security enhancements for mobile adhoc networks with trust management using uncertain reasoning

0018-9545 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. Seehttp://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI10.1109/TVT.2014.2313865, IEEE Transactions on Vehicular Technology

1

Security Enhancements for Mobile Ad HocNetworks with Trust Management Using Uncertain

ReasoningZhexiong Wei, Helen Tang, F. Richard Yu, Maoyu Wang, and Peter Mason

Abstract The distinctive features of mobile ad hoc net-works (MANETs), including dynamic topology and openwireless medium, may lead MANETs suffering from manysecurity vulnerabilities. In this paper, using recent advancesin uncertain reasoning originated from artificial intelligencecommunity, we propose a unified trust management schemethat enhances the security in MANETs. In the proposed trustmanagement scheme, the trust model has two components:trust from direct observation and trust from indirect observa-tion. With direct observation from an observer node, the trustvalue is derived using Bayesian inference, which is a typeof uncertain reasoning when the full probability model canbe defined. On the other hand, with indirect observation, alsocalled secondhand information that is obtained from neighbornodes of the observer node, the trust value is derived using theDempster-Shafer theory, which is another type of uncertainreasoning when the proposition of interest can be derivedby an indirect method. Combining these two components inthe trust model, we can obtain more accurate trust valuesof the observed nodes in MANETs. We then evaluate ourscheme under the scenario of MANET routing. Extensivesimulation results show the effectiveness of the proposedscheme. Specifically, throughput and packet delivery ratio canbe improved significantly with slightly increased average end-to-end delay and overhead of messages.

Index Terms - MANETs, Security, Trust Management, Un-certain Reasoning.

I. INTRODUCTION

With recent advances in wireless technologies and mobiledevices, Mobile Ad hoc Networks (MANETs) [1], [2] have be-come popular as a key communication technology in militarytactical environments such as establishment of communicationnetworks used to coordinate military deployment among the

Copyright (c) 2014 IEEE. Personal use of this material is permitted.However, permission to use this material for any other purposes must beobtained from the IEEE by sending a request to [email protected].

Zhexiong Wei and F. Richard Yu are with the Department of Systemsand Computer Engineering, Carleton University, e-mail: {zhexiong wei,richard yu}@carleton.ca.

Helen Tang and Peter Mason are with the Defence R&D Canada-Ottawa,Ottawa, ON, Canada, e-mail: {helen.tang, peter.mason}@drdc-rddc.gc.ca.

Maoyu Wang is with the Communications Research Centre Canada, Ottawa,ON, Canada, e-mail: [email protected].

soldiers, vehicles, and operational command centers [3]. Thereare many risks in military environments needed to be con-sidered seriously due to the distinctive features of MANETs,including open wireless transmission medium, nomadic anddistributed nature, lack of centralized infrastructure of securityprotection [4]–[6]. Therefore, security in tactical MANETs isa challenging research topic [7].

There are two complementary classes of approaches that cansafeguard tactical MANETs: prevention-based and detection-based approaches [8]. Prevention-based approaches are studiedcomprehensively in MANETs [9]–[12]. One issue of theseprevention-based approaches is that a centralized key man-agement infrastructure is needed, which may not be realisticin distributed networks such as MANETs. In addition, acentralized infrastructure will be the main target of rivalsin battlefields. If the infrastructure is destroyed, then thewhole network may be paralyzed [13]. Furthermore, althoughprevention-based approaches can prevent misbehavior, thereare still chances remained for malicious nodes to participatein the routing procedure and disturb proper routing estab-lishment. From the experience in the design of security inwired networks, multi-level security mechanisms are needed.In MANETs, this is especially true given the low physicalsecurity of mobile devices [14], [15]. Serving as the secondwall of protection, detection-based approaches can effectivelyhelp identify malicious activities [16]–[18].

Although some excellent work has been done on detection-based approaches based on trust in MANETs, most of existingapproaches do not exploit direct and indirect observation (alsocalled secondhand information that is obtained from third partynodes) at the same time to evaluate the trust of an observednode. Moreover, indirect observation in most approaches isonly used to assess the reliability of nodes, which are not inthe range of the observer node [19]–[21]. Therefore, inaccuratetrust values may be derived. In addition, most methods oftrust evaluation from direct observation [19], [20], [22] donot differentiate data packets and control packets. However,in MANETs, control packets usually are more important thandata packets.

In this paper, we interpret trust as the degree of beliefthat a node performs as expected. We also recognize un-certainty in trust evaluation. Based on this interpretation, wepropose a trust management scheme to enhance the securityof MANETs. The difference between our scheme and existingschemes is that we use uncertain reasoning to derive trustvalues. Uncertain reasoning was initially proposed from the



2

artificial intelligence community to solve the problems inexpert systems, which have frequent counter-factual results[23]. The elasticity and flexibility of uncertain reasoning makeit successful in many fields, such as expert systems, multi-agent systems, and data fusion [23]–[26]. The contributionsof this paper are outlined as follows:

• We propose a unified trust management scheme thatenhances the security in MANETs using uncertain rea-soning. In the proposed scheme, the trust model has twocomponents: trust from direct observation and trust fromindirect observation. With direct observation from anobserver node, the trust value is derived using Bayesianinference, which is a type of uncertain reasoning whenthe full probability model can be defined. On the otherhand, with indirect observation from neighbor nodesof the observer node, the trust value is derived usingthe Dempster-Shafer theory, which is another type ofuncertain reasoning when the proposition of interest canbe derived by an indirect method.

• The proposed scheme differentiates data packets andcontrol packets, and meanwhile excludes the other causesthat result in dropping packets, such as unreliable wirelessconnections and buffer overflows.

• We evaluate the proposed scheme in a MANET routingprotocol, the optimized link state routing protocol version2 (OLSRv2) [27], with the Qualnet simulator. Extensivesimulation results show the effectiveness of the proposedscheme. Throughput and packet delivery ratio can beimproved significantly, with slightly increased averageend-to-end delay and overhead of messages.

The remainder of this paper is organized as follows. Relatedwork is presented in Section II. The trust model and its twocomponents are presented in Section III. Section IV depicts theBayesian methodology and how to use it in trust evaluationfrom direct observation. Section V describes the Dempster-Shafer theory and how to use it in trust evaluation fromindirect observation. The performance and effectiveness of ourscheme are evaluated and discussed in Section VII. Finally, weconclude the work in Section VIII.

II. RELATED WORK

Trust-based security schemes are important detection-basedmethods in MANETs, which have been studied recently [19],[20], [22], [24]–[26], [28]–[31]. In [19], [20], the trust value ofa node based on direct observation is derived using Bayesianmethodology. The authors of [22] regard trust as uncertaintythat the observed node performs a task correctly, and entropyis used to formulate a trust model and evaluate trust valuesby direct observation. Compared to direct observation in trustevaluation, indirect observation or second-hand informationcan be important to assess the trust of observed nodes. Forexample, the collection of testimonies from neighbor nodescan detect the situation where a hostile node performs well toone observer, while performing poorly according to anothernode.

The Dempster-Shafer theory (DST) is regarded as a usefulmechanism in uncertain reasoning and is widely used in

expert systems and multi-agent systems [24], [25]. In [26],the Dempster-Shafer theory is used in sensor fusion. Intrusiondetection systems [29], [30] apply the Dempster-Shafer theoryto assess unreliable information from IDS sensors.

In this paper, we use uncertain reasoning theory from artifi-cial intelligence to evaluate the trust of nodes in MANETs.Uncertainty is an old problem from gambler’s world. Thisproblem can be handled by probability theory. Reasoningis another important behavior in everyday life. A lot ofresearchers, even Aristotle (384 BCE - 322 BCE) (GreekPhilosopher), try to understand and formulate it. Reasoningbased on uncertainty has been prosperous in the artificialintelligence community due to the development of probabilitytheory and symbolic logic. Probabilistic reasoning is intro-duced to intelligence systems [23], which is used to tacklethe exceptions in automatic reasoning. In order to overcomethe drawbacks of traditional rule-based systems, which arebased on truth tables with no exceptions, probabilistic rea-soning is proposed, in which the uncertainty of knowledgeis considered and described as subsets of ”possible worlds.”Probabilistic reasoning can be used to different areas, fromartificial intelligence to philosophy, cognitive psychology, andmanagement science. In the area of security in MANETs, wefind that this theory is very suitable for trust evaluation basedon the trust interpretation in this paper. Bayesian inferenceand Dempster-Shafer evidence theory are two approaches inuncertain reasoning. We adopt them to evaluate trust of nodesby direct and indirect observation.

Trust based security systems are also studied in differentnetwork architectures, e.g., wireless sensor networks [32],[33], vehicular ad hoc networks (VANETs) [34], coopera-tive wireless networks [35], etc. Although different types ofnetworks have different specific characteristics, the proposedtrust model based on direct and indirect observation is generalenough and can be customized to a particular network.

To make it easier to understand the proposed trust model,we present an overview of OLSRv2 and the vulnerabilities ofOLSRv2. OLSRv2 is a proactive routing protocol, which isa new version of OLSR [36]. OLSRv2 inherits OLSR’s corealgorithms and also introduces some new features [27]: routingMultipoint Relay (MPR), flexible link metrics, extensiblemessage formats, etc. OLSRv2 has three basic components:Neighborhood Discovery, MPR Selection, and Topology Es-tablishment, as well as two types of control messages: aHELLO message and Topology Control message [27]. Neigh-borhood Discovery [37] is used to facilitate a node’s discoveryof its one-hop neighbors in radio range. HELLO messages,which can carry link status such as symmetric, asymmetric,or multipoint relay, are used in the neighborhood discoveryprocedure. Through periodically sending HELLO messages, anode can establish bi-directional (symmetric) links with itsone-hop neighbors. Two types of MPR selection: floodingMPR selection and routing MPR selection, are performedin OLSRv2. Flooding MPR selection plays a key role inforwarding control traffic in the network. A node selected bya neighbor of this node as a flooding MPR will forward themessage from the neighbor once. This mechanism can effec-tively reduce the total transmission in the network. Routing



3

TABLE IMAIN NOTATIONS

Notation DefinitionTAB The total trust value that Node A

gives Node BTSAB The trust value that Node A gives

Node B based on direct observation of Node ATNAB The trust value that Node A gives

Node B based on indirect observation of Node ATDAB The trust value that Node A gives

Node B based on data packetsTCAB The trust value that Node A gives

Node B based on control packetsλ The weight for the trust value

based on direct observationρ The weight for the trust value

based on data packetsγ A factor of punishment

which is larger than or equal to 1

MPR selection is separated from flooding MPR selection inOLSRv2. The main function of routing MPRs is to disseminatelink state information by TC messages. This mechanism canreduce the size of link statement information that is sent inorder to limit the network topology. A network topology isestablished by link state information diffusion. Routing MPRsconnect multi-hop routes in the network.

Although OLSRv2 provides extensions for customization,there is no specific security mechanism in the protocol. Con-sequently, OLSRv2 is susceptible to various attacks such asworm hole attacks, black hole attacks, spoofing, jamming,and so on [38], [39]. In this paper, our scheme is a securitymechanism that mainly protects OLSRv2 against two types ofmisbehavior, dropping packets and modifying packets. Packetdropping attack is also called as a black hole attack, whichis a type of denial-of-service attacks [40]. Modification ofpackets may have a significant impact on a topology map [9].Through trust evaluation in our scheme, malicious nodes thatdeliberately drop or modify packets can be detected and keptaway.

III. TRUST MODEL IN MANETS

In this section, we describe the definition and properties oftrust in MANETs. Based on the definition, we depict the trustmodel that is used to formulate the trust between two nodes inMANETs, and present a framework of the proposed scheme.The main notations that are used in this paper are summarizedin Table I.

A. Definition and Properties of Trust

Trust has different meanings in different disciplines frompsychology to economy [28]. The definition of trust inMANETs is similar to the explanation in sociology, where trustis interpreted as degrees of the belief that a node in a network(or an agent in a distributed system) will carry out tasks that itshould [28]. Due to the specific characteristics of MANETs,trust in MANETs has five basic properties: subjectivity, dy-namicity, non-transitivity, asymmetry, and context-dependency[28]. Subjectivity means that an observer node has a right todetermine the trust of an observed node. Different observer

nodes may have different trust values of the same observednode. Dynamicity means that the trust of a node should bechanged depending on its behaviors. Non-transitivity meansthat if node A trusts node B and node B trusts node C, thennode A does not necessarily trust node C. Asymmetry meansthat if node A trusts node B, then node B does not necessarilytrust node A. Context-dependency means that trust assessmentcommonly bases on the behaviors of a node. Different aspectsof actions can be evaluated by different trust. For example,if a node has less power, then it may not be able to forwardmessages to its neighbors. In this situation, the trust of powerin this node will decline, but the trust of security in this nodewill not be changed due to its state.

Reputation is another important concept in trust evaluation.Reputation reflects the public opinions from members in acommunity [41]. In MANETs, reputation can be a collectionof trust from nodes in the network. Reputation is more globalthan trust from the perspective of the whole network [41].

B. Trust Model

Based on the definition and properties of trust in MANETs,we evaluate trust in the proposed scheme by a real number,T , with a continuous value between 0 and 1. Although trustand trustworthiness may be different in contexts, in which thetrustor needs to consider risk [28], trust and trustworthinessare treated the same for simplicity in the proposed scheme.

In this model, trust is made up of two components: directobservation trust and indirect observation trust. These compo-nents are similar to those used in [42]. In direction observationtrust, an observer estimates the trust of his one-hop neighborbased on its own opinion. Therefore, the trust value is theexpectation of a subjective probability that a trustor uses todecide whether or not a trustee is reliable. It is similar to first-hand information defined by [19], [20].

We denote T S as a trust value from direct observationand can be calculated by Bayesian inference. The detailedexplanation is in Section IV.

If we only consider direct observation, there would beprejudice in trust value calculation. In order to obtain lessbiased trust value, we also consider other observers’ opinionsin this paper. Although opinions of neighbors are introducedin [42], the method that simply takes arithmetic mean of alltrust values is not sufficient to reflect the real meaning of otherunreliable observers’ opinions because there are two situationsthat may severely disturb the effective evidence from neigh-bors: unreliable neighbors and unreliable observation [29].Unreliable neighbors themselves are suspects. Even thoughneighbors are trustworthy, they may also provide unreliableevidence due to observation conditions. The Dempster-Shafertheory [25], [29] is a good candidate to aid in this situation,in which evidence is collected from neighbors that may beunreliable. Therefore, We denote the trust value derived fromindirect observation of one-hop neighbors as T N . Combiningthe trust value, T S , from direct observation and the trust value,TN , from indirect observation, we can get a more realistic andaccurate trust value of a node in MANETs.

T = λT S + (1− λ)TN , (1)



4

Application

Networking

Trust Repository

Trust Evaluation andUpdate

DirectObservation

IndirectObservation

Trust Scheme

Fig. 1. The framework of the proposed scheme.

where λ is a weight assigned to T S , 0 ≤ λ ≤ 1.

C. Framework of the Proposed Scheme

Based on the trust model, the framework of the proposedscheme is shown in Fig. 1. In the trust scheme component,the module of trust evaluation and update can obtain evidencefrom direct and indirect observation modules and then utilizetwo approaches, Bayesian inference and DST, to calculate andupdate the trust values. Next, the trust values are stored in themodule of trust repository. Routing schemes in the networkingcomponent can establish secure routing paths between sourcesand destinations based on the trust repository module. Theapplication component can send data through secure routingpaths.

The trust from direct observation between an observer nodeA and an observed node B in this trust scheme can be definedfurther as

T SAB = ρTD

AB + (1− ρ)TCAB, (2)

where ρ (0 ≤ ρ ≤ 1)is the weight for data packets; T DAB is the

trust value based on data packets; T CAB is the trust value based

on control packets. Trust from indirect observation between anobserver node A and an observed node B, denoted as T N

AB,can be obtained by DST, which will be explained in SectionV.

In order to explain the basic procedure of trust evaluationin our scenario, an example network is shown in Fig. 2. Inthis example, node 1 is an observer node and node 3 is anobserved node. Node 1 sends data messages to node 5 throughnode 3. When node 3 receives data messages and forwards tonode 5, node 1 can overhear it. Then node 1 can calculate thetrust value of node 3 based on data messages. The same ideais applied to the control message situation. In the meanwhile,node 1 can collect information from node 2 and node 4, whichhave interactions with node 3 in order to evaluate the trustvalue of node 3. This information collected from third partynodes is called indirection observation. In another situation,node 7 sends data messages to node 3, which is the destinationnode. Node 1 cannot overhear the data messages sent to node3 in this situation.

1

2

3

4 6

5

7

Direction of message flow

Data message

Control message

Bi-directional connection

Node

Fig. 2. An example mobile ad hoc network.

IV. TRUST EVALUATION WITH DIRECT OBSERVATION

Based on the model presented in the last section, weevaluate trust values with direct observation on two maliciousbehaviors: dropping packets and modifying packets [40]. Inthe direct observation, we assume that each observer canoverhear packets forwarded by an observed node and comparethem with original packets so that the observer can identifythe malicious behaviors of the observed node. Therefore, theobserver node can calculate trust values of its neighbors byusing Bayesian inference, which is a general framework todeduce the estimation of the unknown probability by usingobservation. As mentioned in the last section of trust model,the degree of belief is a random variable, denoted by Θ and0 ≤ θ ≤ 1. From Bayes’ theorem, we can derive the followingformulation

f(θ, y|x) = p(x|θ, y)f(θ, y)∫ 1

0p(x|θ, y)f(θ, y) dθ

, (3)

where x is the number of packets is forwarded correctly; y isthe number of packets is received by a node; p(x|θ, y) is thelikelihood function, which follows a binomial distribution

p(x|θ, y) =(y

x

)θx(1 − θ)y−x. (4)

We assume that the prior distribution, f(θ, y), follows Betadistribution,

Beta(θ;α, β) =θα−1(1− θ)β−1∫ 1

0θα−1(1 − θ)β−1 dθ

, (5)

where 0 ≤ θ ≤ 1, α > 0, β > 0. Then we have

f(θ, y|x) ∼ Beta(α+ x, β + y − x). (6)

The expectation of Beta distribution is

E[Θ] =α

α+ β. (7)

Due to reproductivity of (6), the trust value is calculatediteratively. At the beginning, there are no observation. Theprior distribution f(θ, y) is Beta(θ; 1, 1) at the beginning.Then we have

En[Θ] =αn

αn + βn, (8)



5

where αn = αn−1 + xn−1, βn = βn−1 + yn−1 − xn−1, α0 =β0 = 1, n ∈ Z

+. Intuitively, this situation is explained that thetrust value of a node is 0.5 at the beginning. That means thenode is seemed as neutral when no history records behaviorsis established. The value trust can be revised continuouslythrough follow-up observation.

Past experience is also an important factor when trust valuesare calculated. Recent activities of a node can seriously affectthe trust evaluation. Consider the case where a node has agood history of past experience, but it drops or modifiespackets recently. In order to handle this, a windowing schemeis proposed. Using weighted evidence from observation isanother method [19].

In our scheme, we introduce a punishment factor for reputa-tion fading, which focuses on recent activities. The punishmentfactor is used to give more weights on misbehavior in theBayesian framework. Firstly, this can lower the trust of anattacker when it misbehaves. Secondly, the trust of the attackwill not recover quickly even if it forwards a large number ofpackets correctly due to the impact of the punishment factor.This can help the proposed scheme distinguish the maliciousnode quickly and avoid them disrupting the normal trafficbetween benign nodes again. The punishment factor is inspiredby our daily lives in human society, where a scandal canbadly affect a person who has a good reputation. What’s more,it is hard to quickly recover a good reputation. The factorof punishment makes the trust evaluation more realistic. Thepunishment factor, γ, in the formula of trust evaluation in (8)is described as follows:

En[Θ] =αn

αn + γβn, (9)

where γ ≥ 1. As the value of γ becomes larger, the trustvalue declines more. This is because the punishment factorgives more weight to misbehavior. Based on this deduction,T S is defined as:

T S = En[Θ]. (10)

V. TRUST EVALUATION WITH INDIRECT OBSERVATION

In this section, indirect observation from neighbor nodesused to evaluate the trust value of the observed node willbe discussed. Although direct observation from an observeris important in assessing the trust value of the observednode, the testimonies from neighbor nodes are also helpfulto judge the trustworthiness of the observed node. Collectionof neighbors’ opinions can help in justifying whether or nota node is hostile. This mechanism may reduce the bias froman observer. A situation in which a node is benign to onenode but malicious to others may be mitigated. In order toimplement this method, the Dempster-Shafer theory, whichis a mathematical theory of evidence, is used as it is welldeveloped for coping with uncertainty or ignorance, and itprovides a numerical measurement of degrees of belief abouta proposition from multiple sources [26], [30]. The core ofthis theory is the belief function that is based on two essentialideas: degrees of belief about a proposition can be obtainedfrom subjective probabilities of a related question, and thesedegrees of belief can be combined together on condition that

B

j1

jn

A

j2

Fig. 3. A scenario for indirect observation.

they are from independence evidence [24], [29]. In the indirectobservation, we assume that there are more than one neighbornodes between an observer and an observed node when thetrust evaluation is performed with DST. We also assume thatevidence provided by different neighbors is independent.

First, we will introduce the theory of belief functions. Thenwe will discuss the rule of combining belief functions thatare used to accommodate testimonies from one-hop neighbornodes in order to assess trust values of nodes in MANETs.

A. Belief Function

In the Dempster-Shafer theory, a frame of discernment is aset of propositions that are mutually exclusive and exhaus-tive, which is denoted by Ω [29]. Based on the frame ofdiscernment, the basic probability value of a focal set, A i,is a function m : 2Ω −→ [0, 1], which satisfies followingconditions: m(∅) = 0; and

∑Ai⊆Ω m(Ai) = 1. For any subset

B of the frame of discernment, the belief function is definedas

bel(B) =∑

Ai⊆B

m(Ai). (11)

In our scenario shown in Fig. 3, we designate two secu-rity states to a node, i.e., {trustworthy, untrustworthy},which is similar to [25], [29]. Therefore, the frameof discernment in the Dempster-Shafer theory, Ω ={trustworthy, untrustworthy}, which demonstrates thatnode B has two states: trustworthy and untrustworthy. NodeA evaluates the trust value of node B through one-hopneighbors between them. One-hop neighbors of node B canprovide evidence to a subset of Ω with hypothesis H , i.e.,node B is trustworthy. The power set of our scenario, 2Ω,includes: ∅; hypothesis H = {trustworthy}; hypothesisH = {untrustworthy}; and hypothesis U = Ω, which meansthat the observed node B is either in the trustworthy stateor untrustworthy state. Each one-hop neighbor gives evidencefrom its observation by assigning its beliefs over Ω. Eachhypothesis is assigned a basic probability value m(H) between0 and 1. In our scheme, the basic probability value can beobtained from direct observation. For example, the trust valueof node j1 is TS

Aj1, from direct observation of node A to node

j1. If node j1 believes that node B is trustworthy, then the basicprobability value mj1(H) is TS

Aj1. The basic probability value



6

mj1(H) is 0. From the definition of belief function, mj1(U)is equal to 1− T S

Aj1. The formulae are as follows [29]:

mj1(H) = T SAj1 ,

mj1(H) = 0,

mj1(U) = 1− T SAj1 , (12)

If node j1 believes that node B is untrustworthy, the formulaeare as follows:

mj1(H) = 0,

mj1(H) = T SAj1 ,

mj1(U) = 1− T SAj1 , (13)

The belief function of each focal set can be obtained from(11). For example,

belj1(H) = mj1(H),

belj1(H) = mj1(H),

belj1(U) = mj1(H) +mj1(H) +mj1(U). (14)

This means that from the testimony of node j1, node A canderive whether or not node B is trustworthy based on the trustvalue of node j1.

B. Dempster’s Rule of Combining Belief Functions

Based on the above description of belief function,Dempster-Shafer theory combines multiple neighbor nodes’belief on the condition that evidence from different neighbornodes is independent [25], [29]. Assuming that bel 1(B) andbel2(B) are two belief functions over the same frame ofdiscernment, Ω, the orthogonal sum of bel1(B) and bel2(B),bel(B), is defined as

bel(B) = bel1(B)⊕ bel2(B)

=

∑i,j,Ai

⋂Aj=B m1(Ai)m2(Aj)∑

i,j,Ai

⋂Aj �=∅

m1(Ai)m2(Aj), (15)

where Ai, Aj ⊆ Ω. The order of the combination of belieffunctions does not affect the result value produced by Demp-ster’ rule due to the commutativity of multiplication [25].

In our scenario, we assume that there are one-hop neighborsbeside node B as shown in Fig. 3. Therefore, the combinedbelief of node j1 and node j2 is calculated as follows [29]:

mj1(H)⊕mj2(H) =1

K[mj1(H)mj2(H)

+mj1(H)mj2(U)

+mj1(U)mj2(H)],

mj1(H)⊕mj2(H) =1

K[mj1(H)mj2(H)

+mj1(H)mj2(U)

+mj1(U)mj2(H)],

mj1(U)⊕mj2(U) =1

Kmj1(U)mj2(U), (16)

where

K = mj1(H)mj2(H) +mj1(H)mj2(U)

+mj1(U)mj2 (U) +mj1(U)mj2 (H)

+mj1(U)mj2 (H) +mj1(H)mj2 (H)

+mj1(H)mj2 (U). (17)

For instance, assuming that

mj1(H) = 0.8,mj1(H) = 0,mj1(U) = 0.2,

mj2(H) = 0.7,mj2(H) = 0,mj2(U) = 0.3,

then we can obtain the result of combining two belief functionsas follows:

bel(H) = 0.8 ∗ 0.7 + 0.8 ∗ 0.3 + 0.7 ∗ 0.2 = 0.94

bel(H) = 0 ∗ 0 + 0 ∗ 0.3 + 0 ∗ 0.2 = 0

bel(U) = 0.2 ∗ 0.3 = 0.06

That means from the result of combination, the trust value ofnode B from indirect observation is 0.94.

Following the rule of combination of belief, we can combinemore results from neighbor nodes. Based on the Dempster-Shafer theory, T N

AB is defined as:

TNAB = mj1(H)⊕mj2(H) . . .⊕mjn(H), (18)

where node ji, 1 ≤ i ≤ n, is an one-hop neighbor of node Aand node B.

VI. SECURE ROUTING BASED ON TRUST

The original OLSRv2 [27] does not provide security mea-surements in the protocol. OLSRv2 assumes that every nodeis cooperative and benevolent. However, this assumption isinappropriate in a military environment. Malicious nodes canattack nodes that are not protected. Based on trust values, asecure route can be established.

Modifications of OLSRv2 include two important parts: routeselection process based on link metrics and trust value calcu-lation algorithms. Although OLSRv2 provides new featuressuch as link metrics and extensible message formats, whichmay be used to improve security of the protocol, OLSRv2implementation [27] [43] still attempts to use hop countwhen the shortest routing path is calculated. In order toimplement route selection process based on link metrics, thereare three components that need to be changed, HELLO andTC messages, protocol information bases, and the shortestpath algorithm. Message format is extensible and flexiblein OLSRv2. Thus link metrics information can be added tomessages as Type Length Value (TLV) blocks. Modification ofprotocol information bases, including local information base,neighbor information base and topology information base, isused to record link metrics in each node. Based on theseinformation bases, route processing set can update the shortestrouting path with link metrics.

Based on the Internet draft of OLSRv2 [27], there aretwo types of control messages, HELLO and TC. In this trustmanagement, we only consider the TC messages because ofthe need for forwarding TC. The message type of TC, which



7

is defined in OLSRv2 Internet draft, can be used to check thetype of the message. The trust management scheme can sepa-rate the data and control messages by the message type duringtrust evaluation. For other standard protocols, like AODV [44],the trust management scheme also can differentiate the controlmessages, e.g., RREQs, RREPs in AODV, by message typechecking when a trust evaluation procedure is performed.

We assume that each node works in the promiscuous modeimplemented by the MAC layer. We also assume that, in atime slot, the observed node (sender) does not move out ofthe transmission range. As the time of packets processingin a node is short, our assumptions are realistic in practicalnetworks. This means that the observer can detect whether ornot the neighboring node sends the received packets beforethe observed node moves out the transmission range.

Every node needs to record its one-hop neighbors, howmany data packets each neighbor received, how many controlpackets each neighbor received, how many data packets eachneighbor forwards correctly, and how many control packetseach neighbor forwards correctly. In OLSRv2, there are twotypes of control messages: HELLO and TC. TC message isonly recorded for trust evaluation because HELLO messageis transmitted with one hop in the network. When a nodereceives a packet, the number of received packets, accordingto the type, will increase one. If the node forwards thereceived packet correctly, the number of forwarded packetswill increase one. There are three scenarios that the numberof received packets will not increase. Firstly, if the packet isdropped because of time to live (TTL), then the number ofreceived packets should not increase. Secondly, if a node thatreceives a packet drops it due to overflow of buffers. Thirdly,a packet is dropped by a node because the state of wirelessconnection is bad. Considering these significant factors, weimprove the accuracy of trust calculation.

In this paper, we consider the condition that packets aredropped due to unreliable wireless connections. During thetrust evaluation with direct observation, the scheme can re-move the number of packets dropped by this condition (inAlgorithm 1). We assume that there is a probability that pack-ets are dropped because of unreliable wireless connections.

Algorithm 1 depicts the details of each iteration. Algorithm2 describes that an observer node collects evidence fromits one-hops neighbors between the observer node and theobserved node. Then the trust values from indirect observationare evaluated by (18). After T S and TN are obtained, wecan get the total trust value of the observed node by (1).In proactive routing protocols, such as OLSRv2, an observernode can obtain the information from its neighbor nodesperiodically by control messages (e.g., HELLO and TC),which can be used to carry the trust values.

Compared to the existing OLSRv2 scheme that uses theshortest path based on hop count, we derive the best routingpath considering both trust values and hop count. We use theDijkstra’ algorithm to calculate the best routing path. Sinceminimization is used in the Dijkstra’ algorithm (e.g., to findthe shortest path with the minimal hop count in traditionalOLSRv2), we need to convert the trust value to untrustworthyvalue. Then, we can minimize the untrustworthy value of a

Algorithm 1 Trust Calculation with Direct Observation1: if node A, which is an observer, finds that its one-hop

neighbor, Node B that is a trustee, receives a packet then2: the number of packets received increases one3: if node A finds that node B forwards the packet suc-

cessfully then4: the number of packets forwarded increases one5: else6: if TTL of the packet becomes zero or overflow of

buffers in node B or the state of wireless connectionof node B is bad then

7: the number of packets received decreases one8: end if9: end if

10: end if11: calculate the trust value, T S , from (8) and update the old

one.

Algorithm 2 Trust Calculation with Indirect Observationif node A, which is an observer, has more than one one-hop neighbors between it and the trustee, node B then

2: calculates the trust value, TN , from (18)else

4: set TN to 0set λ to 1

6: end if

path using the Dijkstra’ algorithm. To this end, we define theuntrustworthy value between node A and node B as UAB ,which can be calculated as UAB = 1 − TAB. The sum ofuntrustworthy values of a path is

Upath =n−1∑i=1

Ukiki+1 =n−1∑i=1

(1− Tkiki+1), (19)

where Tkiki+1 is the trust value between node ki and its one-hop neighbor, node ki+1. Nodes k1, k2, . . . , kn belong tothe path with n − 1 hops. The best routing path satisfies theminimum of Upath.

The trust values and routing table of each node can be storedin the Trust Platform Module (TPM) [45], which providesadditional security protection in open environments with thecombination of software and hardware. Since the trust valuesin each node are the key facilities to detect malicious nodes,the TPM is able to provide effective protection to securerouting to avoid malicious attacks by enemies in battlefields.

VII. SIMULATION RESULTS AND DISCUSSIONS

The proposed scheme is simulated on the Qualnet platformwith the OLSRv2 protocol. In the simulations, the effective-ness of the scheme is evaluated in an insecure environment.We compare the performance of the proposed scheme withthat of OLSRv2 without security mechanisms.

A. Environment Settings

We randomly place nodes in the defined area. Each scenariohas a pair of nodes as the source and destination with Constant



8

4

2

5

16

7

10

8

3

9

Fig. 4. An example of the network setup.

Bit Rate (CBR) traffic. The simulation parameters are listedin Table II. In our simulations, we assume that there are twotypes of nodes in the network: normal nodes, which followthe routing rules, and compromised nodes, which drop ormodify packets maliciously. We also assume that the numberof compromised nodes is minor compared to the total numberof nodes in the network. In this adversary mode, the proposedscheme is evaluated and compared with the original OLSRv2protocol. We have simulated networks with different numbersof nodes. Fig. 4 is an example of the network setup wherenode 1 is the source node that generates the CBR traffic, node3 is the destination node, and node 2 is compromised by anadversary. For node mobility, the random waypoint mobilitymodel is adopted in a 30-node MANET. The maximumvelocity of each node is set from 0 to 10 m/s. The pausetime is 30 seconds.

There are four performance metrics considered in the sim-ulations: 1) Packet delivery ratio (PDR) is the ratio of thenumber of data packets received by a destination node andthe number of data packets generated by a source node; 2)Throughput is the total size of data packets correctly receivedby a destination node every second; 3) Average end-to-enddelay is the mean of end-to-end delay between a sourcenode and a destination node with CBR traffic; 4) MessageOverhead is the size of Type Length Value (TLV) blocks intotal messages, which are used to carry trust values; 5) Routingload is the ratio of the number of control packets transmittedby nodes to the number of data packets received successfullyby destinations during the simulation.

B. Performance Improvement

The original OLSRv2 and our scheme are evaluated in thesimulations, where some nodes misbehave through dropping ormodifying packets. In Fig. 5, we compare our scheme with andwithout indirect observation and original OLSRv2 in scenariosthat a source node sends data packets to a destination node inthe network, which includes nodes from 5 to 30.

From Fig. 5, we can see that the proposed scheme hasa much higher PDR than the existing scheme because thetrust based routing calculation can detect the misbehaviorof malicious nodes. The results also demonstrate that theproposed scheme with indirect observation has the highestPDR among these three schemes. In Fig. 5, we also can

TABLE IISIMULATION PARAMETERS

Parameter ValueApplication protocol CBR

CBR transmission time 1s to 100sCBR transmission interval 0.5s

Packet size 512 bytesTransport protocol UDPNetwork protocol IPv4Routing protocol OLSRv2MAC protocol IEEE 802.11

Physical protocol IEEE 802.11bData rate 2Mbps

Transmission power 6dBmRadio range 180m

Propagation pathloss model Two-raySimulation area 300m × 300m, 500m × 500m,

800m × 800m, 1000m × 1000mNumber of nodes 5, 10, 15, 20, 25, 30Simulation time 300s

5 10 15 20 25 300.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Number of Nodes

Pac

ket D

eliv

ery

Rat

io

Proposed Scheme w/ Indirect ObservationProposed Scheme w/o Indirect ObservationExisting Scheme

Fig. 5. Packet delivery ratio (PDR) versus the number of nodes in thenetwork.

find that the PDR of three schemes decreases gradually whenthe number of nodes grows. This is because the collision ofsending messages becomes more frequent as the number ofnodes increases in the MANET. Although the PDR declinesin three schemes, the proposed scheme is apparently betterthan the existing scheme. In Fig. 6, we evaluate throughputin our scheme and the original one. Although the number ofpackets received correctly decreases as long as the numberof nodes increases, the performance of our scheme has abig improvement. Fig. 5 and Fig. 6 both reveal that thetrust based routing algorithm can improve the performance ofOLSRv2. Fig. 7 and Fig. 8 show the impact of node mobilityin a 30-node MANET. We can observe that, as the nodevelocity increases, PDR and throughput decrease gradually.This is because the higher speed of a node may increase theprobability of packets lost. Nevertheless, the proposed schemehas better performance than the existing one.

The number of malicious nodes in the MANET also hasa significant impact on the throughput of the network. Here,we assume the attackers are independent. Hence, there is nocollusion attack in the MANET. We investigate the throughputwith malicious nodes, from 2 to 10, in a 30-node MANET



9

5 10 15 20 25 301

2

3

4

5

6

7

8

9

Number of Nodes

Thr

ough

put (

kbps

)


Fig. 6. Throughput versus the number of nodes in the network.

0 1 2 3 4 5 6 7 8 9 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Node Velocity

Pac

ket D

eliv

ery

Rat

io


Fig. 7. Packet delivery ratio versus node velocity.

environment. The basic parameter is the same as above. Fig.9 shows that, as the number of malicious nodes increases, thethroughput drops dramatically. When the number of maliciousnodes reaches to one third of the total number of nodes inthe network, the throughput decreases to about half of thethroughput in the network with 2 malicious nodes. From thisfigure, we can see that the proposed scheme is affected deeplyby the number of malicious nodes. Compared to the proposedscheme, the existing scheme has a very low throughput evenif the number of malicious nodes is very small.

From these three figures, we can observe that our proposedscheme based on trust outperforms the existing scheme signifi-cantly in terms of both PDR and throughput. Our scheme takesadvantage of trust evaluation of nodes in the network so thatmore reliable routing paths can be established. The existingscheme is severely affected by malicious nodes that drop ormodify packets. We can observe that the proposed schemewith trust can steer clear of malicious nodes dynamically.Therefore, the PDR and throughput of our scheme are betterthan those of the existing scheme.

C. Cost

The cost of security enhancement in OLSRv2 mainly in-cludes the increased average end-to-end delay and overhead of

0 1 2 3 4 5 6 7 8 9 100

1

2

3

4

5

6

7

Node Velocity

Thr

ough

put (

kbps

)


Fig. 8. Throughput versus node velocity.

2 4 6 8 100

1

2

3

4

5

6

7

Number of Malicious Nodes

Thr

ough

put (

kbps

)


Fig. 9. Throughput versus the number of malicious nodes in the network.

5 10 15 20 25 305

10

15

20

25

30

Number of Nodes

Ave

rage

End

−to

−en

d D

elay

(m

s)


Fig. 10. Average end-to-end delay versus the number of nodes in the network.



10

0 1 2 3 4 5 6 7 8 9 1020

25

30

35

40

45

50

Node Velocity (m/s)

Ave

rage

End

−to

−en

d D

elay

(m

s)


Fig. 11. Average end-to-end delay versus node velocities.

5 10 15 20 25 300

2

4

6

8

10

12

14x 10

5

Number of Nodes

Tot

al B

ytes

of M

essa

ges

Sen

t

Proposed Scheme w/o Indirect ObservationProposed Scheme w/ Indirect ObservationExisting Scheme

Fig. 12. Total bytes of messages sent versus the number of nodes in thenetwork.

5 10 15 20 25 3010

11

12

13

14

15

16

17

18

Number of Nodes

Per

cent

age

of T

rust

Val

ue O

verh

ead

in M

essa

ge

Fig. 13. Percentage of overhead in message versus the number of nodes inthe network.

5 10 15 20 25 300

50

100

150

200

250

300

350

400

Number of Nodes

Rou

ting

Load

Existing SchemeProposed Scheme w/o Indirect ObservationProposed Scheme w/ Indirect Observation

Fig. 14. Routing load versus the number of nodes in the network.

messages that are used to carry trust values of nodes. Fig. 10shows that the proposed scheme has a slightly higher averageend-to-end delay than the existing scheme in the maliciousenvironment. In Fig. 11, we can see that, as the node velocityincreases, the average end-to-end delay becomes longer. Thereason is that trust based routing path is usually a longer routefrom a source node to a destination node. Therefore, there is atrivial delay introduced by the proposed scheme. Nevertheless,higher security is guaranteed in the proposed scheme.

Compared to local computing capacity, sending and re-ceiving message is an important issue in MANETs becausemessage transmission is energy-consuming. Thus, we studyhow much overhead of messages is imported when the trustvalue is calculated in the OLSRv2 protocol. Since the metriclink value is introduced in OLSRv2, one new address blockTLV, which occupies 12 bytes, is added to the message formatdescribed in Section VI. Fig. 12 shows how much the overheadof messages is imported compared to the original version ofOLSRv2. Because trust values are embedded in the HELLOmessages and TC messages, there is no more messages need tobe sent. The overhead is not very high. However, as the numberof nodes increases, the percentage of overhead in messagesdrops dramatically, as shown in Fig. 13. This is because, whenthe number of nodes increases, the total message becomeslarge. Then the 12-byte overhead is trivial compared to thesize of messages. In Fig. 14, the results demonstrate that theproposed scheme has a lower routing load because of thehigher number of packets received correctly by the destinationnode. As the number of nodes increases, the routing load ofthe existing and proposed schemes climb up due to the natureof proactive routing protocol: periodical generation of controlmessages in every node.

VIII. CONCLUSIONS AND FUTURE WORK

In this paper, we proposed a unified trust managementscheme that enhances the security of MANETs. Using re-cent advances in uncertain reasoning, Bayesian inference andDempster-Shafer theory, we evaluate the trust values of ob-served nodes in MANETs. Misbehaviors such as droppingor modifying packets can be detected in our scheme through



11

trust values by direct and indirect observation. Nodes withlow trust values will be excluded by the routing algorithm.Therefore, secure routing path can be established in mali-cious environments. Based on the proposed scheme, moreaccurate trust can be obtained by considering different typesof packets, indirect observation from one-hop neighbors andother important factors such as buffers of queues and statesof wireless connections, which may cause dropping packetsin friendly nodes. The results of MANET routing scenariopositively support the effectiveness and performance of ourscheme, which improves throughput and packet delivery ratioconsiderably, with slightly increased average end-to-end delayand overhead of messages. In our future work, we will extendthe proposed scheme to MANETs with cognitive radios [46].

REFERENCES

[1] S. Corson and J. Macker, “Mobile ad hoc networking (MANET): routingprotocol performance issues and evaluation considerations,” IETF RFC2501, Jan. 1999.

[2] F. R. Yu, Cognitive Radio Mobile Ad Hoc Networks. New York:Springer, 2011.

[3] J. Loo, J. Lloret, and J. H. Ortiz, Mobile Ad Hoc Networks: CurrentStatus and Future Trends. CRC Press, 2011.

[4] Q. Guan, F. R. Yu, S. Jiang, and V. Leung, “Joint topology controland authentication design in mobile ad hoc networks with cooperativecommunications,” IEEE Trans. Veh. Tech., vol. 61, pp. 2674 –2685, July2012.

[5] F. R. Yu, H. Tang, S. Bu, and D. Zheng, “Security and quality of service(QoS) co-design in cooperative mobile ad hoc networks,” EURASIP J.Wireless Commun. Networking, vol. 2013, pp. 188–190, July 2013.

[6] Y. Wang, F. R. Yu, H. Tang, and M. Huang, “A mean field game theoreticapproach for security enhancements in mobile ad hoc networks,” IEEETrans. Wireless Commun., vol. 13, pp. 1616–1627, March 2014.

[7] J. Chapin and V. W. Chan, “The next 10 years of DoD wirelessnetworking research,” in Proc. IEEE Milcom’11, (Baltimore, MD, USA),Nov. 2011.

[8] S. Bu, F. R. Yu, P. Liu, P. Manson, and H. Tang, “Distributed combinedauthentication and intrusion detection with data fusion in high-securitymobile ad hoc networks,” IEEE Trans. Veh. Tech., vol. 60, pp. 1025–1036, Mar. 2011.

[9] C. Adjih, D. Raffo, and P. Muhlethaler, “Attacks against OLSR: dis-tributed key management for security,” in Proc. 2nd OLSR Workshop,(Domaine de Voluceau, France), Dec. 2005.

[10] Y. Zhang, W. Liu, W. Lou, and Y. Fang, “Securing mobile ad hocnetworks with certificateless public keys,” IEEE Trans. Dependable andSecure Computing, vol. 3, pp. 386–399, Oct.–Dec. 2006.

[11] Y. Fang, X. Zhu, and Y. Zhang, “Securing resource-constrained wirelessad hoc networks,” IEEE Wireless Comm., vol. 16, no. 2, pp. 24–30, 2009.

[12] F. R. Yu, H. Tang, P. Mason, and F. Wang, “A hierarchical identitybased key management scheme in tactical mobile ad hoc networks,”IEEE Trans. on Network and Service Management, vol. 7, pp. 258 –267, Dec. 2010.

[13] S. Marti, T. Giuli, K. Lai, and M. Maker, “Mitigating routing misbe-havior in mobile ad hoc networks,” in Proc. ACM MobiCom’00, (NewYork, NY, USA), Aug. 2000.

[14] W. Lou, W. Liu, Y. Zhang, and Y. Fang, “SPREAD: improving networksecurity by multipath routing in mobile ad hoc networks,” ACM WirelessNetworks, vol. 15, no. 3, pp. 279–294, Mar. 2009.

[15] R. Zhang, Y. Zhang, and Y. Fang, “AOS: An anonymous overlay systemfor mobile ad hoc networks,” ACM Wireless Networks, vol. 17, no. 4,pp. 843–859, May 2011.

[16] P. Albers, O. Camp, J.-M. Percher, B. Jouga, and L. M. R. S. Puttini,“Security in ad hoc networks: a general intrusion detection architectureenhancing trust based approaches,” in Proc. 1st Int’l Workshop onWireless information Systems, (Ciudad Real, Spain), Apr. 2002.

[17] A. Mishra, K. Nadkarni, and A. Patcha, “Intrusion detection in wirelessad hoc networks,” IEEE Wireless Comm., vol. 11, pp. 48–60, Feb. 2004.

[18] S. Bu, F. R. Yu, X. P. Liu, and H. Tang, “Structural results for combinedcontinuous user authentication and intrusion detection in high securitymobile ad-hoc networks,” IEEE Trans. Wireless Commun., vol. 10,pp. 3064 –3073, Sept. 2011.

[19] S. Buchegger and J.-Y. L. Boudec, “A robust reputation system for P2Pand mobile ad-hoc networks,” in Proc. 2nd Workshop on the Economicsof Peer-to-Peer Systems, (Bologna, Italy), Nov. 2004.

[20] C. Zouridaki, B. L. Mark, M. Hejmo, and R. K. Thomas, “A quantita-tive trust establishment framework for reliable data packet delivery inMANETs,” in Proc. 3rd ACM Workshop on SASN’05, (Alexandria, VA,USA), Nov. 2005.

[21] S. Buchegger and J.-Y. L. Boudec, “Performance analysis of the confi-dant protocol,” in Proc. ACM MOBIHOC’02, (Lausanne, Switzerland),Jun. 2002.

[22] Y. Sun, W. Yu, Z. Han, and K. J. R. Liu, “Information theoreticframework of trust modeling and evaluation for ad hoc networks,” IEEEJ. Sel. Areas Commun., vol. 24, no. 2, pp. 305–317, 2006.

[23] J. Pearl, Probabilistic Reasoning in Intelligent Systems: Networks ofPlausible Inference. Morgan Kaufmann, 1988.

[24] G. Shafer and J. Pearl, Readings in Uncertain Reasoning. MorganKaufmann, 1990.

[25] B. Yu and M. P. Singh, “An evidential model of distributed reputationmanagement,” in Proc. ACM AAMAS’02, (Bologna, Italy), Jul. 2002.

[26] H. Wu, M. Siegel, R. Stiefelhagen, and J. Yang, “Sensor fusion usingDempster-Shafer theory,” in Proc. IEEE Instrumentation and Measure-ment Technology Conf., (Alaska, USA), May 2002.

[27] T. Clausen, C. Dearlove, and P. Jacquet, “The optimized link staterouting protocol version 2,” IETF draft-ietf-manet-olsrv2-13, Oct. 2011.

[28] J. H. Cho, A. Swami, and I. R. Chen, “A survey on trust management formobile ad hoc networks,” IEEE Communications Surveys and Tutorials,vol. 13, no. 4, pp. 562–583, 2011.

[29] T. M. Chen and V. Venkataramanan, “Dempster-Shafer theory forintrusion detection in ad hoc networks,” IEEE Internet Comput., vol. 9,no. 6, pp. 35–41, 2005.

[30] S. Bu, F. Yu, P. Liu, P. Mason, and H. Tang, “Distributed combinedauthentication and intrusion detection with data fusion in high securitymobile ad-hoc networks,” IEEE Trans. Veh. Tech., vol. 60, no. 3,pp. 1025–1036, 2011.

[31] R. Changiz, H. Halabian, F. R. Yu, I. Lambadaris, and H. Tang, “Trustestablishment in cooperative wireless relaying?networks,” Wireless Com-munications and Mobile Computing, 2012.

[32] H. Deng, Y. Yang, G. Jin, R. Xu, and W. Shi, “Building a trust-awaredynamic routing solution for wireless sensor neworks,” in Proc. IEEEGLOBECOM’10 Workshop on Heterogeneours, Multi-hop Wireless andMobile Networks, (Miami, FL, USA), Dec. 2010.

[33] S. Ganeriwal and M. B. Srivastava, “Reputation-based framework forhigh integrity sensor networks,” in Proc. ACM SASN’04, (Washington,D.C., USA), Oct. 2004.

[34] M. Raya, P. Papadimitratos, V. D. Gligor, and J.-P. Hubaux, “On data-centric trust establishment in ephemeral ad hoc networks,” in Proc. IEEEINFOCOM’08, (Phoenix, AZ, USA), Mar. 2008.

[35] R. Changiz, H. Halabian, F. R. Yu, I. Lambadaris, H. Tang, and P. Mason,“Trust establishment in cooperative wireless networks,” in Proc. IEEEMilcom’10, (San Jose, CA, USA), Nov. 2010.

[36] T. Clausen and P. Jacquet, “Optimized link state routing protocol(OLSR),” IETF RFC 3626, Oct. 2003.

[37] T. Clausen, C. Dearlove, and J. Dean, “Mobile ad hoc network(MANET) neighborhood discovery protocol (NHDP),” IETF RFC 6130,Apr. 2011.

[38] C. Adjih, T. Clausen, P. Jacquet, A. Laouiti, P. Muhlethaler, andD. Raffo, “Securing the OLSR protocol,” in Proc. 2nd IFIP AnnualMediterranean Ad Hoc Networking Workshop (Med-Hoc-Net 2003),(Mahdia, Tunisia), Jun. 2003.

[39] T. Clausen and U. Herberg, “Vulnerability analysis of the optimized linkstate routing protocol version 2 (OLSRv2),” in Proc. IEEE WCNIS’10,(Beijing, China), Feb. 2010.

[40] H. Deng, W. Li, and D. Agrawal, “Routing security in wireless ad hocnetworks,” IEEE Comm. Mag., vol. 40, no. 10, pp. 70–75, 2002.

[41] H. Yu, Z. Shen, C. Miao, C. Leung, and D. Niyato, “A survey ontrust and reputation management systems in wireless communications,”Proceedings of the IEEE, vol. 98, no. 10, pp. 1755–1772, 2010.

[42] N. Marchang and R. Datta, “Light-weight trust-based routing protocolfor mobile ad-hoc networks,” IET Inf. Secur., vol. 6, no. 2, pp. 77–83,2012.

[43] Y. Owada, T. Maeno, and H. Imai, “OLSRv2 implementation and perfor-mance evaluation with link layer feedback,” in Proc. ACM IWCMC’07,(Honolulu, Hawaii, USA), Aug. 2007.

[44] C. Perkins, E. Belding-Royer, and S. Das, “Ad hoc on-demand distancevector (AODV) routing,” IETF RFC 3561, Jul. 2003.

[45] “Trust platform module.” website: http://www.trustedcomputinggroup.org/.



12

[46] F. R. Yu, M. Huang, and H. Tang, “Biologically inspired consensus-based spectrum sensing in mobile ad hoc networks with cognitiveradios,” IEEE Network, vol. 24, pp. 26 –30, May 2010.

Zhexiong Wei received his B.S. degree in ComputerScience and Technology from Tianjin Institute ofTechnology, China and a M.E. degree in ComputerApplications from Tianjin University, China. Hecurrently is a PhD candidate with the Department ofSystems and Computer Engineering, Carleton Uni-versity, Canada. He also is a research assistant andteaching assistant in Carleton. His research interestsmainly include trust-based security in distributed en-vironments with uncertain reasoning and MANETsnetworking.

Helen Tang is a Defence Scientist at DefenceR&D Canada Ottawa. She received her Ph.D. degreein Electrical Engineering from Carleton University.From 1999 to 2005, she had worked in severalR&D organizations in Canada and USA includingAlcatel-Lucent, Mentor Graphics and Communica-tions Research Center Canada. In Oct. 2005, shejoined DRDC as a Defence Scientist. She has servedas technical program chair, reviewer and sessionchair for numerous conferences. She received theBest Paper Award at IEEE/IFIP TrustCom 2009 and

Outstanding Contribution Award at DRDC Ottawa in 2009. Her researchinterests include architectural and protocol design for computer networks,with a current focus on wireless network security. Helen is also an AdjunctProfessor at the Department of System and Computer Engineering of CarletonUniversity, where she is the supervisor of several graduate students.

F. Richard Yu (S’00-M’04-SM’08) received thePhD degree in electrical engineering from the Uni-versity of British Columbia (UBC) in 2003. From2002 to 2004, he was with Ericsson (in Lund,Sweden), where he worked on the research anddevelopment of wireless mobile systems. From 2005to 2006, he was with a start-up in California, USA,where he worked on the research and developmentin the areas of advanced wireless communicationtechnologies and new standards. He joined Car-leton School of Information Technology and the

Department of Systems and Computer Engineering at Carleton Universityin 2007, where he is currently an Associate Professor. He received theIEEE Outstanding Leadership Award in 2013, Carleton Research AchievementAward in 2012, the Ontario Early Researcher Award (formerly Premier’sResearch Excellence Award) in 2011, the Excellent Contribution Award atIEEE/IFIP TrustCom 2010, the Leadership Opportunity Fund Award fromCanada Foundation of Innovation in 2009 and the Best Paper Awards atIEEE Globecom 2012, IEEE/IFIP TrustCom 2009 and Int’l Conference onNetworking 2005. His research interests include cross-layer design, security,green IT and QoS provisioning in wireless networks.

Dr. Yu is a senior member of the IEEE. He serves on the editorial boards ofseveral journals, including IEEE Transactions on Vehicular Technology, IEEECommunications Surveys and Tutorials, ACM/Springer Wireless Networks,EURASIP Journal on Wireless Communications Networking, Ad Hoc andSensor Wireless Networks, Wiley Journal on Security and CommunicationNetworks, and International Journal of Wireless Communications and Net-working, and a Guest Editor for IEEE Transactions on Emerging Topics inComputing for the special issue on Advances in Mobile Cloud Computingand IEEE Systems Journal for the special issue on Smart Grid Commu-nications Systems. He has served on the Technical Program Committee(TPC) of numerous conferences, as the TPC Co-Chair of IEEE Globecom’14,WiVEC’14, INFOCOM-MCC’14, Globecom’13, GreenCom’13, CCNC’13,INFOCOM-CCSES’12, ICC-GCN’12, VTC’12S, Globecom’11, INFOCOM-GCN’11, INFOCOM-CWCN’10, IEEE IWCMC’09, VTC’08F and WiN-ITS’07, as the Publication Chair of ICST QShine’10, and the Co-Chair ofICUMT-CWCN’09.

Maoyu Wang holds B.E. and M.E. degrees inElectrical and Computer Engineering from SichuanUniversity, Chengdu, China and a M.Sc. degree inComputer Science from Carleton University, Ottawa,Canada. She has been working at CommunicationResearch Center Canada, Ottawa since 2011 andcurrently is a senior researcher at the researchcenter. Her research interests mainly focus on het-erogeneous network, machine-to-machine commu-nication, wireless cellular network, wireless sen-sor networks (WSNs) and mobile ad hoc networks

(MANETs) as well as security in WSNs and MANETs. She has contributedover 15 papers related mobile ad hoc networking and networking security.She has been a reviewer for many conference related to routing and securityof WSNs and MANETs.

Peter Mason received the B.Sc. degree in mathemat-ics and the M.Sc. and Ph.D. degrees in physics. Heis a Scientist and Head of the Cyber Operations andSignals Warfare Section with the Defence Researchand Development Canada (DRDC), Ottawa, ON,Canada. He is also an Adjunct Professor with theUniversity of Ottawa and the University of OntarioInstitute of Technology, Oshawa, ON, where he isthe supervisor of several graduate students.

security enhancements for mobile adhoc networks with trust management using uncertain reasoning

Documents