy will consume.

a e : t e average energy consumptIOn III t e networ s T bl 2 h .

h k

Malicious% Network size=50 Network size=100 Network size=300 Network sensor sensor sensor size=500 sensor

10 1.8*101'14 1.4*101'16 1.4*101'19 3* 1 01'21 20 1.9*101'14 1.5*10"'16 1.6* 10"'19 3.2*101'21 30 2.1*101'14 1.9*101'16 1.7* 101'19 3.3*101'21 40 3.1*101'14 2.1*10"'16 1.9* 101'19 3.4*10"'21 50 3.4*101'14 2.5*10"'16 2.2*101'19 4* 101'21 60 4.6*101'14 3.2*10"'16 8.3*101'19 4.7*10"'21

70 5.1 *101'14 4.3*101'16 1*101'20 6.3*10"'21 80 6.4*101'14 8.6*101'16 3*101'20 2*101'22 90 9.9*10"'14 52*10/\16 2.2*10/\22 2.1 *10/\24

v. CONCLUSIONS

In this paper, we have proposed new trust scheme for static wireless sensor network based on cross-layer concept , called Trust-Based Cross-Layer Model. The description of this scheme is presented and evaluated using two scenario. In the first one we evaluate the performance of the proposed scheme when the network consist of big number of sensors and we have found that the proposed model is scalable. While in second scenario we evaluate our model when the number of malicious nodes increase in the network, and we have found that the proposed model work excellently even if the percent of malicious is high. As a future work we can test our model against dynamic Wireless Sensor Networks, where some nodes switched off or when some sensors are mobile. Also, we can compare our trust model with other models. we also suggest applying energy-aware technique by letting some nodes to operate in on/off and! or relay destination mode to improve the performance of the network and saving in the consumption energy.

REFERENCES

[1] C. Cordeiro, D. Agrawal," AD HOC & SENSOR NETWORKS: Theory and Applications", book Published by World Scientific, 2006.

[2] A. Willig, "Wireless sensor networks: concept, challenges and approaches", Elektrotechnik &Informationstechnik, 2006.

[3] A Bharathidasan, V. Ponduru, "Sensor Networks: An Overview", Technical report, University of California, Davis. 2002.

[4] A. Zerger, R.A. Viscarra Rossel, D.L. Swain, T. Wark, R.N. Handcock, V.AJ. Doerr, G.J. Bishop-Hurley, E.D. Doerr, P.G. Gibbons, C. Lobsey, "Environmental sensor networks for vegetation, animal and soil sciences", International Journal of Applied Earth Observation and Geoinformation, Volume 12, Issue 5, Pages: 303-316, October 2010.

[5] P. Michiardi and R. Molva. "CORE: A Collaborative REputation Mechanism to Enforce Node Cooperation in Mobile Ad Hoc Networks". IFIP Conference proceedings In Communication and Multimedia Security, Vol. 228,PP: 107 - 121, 2002.

[6] S. Buchegger and J.-Y. Le Boudec. "Performance Analysis of the CONFIDANT Protocol (Cooperation Of Nodes- Fairness In Dynamic Ad-hoc NeTworks)". In proceedings of ACM International Symposium on Mobile Ad Hoc Networking and Computing (MobiHoc), 2002.

[7] S. Ganeriwal and M. Srivastava. "Reputation-based Framework for High Integrity Sensor Networks". In Proceedings of ACM Workshop on Security of Ad Hoc and Sensor Networks (SASN), pages 66-77, 2004.

28th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2011)

April 26-28, 2011, National Telecommunication Institute, Egypt

[8] M. Pushpalatha, R. Venkataraman, T. Ramarao,"Trust Based Energy Aware Reliable Reactive Protocol in Mobile Ad Hoc Networks", in World Academy of Science, Engineering and Technology, Vol. 56, No.68, PP.356-359, 2009.

[9] Y. Wang and 1. Vassileva, "Bayesian network-based trust model", In the International conference on Web Intelligence (WI '03), Halifax, Canada, 2003.

[10] C. Zouridaki, B. L. Mark, M. Hejmo, R. K. Thomas, "A quantitive trust establishment framework for reliable data packet delivery in MANETs", In Proceedings of the 3rd ACM workshop on Security of ad hoc and sensor networks (SASN), vol. ,PP: 1 - 10 , 2005.

[11] Crosby, G.V. Pissinou, N. Gadze, J." A framework for trust-based cluster head election in wireless sensor networks", in the Second IEEE Workshop on Dependability and Security in Sensor Networks and Systems, Pages: 13 - 22, 2006.

[12] X. Xiao, W. Peng, C. Hung, and W. Lee, "Using SensorRanks for In-Network Detection of Faulty Readings in Wireless Sensor Networks."in MobiDE, 2007.

[13] Haiguang Chen, , Gangfeng Gu, , Huafeng Wu, , Chuanshan Gao,"Reputation and Trust Mathematical Approach for Wireless Sensor Networks ", in International Journal of Multimedia and Ubiquitous Engineering, Vol. 2, No. 4, October, 2007.

[14] F. Song, B. Zhao, "Trust-based LEACH Protocol for Wireless Sensor Networks", In Second International Conference on Future Generation Communication and Networking (FGCN), vol. 1, pp.202-207, 2008.

[15] D. Hui-hui, G. Ya-jun, Y. Zhong-qiang, C. Hao, "A Wireless Sensor Networks Based on Multi-angle Trust of Node", In 2009 International Forum on Information Technology and Applications (lFITA), vol. 1, pp.28-31, 2009.

[16] Haiguang Chen , "Task-based Trust Management for Wireless Sensor Networks ", in international Journal of Security and Its Applications, Vol. 3, No. 2, pp: 21-26, 2009.

[17] F. G. Marmol ," Providing trust in wireless sensor networks using a bio-inspired technique ", In Springer journal of Telecommunication Systems , DOl :10.1007/s11235-010-9281-7, 18 February 2010.

[18] C. Zouridaki, B. L. Mark, M. Hejmo, R. K. Thomas, "E-Hermes: A robust cooperative trust establishment scheme For mobile ad hoc networks", In Elsevier journal of ad hoc networks, Volume 7, Issue 6, Pages 1156-1168, August 2009.

[19] S. Buchegger and J.-Y.L. Boudec. A Robust Reputation System for P2P and Mobile Ad-hoc Networks. In Proc.2nd Workshop on Economics of Peer-to-PeerSystems,June,2004.

[20] W. M. Bolstad, "Introduction to Bayesian Statistics", Second Edition, book,ISBN 978-0-470-141 15-1, John Wiley & Sons,2007.

[21] L. Li, &, J. Y. Halpern ," Minimum-energy, mobile wireless networks revisited". In IEEE International Conference on ommunications,ICC 2001 (Vol. 1, pp. 278-283),2001.

[22] J.A. Sanchez, & P.M. Ruiz, "Improving delivery ratio and power efficiency in unicast geographic routing with a realistic physical layer for wireless sensor networks". In Proc. 9th Euromicro conference on digital system design (DSD'06) (pp. 591-597),2006.