Research Article LSOT: A Lightweight Self-Organized Trust Model...

Research ArticleLSOT A Lightweight Self-Organized Trust Model in VANETs

Zhiquan Liu1 Jianfeng Ma12 Zhongyuan Jiang2 Hui Zhu2 and Yinbin Miao3

1School of Computer Science and Technology Xidian University Xirsquoan 710071 China2School of Cyber Engineering Xidian University Xirsquoan 710071 China3School of Telecommunication Engineering Xidian University Xirsquoan 710071 China

Correspondence should be addressed to Jianfeng Ma jfmamailxidianeducn

Received 23 June 2016 Accepted 13 November 2016

Academic Editor Elio Masciari

Copyright copy 2016 Zhiquan Liu et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

With the advances in automobile industry and wireless communication technology Vehicular Ad hoc Networks (VANETs) haveattracted the attention of a large number of researchers Trust management plays an important role in VANETs However it is still atthe preliminary stage and the existing trust models cannot entirely conform to the characteristics of VANETs This work proposesa novel Lightweight Self-Organized Trust (LSOT) model which contains trust certificate-based and recommendation-based trustevaluations Both the supernodes and trusted third parties are not needed in our model In addition we comprehensively considerthree factor weights to ease the collusion attack in trust certificate-based trust evaluation and we utilize the testing interactionmethod to build and maintain the trust network and propose a maximum local trust (MLT) algorithm to identify trustworthyrecommenders in recommendation-based trust evaluation Furthermore a fully distributed VANET scenario is deployed based onthe famous Advogato dataset and a series of simulations and analysis are conducted The results illustrate that our LSOT modelsignificantly outperforms the excellent experience-based trust (EBT) and Lightweight Cross-domain Trust (LCT) models in termsof evaluation performance and robustness against the collusion attack

1 Introduction

Nowadays an increasing number of vehicles are beingequipped with position and wireless communication deviceswhich forms an independent research area known asVANETs[1 2] Furthermore VANETs have become one of the mostprominent branches of Mobile Ad hoc Networks (MANETs)as they contribute to the increased road safety and passengercomfort [3ndash5]

In VANETs the participating nodes (ie vehicles) caninteract and cooperate with each other by exchanging mes-sages through nearby roadside units (ie vehicle to infras-tructure) and intermediate vehicles (ie vehicle to vehicle)[6] However due to the characteristics of VANETs namelybeing large open distributed highly dynamic and sparsethey are vulnerable to some malicious behaviors and attacks[7]

Traditional cryptography and digital signature technolo-gies mainly focus on ensuring the verifiability integrity and

nonrepudiation of messages among nodes and little concernshave been placed on evaluating the quality of messagesand nodes to deal with unreal information from maliciousnodes which may compromise VANETs [13 14] In factauthenticated nodes may also send out unreal information orcollude with others to cheat honest nodes for their own sake[15 16]

Trustmanagement plays a significant role inVANETs as itenables each node to evaluate the trust values of other nodesbefore acting on a message from other nodes for the purposeof avoiding the dire consequences caused by the unrealmessages frommalicious nodes [17] However recently only afew trust models in VANETs have been proposed [8 9 11 18ndash22] and they can be roughly divided into two categoriesnamely infrastructure-based and self-organized models [723]

Infrastructure-based trust models (as shown in Fig-ure 1(a)) [8 9 18 19] usually include the hierarchical Cer-tificate Authorities (CAs) which are supposed to be totally

Hindawi Publishing CorporationMobile Information SystemsVolume 2016 Article ID 7628231 15 pageshttpdxdoiorg10115520167628231

2 Mobile Information Systems

A

B C

a b c d

(a) Infrastructure-based models

a

b c

f

ed

(b) Self-organized models

Figure 1 Classic trust models in VANETs (where AsimC denote CAs and asimf represent vehicles)

trusted and able to satisfy a variety of security needs suchas authentication integrity nonrepudiation and privacyHowever this kind of trust models requires too strongassumption For example in these models the CAs must betotally trusted and online at all the time and every vehiclemust be able to access to the CAs at any time while in realitytheCAsmay break downor even colludewith somemaliciousvehicles to cheat other honest ones and the vehicles may notbe able to connect to the CAswhere the roadside units are notavailable (eg outside the city)

Since the self-organized models are more applicable tothe distributed and highly dynamic environment of VANETsmost of the recent trust models are built in this manner (asshown in Figure 1(b)) [11 20ndash22] In these models the CAsare not guaranteed at all the time and each node evaluatesthe trust value of target node based on the local knowledgeobtained from its past experiences and the recommendationsof neighbor nodes during a short period of time Though afew self-organized trust models have been proposed therestill exist the following drawbacks in them

(a) Due to the high dynamic characteristic VANETsare indeed temporary networks and the connectionsamong nodes are short-lived In most cases a nodewill not interact with the other same nodes more thanonce [24] As a result the self rsquos past experiences areusually not available for trust evaluation

(b) Most of the messages in VANETs are time-critical(eg the reports about traffic jams or accidents) andthe nodes need to evaluate their trust quickly anddecide whether to act on them or not while collectingthe trust recommendations requires large amounts oftime and bandwidth resources [12] which does notconform well to the natures of VANETs

(c) Though trust management can effectively detect themalicious nodes and false messages and promote thenode collaboration the trust model self may becomethe target of attacks such as the notorious collusionattack which is an open problem in the area of trustand reputation system [14] while the existing self-organized trust schemes rarely consider the robustnessagainst the collusion attack

To the best of our knowledge there is no existingdistinguished trust model for VANETs that has overcome allthe above limitationsThis is just the motivation of our workIn this paper we introduce the trust certificate [10 12] andtesting interaction [25 26] and propose a novel LSOT modelfor VANETs The major characteristics and contributions ofour proposed model are summarized as follows

(a) Our LSOT Model Is Built in a Lightweight and FullyDistributed Manner In our proposed model the nodes areself-organized and both the supernodes (eg the nodes withspecial roles) and trusted third parties (eg CAs) are notneeded Moreover as our LSOT model aggregates both trustcertificate-based and recommendation-based trust evalua-tions the evaluations in our model can be made quickly andreach an excellent performance in a lightweight manner

(b) Our LSOT Model Has High Evaluation Performance Todemonstrate the performance of our proposed model wedeploy a VANET scenario based on the noted Advogatodataset (httpkonectuni-koblenzdenetworksadvogato)and conduct a series of simulations and analysis Theresults demonstrate that our proposed model significantlyoutperforms the excellent EBT model [25] and LCT model[12] in terms of the evaluation performance

(c) Our LSOT Model Has Strong Robustness against theCollusion Attack In our LSOT model we adopt the testinginteraction method to build and maintain the trust recom-mendation network and combine trust certificate-based andrecommendation-based trust evaluationsThus our proposedmodel has stronger robustness against the collusion attackthan LCT model which has been verified by the simulationsand analysis

The rest of this paper is organized as follows Section 2includes some related work and its limitations Section 3demonstrates the motivation and general evaluation pro-cedure of our LSOT model and the trust certificate-basedand recommendation-based trust evaluations are detailed inSections 4 and 5 respectively Afterwards Section 6 intro-duces the aggregation evaluation method Comprehensivesimulations and analysis are presented in Section 7 andSection 8 concludes this paper

Mobile Information Systems 3

2 Related Work

In recent years a great deal of research work for VANETshas been done by utilizing digital signature and cryptog-raphy technologies Security and privacy have been widelyconcerned and the architectures challenges requirementsattacks and solutions in VANETs have been analyzed byseveral researchers [13 27ndash30] However these schemesmainly pay attention to ensuring the verifiability integrityand nonrepudiation of messages among nodes and littleconcerns have been placed on evaluating the quality ofmessages and nodes In actual fact an authenticated nodemay also send out unrealmessages for its own sake and otherscannot perceive them in advance

Trust management has been proved to be a very usefulsolution for themobile distributed environments as it enableseach node to evaluate the trust values of others in advanceso as to avoid interacting with malicious or selfish nodesA large number of trust models have been proposed forMANETs [31] Wireless Sensor Networks (WSNs) [32ndash34]and Mobile Peer to Peer networks (MP2Ps) [35] Howeverthese trust models are not suitable to VANETs due to theunique characteristics and requirements in this field

Currently trust management in VANETs is still at apreliminary stage and only a few trust models have been pro-posed These trust models can mainly be classified into twocategories namely infrastructure-based and self-organizedmodels

In the infrastructure-based schemes CAs are taskedwith maintaining the trust scores of vehicles Wu et al[18] proposed a Roadside-unit Aided Trust Establishment(RATE) model for VANETsThis model contains three prop-erties namely infrastructure-based architecture data-centricpattern and integration of observation and feedback Parket al [8] introduced a simple Long-Term Reputation (LTR)scheme based on the fact that plenty of vehicles have prede-fined constant daily trajectories In this model roadside unitsmonitor the daily behaviors of vehicles and update their rep-utation values To ensure the freshness of reputation scoresthe users have to query the roadside units frequently GomezMarmol and Martınez Perez [19] surveyed the deficiencyof existing trust models in VANETs and suggested a set ofdesign requirements for trust schemes which are specificallysuitable to VANETs Furthermore they also presented anoriginal Trust and Reputation Infrastructure-based Proposal(TRIP) from a behavioral perspective instead of an identity-based one Li et al [9] introduced a Reputation-basedGlobal Trust Establishment (RGTE) scheme in which thereputation management center is responsible for collectingthe trust information from all legal nodes and calculating thereputation scores of nodes

As we mentioned earlier the infrastructure-basedschemes require too strong assumptions and may lead tosome issues such as single point of failure and high mainte-nance cost Thus most of the recent trust models forVANETs are built in a self-organized manner Yang [20]proposed a novel Trust and Reputation Management Frame-work based on the Similarity (TRMFS) between messagesand between vehicles They also presented a similarity

mining technique to identify similarity and an updatingalgorithm to calculate the reputation values Bamberger etal [21] introduced an Inter-vehicular Communication trustmodel based on Belief Theory (ICBT) This model mainlyfocuses on the direct experiences among vehicles and utilizesbinary error and erasure channel to make a decision basedon the collected data Hong et al [22] noticed that VANETsface lots of situations and quickly change among differentsituations then they described a novel Situation-Aware Trust(SAT) model which includes three important componentsHuang et al [11] absorbed the Information Cascading andOversampling (ICO) into VANETs and proposed a novelvoting scheme in which each vote has different weight basedon the distance between sender and event

Though the above schemes provide many brilliant ideasthere exist several limitations as we analyzed earlier Inour previous work [12] we improved the classic CertifiedReputation (CR) model [10] and proposed a LCT model forthe mobile distributed environment In this model the trustcertificates are adopted as they can be carried by trustees andcontribute to establishing the trust relationships in highlydynamic environment in a fast and lightweight mannerHowever this model is intuitively vulnerable to the collusionattack In addition to tackle the sparse issue of VANETsMinhas et al [25] introduced a novel EBT scheme in whichthe vehicles send the testing requests to each other andinteractively compute the trust values of others based onthe quality of responses By this way a trust network canbe built and updated dynamically However the supernodeswith special roles are needed in this model thus in essencethis model is not built in a fully self-organized way

Aiming at building a lightweight trust model for VANETsin a fully self-organized way as well as overcoming thelimitations of aforementioned schemes we propose a novelLSOTmodel in this paper and the intuitive comparisons withsome other trust models are illustrated in Table 1

3 The Framework of Our LSOT Model

In this section we first show the motivation of our workwith a fully self-organized VANET scenario Afterwards weintroduce the general evaluation procedure in our proposedmodel through a simple example

31 The Motivation of Our Work Before introducing ourLSOT model we first illustrate our motivation with thefollowing VANET scenario (as demonstrated in Figure 2) Inthe past interactions (as shown in Figure 2(a)) the vehicleA interacted with several nearby vehicles (eg BsimF) andaccumulated certain trust level In a potential interaction (asshown in Figure 2(b)) A and its new neighbors (eg G) arestrange to each other Due to the highly dynamic feature ofVANETs the majority of previous interaction partners of A(eg B D and F) are far from G and there exists no reliabletrust path between them So G can merely collect the trustinformation aboutA from a few previous interaction partnersof A (eg C and E in fact they may not exist) and most ofprevious trust information of A (eg with B D and F) has tobe ignoredwhen building the new trust relationships between


Table 1 Intuitive comparisons between our LSOT model and some other trust models in VANETs

Trust models Architecture Trust certificate Recommendation Cost RobustnessLTR [8] Infrastructure-based times times High mdashRGTE [9] Infrastructure-based times times High mdashEBT [10] Self-organized with supernodes times radic Midterm mdashICO [11] Fully self-organized times times Low WeakLCT [12] Fully self-organized radic times Low WeakLSOT Fully self-organized radic radic Low Strongldquoradicrdquo support ldquotimesrdquo nonsupport ldquomdashrdquo without consideration

A

B C

D

EF

(a) Past interactions

A

C

G

E

(b) Potential interaction

Figure 2 Fully self-organized VANET scenario (where AsimG denote vehicles)

Trust certificate-based Recommendation-based

B

F

A G

C

E

TC(B A)

TC(F A)

MS(A)

TR(C A G)

TR(E A G)

Figure 3 An example of our LSOT scheme (where AsimG denote vehicles)

A and G As a result with the high-speed movement of A itstrust information is mostly discarded and rebuilt again andagain It is distinctly unreasonable and is just the motivationof this work How to utilize the previous trust information toquickly build the new trust relationships is the key focus ofthis paper

32 The Evaluation Procedure in Our LSOT Model To dealwith the above problem we propose a novel LSOTmodel anda simple example is illustrated in Figure 3 It is assumed thatprevious interactions occur between A and BsimF At the end ofpast interactions BsimF provide A with their trust certificates(ie TC(BA) sim TC(FA)) which are generated with digitalsignatures by BsimF Then A stores and updates the trustcertificates in its local storage In a potential interactionA canrelease a message (ie MS(A)) which includes six parts thatis the identification of A (ID) message type (MT) message

content (MC) trust certificates (TCs) timestamp (TS) anddigital signature (DS) to neighboring vehicles (eg G)WhenG receives the message it can check the authentication andintegrity of MS(A) through digital signature technology andcompute the trust certificate-based trust value of A accordingto the trust certificates Moreover G can also collect thetrust recommendations (eg TR(CAG) and TR(EAG))about A from its trustworthy neighbors (eg C and E) andthen derive the recommendation-based trust value of AAfterwards G can calculate the final trust value of A anddecide whether to trust the message content or not In theabove process A and G are defined as trustee and trustorrespectively BsimF are referred to as certifiers and C and E arecalled recommenders

Being consistent with the above example the generalevaluation procedure in our LSOT model is illustrated inFigure 4 Generally speaking it involves four kinds of roles


CertifiersTrustee

Time

TimeTrustor

Time

(b) Release amessage with TCs

(a) Provide TCs

Potentialinteraction

Pastinteractions

(d) Provide TRs

Recommenders(c) Request for TRs

Time

Figure 4 General evaluation procedure in our LSOT model

namely trustor (ie the receiver of message) trustee (ie thesender of message) certifier (ie the vehicle which providesthe trust certificate) and recommender (ie the vehiclewhich has past interactions with the trustee and provides thetrust recommendation to the trustor) Moreover it mainlyincludes four steps (a) At the end of past interactions thecertifiers provide their TCs to the trustee (b) In the beginningof a potential interaction the trustee can send out a messagewith TCs when needed (c) When the trustor receives thismessage it can derive the trust certificate-based trust value ofthe trustee based onTCs Besides it can also send the requeststo its trustworthy neighbors for TRs (d) The trustworthyrecommenders provide TRs to the trustor and then thetrustor can obtain the recommendation-based trust value ofthe trustee Afterwards the trustor can calculate the finaltrust value of the trustee and decide whether to trust themessage content from the trustee or not It should be notedthatwe donot distinguish between the trust value of node andthat of message in this paper aiming at building a lightweighttrust model for VANETs That is to say we utilize the trustvalue of a node to directly derive the trust value of messagesent by the node

In our proposed model the trust certificates for a nodeare stored by itself thus this part of trust information can becarried with the movement of node Furthermore the trustcertificates include the digital signatures and any change tothem can be easily detected [10 12] thus the node cannotmodify the trust certificates for self-praise Besides themessage is also attached with the digital signature thus itcannot be tampered even relayed by other nodes Benefitingfrom trust certificates the previous trust information can becarried and utilized to conduct the trust evaluation quickly ina fully self-organized way

4 Trust Certificate-Based Trust Evaluation

In this section we first introduce the formal representationsof trust certificate and message Moreover we comprehen-sively consider three factor weights that is number weighttime decay weight and context weight for trust certificateFinally we present the trust certificate-based trust calculationmethod in detail

41 The Formal Expressions of Trust Certificate and MessageIn our LSOT scheme the trust certificate generated bycertifier 119894 for trustee 119895 is denoted as

TC (119894 119895) = (ID (119894) ID (119895) TY (119894 119895) RV (119894 119895) LC (119894) TS (119894 119895) DS (119894 119895)) (1)

where ID(119894) and ID(119895) mean the identifications of certifier119894 and trustee 119895 respectively TY(119894 119895) denotes the type ofcorresponding message and RV(119894 119895) represents the ratingvalue which is a real number within the range of [0 1]Larger RV(119894 119895) means higher satisfaction degree and viceversa LC(119894) represents the location coordinate of certifier 119894and TS(119894 119895) denotes the timestamp when the trust certificateis generated DS(119894 119895) represents the digital signature Themessage released by trustee 119895 is denoted as

MS (119895)= (ID (119895) MY (119895) MC (119895) TCs (119895) TS (119895) DS (119895)) (2)

where ID(119895) denotes the identification of trustee 119895MY(119895) andMC(119895) stand for the type and content ofmessage respectivelyTCs(119895) denotes the set of trust certificates for trustee 119895 TS(119895)and DS(119895) represent the timestamp and digital signaturerespectively

42 Three Factor Weights for Trust Certificate Due to theunique feature of our LSOT scheme the trustee may merelyprovide profitable trust certificates to the potential trustoror even collude with others to improve its trust value andslander its competitors (ie collusion attack) Besides thetrustee may first accumulate high trust value through releas-ing authentic but unimportant (eg entertainment-related)message and cheat others by issuing important (eg security-related) but unreal message (ie value imbalance attack) Inorder to ease these two kinds of attacks we comprehensivelyconsider three factor weights that is number weight timedecay weight and context weight

421 Number Weight To balance the robustness againstcollusion attack and bandwidth consumption TCs(119895)merelyconsists of 119873(119895) (119873(119895) le 120578) most favorable trust certificateswhich come from diverse certifiers where 120578 is a systemparameter which relies on current network status in terms ofthe collusion attack The number weight WN(119895) correspond-ing to119873(119895) is denoted as a piecewise function [12]

WN (119895) = 0 if 119873(119895) lt 1205781 otherwise (3)

If 119873(119895) is less than 120578 the trust certificates are consideredincredible thus WN(119895) is set as 0 Otherwise the trustcertificates are viewed as reliable so WN(119895) is set as 1422 Time Decay Weight As we well know the relativelyrecent trust certificate is more convincing than the less recentone and the outdated trust certificate may be unreliable at


all as the behavior of trustee may change from honest tomalicious in VANETs thus the time decay weight WT(119894 119895)for TC(119894 119895) is denoted as [36]

WT (119894 119895) = 0 if TN minus TS (119894 119895) gt 120596119890minus(TNminusTS(119894119895))120572 otherwise (4)

where TN is the current timestamp and 120596 is a time window120572 is a time unit which controls the speed of time decayIf the time difference between TN and TS(119894 119895) exceeds 120596TC(119894 119895) is considered unreliable therefore WT(119894 119895) is set as0 Otherwise WT(119894 119895) is represented as an exponential decayfunction of time difference

423 Context Weight Last but not least we also take thecontext weight into account for TC(119894 119895) Specifically weconsider two kinds of most important contextual propertiesnamely message type and location

(a) Message Type As we mentioned earlier the node mayfirst accumulate high trust value through releasing authenticbut unimportant message and then cheat the other nodes byissuing important but unreal message (ie value imbalanceattack) thus we consider the message type similarity weightWY(119894 119895) for TC(119894 119895) as

WY (119894 119895) = 1 if 120588 (TY (119894 119895)) ge 120588 (MY (119895)) 120573 otherwise (5)

where 120588(lowast) is the importance function of message type and120573 is a constant within the range of [0 1) If the importance ofTY(119894 119895) is no less than that of MY(119895) TC(119894 119895) is consideredreliable andWY(119894 119895) is set as 1 Otherwise TC(119894 119895) is regardedas not entirely credible and WY(119894 119895) is set as 120573

(b) Location As discussed in some related work [1 7 14] thelocation is also an important contextual property In the viewof trustor a trust certificate from a nearby certifier is morereliable than that from a remote certifier as the latter has ahigher likelihood of colluding with trustee than the formerThus the location similarity weight WL(119894 119896) between trustor119896 and certifier 119894 is denoted as

WL (119894 119896)=

0 if LC (119894) minus LC (119896) gt 120575119890minusLC(119894)minusLC(119896)120582 otherwise

(6)

where 120575 is a distance threshold and 120582 is a constant whichcontrols the speed of distance decay If the distance betweencertifier 119894 and trustor 119896 exceeds 120575 TC(119894 119895) is viewed asunreliable thus WL(119894 119896) is set as 0 Otherwise WL(119894 119896) isdenoted as an exponential decay function of distance

43 Trust Calculation Method Next we detail the trustcertificate-based trust calculation method At the end ofeach past interaction the certifier (eg 119894) generated a trustcertificate (eg TC(119894 119895)) and sent it to trustee 119895When trustee119895 needs to release a message MS(119895) it first chooses119873(119895)mostadvantageous trust certificates from its local storage based ontheweighted rating valueRW(119894 119895) which can be derived from

RW (119894 119895) = RV (119894 119895) lowastWT (119894 119895) lowastWY (119894 119895) (7)

It should be noted that in VANETs the messages are usuallybroadcasted in a one-to-many manner thus RW(119894 119895) isindependent of WL(119894 119896) in our scheme

When trustor 119896 receives MS(119895) it can extract 119873(119895) trustcertificates and then calculate the trust certificate-based trustvalue CT(119895 119896) of MS(119895) as

CT (119895 119896) = sum120578119894=1 RV (119894 119895) lowastWT (119894 119895) lowast (WY (119894 119895) +WL (119894 119896))

2 lowast 120578 if 119873(119895) = 120578120583 otherwise (8)

If 119873(119895) equals 120578 the trust certificates are viewed as reliableand CT(119895 119896) is calculated as the weighted average value of120578 ratings which come from diverse certifiers Otherwise thetrust certificates are considered unreliable and CT(119895 119896) isset as a default low value 120583 (0 lt 120583 lt 1) From (8) wecan easily find that CT(119895 119896) falls in the range of 0sim1 Inactual fact newcomer trustees may have no sufficient trustcertificates andmalicious trusteesmay also act as newcomersand refuse to provide unfavorable trust certificates so theirtrust certificate-based trust values equal 1205835 Recommendation-Based Trust Evaluation

In this section we first present the formal representationof trust recommendation Next we introduce the formation

of trust network based on testing interactions Moreoverwe propose an effective MLT algorithm to identify all thetrustworthy recommenders and introduce the details ofrecommendation-based trust calculation method

51 The Formal Representation of Trust Recommendation Inour LSOT scheme the trust recommendation on trustee 119895generated by recommender 119897 for trustor 119896 is denoted as

TR (119897 119895 119896)= (ID (119897) ID (119895) ID (119896) RV (119897 119895 119896) DS (119897 119895 119896)) (9)

where ID(119897) ID(119895) and ID(119896) stand for the identificationsof recommender 119897 trustee 119895 and trustor 119896 respectively


(a) Initial stage

07

08 08

(b) Forming stage

06

09 08

09

05

09

09

03

07

(c) Dynamic update stage

Figure 5 Trust network formation based on testing interactions

RV(119897 119895 119896) denotes the rating value and DS(119897 119895 119896) representsthe digital signature

52 The Formation of Trust Network Due to the sparse andhighly dynamic characteristic there are no sufficient or long-term trust relationships among nodes in VANETs In orderto tackle this problem we introduce the idea of allowingnodes to send several testing requires (to which the sendershave known the corresponding solutions in advance) to eachother and calculate the trust values of receivers accordingto the accuracy and timeliness of responses Inspired by theprevious work [25 26] we adopt and improve the classicexperience-based trust evaluation scheme [37]

Let TV(119904 119903) isin [0 1] be the trust value demonstrating thesatisfaction degree of sender 119904 to the responses of receiver119903 If sender 119904 does not receive any response from receiver 119903TV(119904 119903) is set as 0 Whenever sender 119904 receives a responsefrom receiver 119903 it updates TV(119904 119903) based on the followingrules

(a) If sender 119904 is satisfied with the new response ofreceiver 119903 TV(119904 119903) increases asTV (119904 119903) larr997888 TV (119904 119903) + 120601 lowast (1 minus TV (119904 119903)) (10)

(b) Otherwise TV(119904 119903) decreases asTV (119904 119903) larr997888 TV (119904 119903) minus 120595 lowast TV (119904 119903) (11)

where 120601 and 120595 are the increment and decrement factorsrespectively and their ranges are (0 1) Moreover we set 120601 lt120595 due to the fact that trust is difficult to build up but easy todrop off

We can easily find that the experience-based trust isaccumulated and the trust values of nodes can be updatedrecursively as (10) and (11) Moreover the difficulty of theabove calculations is very small and each node can evaluatethe trust values of other nearby nodes easily through testinginteractions thus the trust network can be generated anddynamically updated in a lightweight manner A simpleexample is shown in Figure 5

53 Trust Calculation Method In recommendation-basedtrust evaluation only the ratings from trustworthy recom-menders are considered For identifying trustworthy recom-menders we propose a novel MLT algorithm (ie Algo-rithm 1) to calculate the maximum local trust values of all therecommenders in the view of trustor

As we know the trust network in VANETs has the highlydynamic characteristic and the reliability of trust evaluationwill be very low when the trust path is too long [38]Therefore we consider the trust decay in ourMLT algorithmSpecifically suppose 1199010 rarr 1199011 rarr sdot sdot sdot rarr 119901ℎ (where 1199010 =119896 119901ℎ = 119897 and recommender 119897 has previous interactionswith trustee 119895) is one of the optimal trust paths from trustor119896 to recommender 119897 then the maximum local trust valueMT(119896 119897) (ie M119879[119897] in Algorithm 1) of recommender 119897 fromthe perspective of trustor 119896 can be obtained from [39]

MT (119896 119897) = prodℎminus1119898=0TV (119901119898 119901119898+1)ℎ120579 if ℎ le MH0 otherwise (12)

where ℎ is the hop from trustor 119896 to recommender 119897 and 120579 is aparameter which controls the speed of trust decay If MT(119896 119897)reaches the trust threshold TH(119896) of trustor 119896 recommender119897 is viewed as trustworthy and vice versa Similarly we canobtain all the elements of trustworthy recommender setRS(119895 119896) and calculate the recommendation-based trust valueRT(119895 119896) of trustee 119895 in the view of trustor 119896 as [40]RT (119895 119896)

= sum119897isinRS(119895119896) RV (119897 119895 119896) lowastMT (119896 119897)

sum119897isinRS(119895119896)MT (119896 119897) if RS (119895 119896) = 0] otherwise

(13)

If RS(119895 119896) is not empty RT(119895 119896) is calculated as theweighted average value of ratings from all the trustworthyrecommenders Otherwise RT(119895 119896) is set as a default lowvalue ] (0 lt ] lt 1) From (10)sim(13) we can find that therange of RT(119895 119896) is also 0sim1


Input TN = (NDTV) MH 119896 120579 lowastTN Trust network based on testing interactions NDNode set TV Set of experience-based trust values among nodes MH Maximum allowable hop119896 Trustor 120579 Trust decay factor lowastOutput MT lowastMT Maximum local trust array of nodes in ND from the perspective of 119896 lowast(1) VN lArr 0 lowastVN Visited node set which is initialized to an empty set lowast(2) MT HP lowastHP Hop array of nodes in ND from the perspective of 119896 lowast(3) MT[119896] lArr 1 HP[119896] lArr 0(4) for each 119901 isin ND minus 119896 do(5) if TV(119896 119901) gt 0 then(6) MT[119901] lArr TV(119896 119901) HP[119901] lArr 1(7) else(8) MT[119901] lArr 0 HP[119901] lArr infin(9) end if(10) end for(11) Add 119896 into VN(12) while ND minus VN = 0 do(13) Choose the node (named 119901) with the maximum local trust value from ND minus VN(14) if HP[119901] lt MH then(15) for each 119902 isin ND minus VN minus 119901 do(16) if TV(119901 119902) gt 0 and MT[119901] lowast TV(119901 119902) lowast (HP[119901](HP[119901] + 1))120579 gt MT[119902] then(17) MT[119902] lArr MT[119901] lowast TV(119901 119902) lowast (HP[119901](HP[119901] + 1))120579 HP[119902] lArr HP[119901] + 1(18) end if(19) end for(20) end if(21) Add 119901 into VN(22) end while(23) return MT

Algorithm 1 Our MLT algorithm

6 Aggregation Trust Evaluation

As we mentioned earlier trust certificate-based and recom-mendation-based trust evaluations have diverse advantagesand weaknesses as follows

(a) Comparing to recommendation-based trust evalu-ation trust certificate-based one can be conductedin a more fast and lightweight manner (the detailedanalysis is provided in our previous work [12]) whileit is intuitively more vulnerable to the collusion attackas the certifiers are strange to the trustor inmost cases

(b) Recommendation-based trust evaluation seems tobe more credible than trust certificate-based oneas in the former only the ratings of trustworthyrecommenders are considered But collecting theopinions from trustworthy recommenders consumeslarge amounts of time and bandwidth resourcesespecially when MH is set as a relatively high value(eg 6)

Thus it is beneficial to aggregate these two kinds of trustevaluations to achieve the more accurate evaluation result Inour scheme the final trust value FT(119895 119896) of trustee 119895 in thesight of trustor 119896 is calculated as

FT (119895 119896) = 120591 lowast CT (119895 119896) + (1 minus 120591) lowast RT (119895 119896) (14)

where 120591 is a weight parameter within the range of [0 1]which controls theweights of two kinds of trust evaluations in

aggregation trust evaluation So the range of FT(119895 119896) is also0sim1 Specifically when 120591 equals 1 or 0 the aggregation trustevaluation reduces tomere trust certificate-based one ormererecommendation-based one respectively In other cases (ie0 lt 120591 lt 1) the aggregation trust evaluation falls in betweentrust certificate-based one and recommendation-based one

7 Simulations and Analysis

To demonstrate the performance of our LSOT model wepresent a series of simulations and analysis in this sectionSpecifically we first deploy a fully distributed VANET sce-nario based on the famous Advogato dataset Then we vali-date the variations of both average trust values and averageacceptance rates of three kinds of messages Moreover wecompare the evaluation performance of our proposed modelwith that of EBT and LCT models Finally we analyze andverify the robustness of our LSOTmodel against the collusionattack comparing to that of LCT model

71 Simulation Settings In this work the comprehensivesimulations are implemented by Java language on an Ubuntuserver with 283GHz CPU and 4G RAM In concrete termswe first deploy a fully distributed VANET scenario Thetrust recommendation network is built based on the famousAdvogato dataset which includes 6541 nodes and 51127directed edges (denoting three kinds of trust relationshipsamong nodes namely apprentice journeyer and master of


Table 2 Parameter settings in our simulations

Parameters Descriptions Values120578 Number threshold in (3) 20120596 Time window in (4) 100120572 Time decay factor in (4) 40120573 Constant in (5) 051205751 Distance threshold in (6) infin1205821 Distance decay factor in (6) infin120583 Default trust value in (8) 01120601 Increment factor in (10) 02120595 Decrement factor in (11) 031198721198672 Maximum allowable hop in Algorithm 1 3120579 Trust decay factor in (12) 05] Default trust value in (13) 01120591 Weight parameter in (14) 051As the nodes in Advogato dataset do not contain location information we set 120575 = infin and 120582 = infin in our simulations so as to ensureWL(119894 119896) equiv 12MH is set as a relatively low value (ie 3) due to the highly dynamic and time-critical features of VANETs

which corresponding trust values are 06 08 and 10 resp)The nodesrsquo trust thresholds are randomly generated Threekinds of different messages namely honest (ie authenticand helpful) general (ie authentic but valueless) andmalicious (ie unreal and harmful) messages are sent fromdifferent senders In each test a random node receives amessage from certain sender and evaluates its trust value byutilizing our LSOT scheme If the messagersquos derived trustvalue reaches the nodersquos trust threshold the node accepts thismessage and provides a new trust certificate to the senderaccording to its satisfaction degree to thismessage After eachtest the timestamp adds 1The parameters in our simulationsare set as illustrated in Table 2

72 Validating the Evaluation Performance In this part wemainly validate the average trust value variations of threekinds of messages in honest environment and we also revealthe variations of average acceptation rates of three kinds ofmessages In concrete terms we divide the 500 timesrsquo testsinto 5 equal intervals (ie I1simI5) and then calculate theaverage acceptation rate in each interval respectively Thesimulation is repeated 1000 times for each kind of messagesand average results are shown in Figures 6 and 7

We first analyze the variations of average trust values asshown in Figure 6 In the initial stage three kinds ofmessageshave the same trust value (ie 010) With the increase oftest times (0sim300 times) the average trust value of honestmessages rises rapidly from 010 to 064 due to their excellentquality while that of general messages grows slowly from 010to 036 Besides the average trust value ofmaliciousmessagesremains about the same at 010 on account of their terribleperformance In the latter tests (300sim500 times) all the threekinds of messages dynamically keep constant average trustvalues (ie 064 036 and 010 resp)

Next we analyze the variations of average acceptationrates as shown in Figure 7 In the first three intervals (ieI1simI3) the average acceptation rate of honest messages growsfrom2746 to 6301 and that of generalmessages rises from

0 50 100 150 200 250 300 350 400 450 5000

01

02

03

04

05

06

07

08

Test times

Aver

age t

rust

valu

e

HonestGeneralMalicious

Figure 6 Average trust value variations of three kinds of messagesin our LSOT model

1860 to 3649 while that of malicious messages basicallystays unchanged at 1143 In the latter intervals (ie I4 andI5) all the three kinds of messages almost maintain constantaverage acceptation rates (ie 6465 3740 and 1143resp)

As we know honest messages bring benefits and mali-cious messages mean risks thus the higher the average trustvalue and average acceptance rate of honest messages thebetter and the lower the average trust value and averageacceptance rate of malicious messages the better Thereforethe above results show that our LSOT model significantlyimproves the average trust value and average acceptance rateof honest messages without increasing the risks caused bymalicious messages


I1 I2 I3 I4 I50

10

20

30

40

50

60

70

Test times intervals

Aver

age a

ccep

tanc

e rat

e (

)


Figure 7 Average acceptation rate variations of three kinds ofmessages in our LSOT model

73 Comparing the Evaluation Performance In this simula-tion we mainly compare the evaluation performance of ourLSOT model with that of EBT and LCT models as they aresimilar to our model Moreover we deploy and necessarilymodify these two models in our VANET scenario As weknow the trust ranges in EBT and LCT models are [minus1 1]and [0 100] respectively different from that in our proposedmodel (ie [0 1]) thus they are all converted to [0 1] forcomparison Besides the role-based trust is removed fromEBTmodel as it is not consistent with the fully self-organizedwayThis simulation is also repeated 1000 times for each kindof messages in EBT and LCT models and the average resultsare shown in Figure 8Moreover we also compare the averageacceptation rates of honest and general messages in everyinterval (ie I1simI5) in three kinds of models as illustrated inFigure 9

We first analyze the average acceptation rate varia-tions of honest messages in three kinds of trust modelsas demonstrated in Figure 9(a) In the first interval (ieI1) LCT model has distinctly lower average acceptationrate (ie 1099) than EBT model (ie 3074) and ourLSOT model (ie 2746) It is because that LCT modelmerely includes trust certificate-based evaluation and thesenders of honest messages are not able to provide suffi-cient trust certificates to improve their own trust valueswhile EBT model has no restriction about the number ofrecommenders in recommendation-based trust evaluationand the average trust value of honest messages rises with theincreasing test times Our LSOT model absorbs the meritsof recommendation-based evaluation thus in I1 the averageacceptation rate in our LSOTmodel is greatly higher than thatin LCT model and slightly lower than that in EBT model

In the latter intervals (ie I2simI5) the average acceptationrate in EBTmodel rises slowly and then dynamically remainsat a distinctly lower rate (ie 3762) than that in LCTmodel (ie 6351) and that in LSOT model (ie 6410)

It is because EBT model only contains recommendation-based evaluation and a portion of recommenders cannot bereached within the maximum allowable hop (ie 3) while inLCT model the trust certificates are attached to the messagesand they contribute to improving the trust values of honestmessages Our LSOT model includes the trust certificate-based and recommendation-based trust evaluations thus inI2simI5 the average acceptation rate in our LSOT model isgreatly higher than that in EBT model and generally higherthan that in LCT model

Next we analyze the average acceptation rate variationsof general messages in three kinds of trust models as shownin Figure 9(b) In the first interval (ie I1) the averageacceptation rate in our LSOT model (ie 1860) is greatlyhigher than that in LCT model (ie 1098) and slightlylower than that in EBT model (ie 2241) In the latterintervals (ie I2simI5) the average acceptation rate in ourLSOT model rises rapidly and stays basically unchanged ata relatively higher rate (ie 3709) than that in EBT model(ie 2984) and LCT model (ie 3530) The detailedanalysis is omitted as it is similar to that of honest messages

Besides we analyze the average acceptation rate varia-tions of malicious messages in three kinds of trust models (asthe average acceptation rate of malicious messages in everymodel remains about the same as 1146 the comparisonchart is omitted for space limitation) In LCT model thesenders of malicious messages act as newcomers and refuseto provide any unfavorable trust certificates thus both theaverage trust value and average acceptation rate keep largelyconstant In EBT model due to the malicious behaviorsand ldquoreentryrdquo strategy [41] the average trust value andaverage acceptation rate of malicious messages also remainbasically unchanged Our LSOT model aggregates EBT andLCT models thus the average acceptation rate of maliciousmessages also remains largely untouched

Through the above analysis we can easily discover thatour LSOTmodel not only limits the risks caused bymaliciousmessages as well as EBT and LCT models do but also greatlyraises the average acceptation rate of honest messages andimproves that of general messages to some extent whencomparing to the other trust models Thus our LSOT modelhas better evaluation performance than EBT and LCTmodelsin general

74 Comparing the Robustness Characteristics In the pre-vious parts we mainly consider the performance of ourmodel in honest environment while in this part we focuson verifying and analyzing the robustness of our modelagainst the collusion attack through comparing to that of LCTmodel The comparison with EBT model is omitted as thereis no consideration of collusion attack in this model Dueto the distributed feature of VANETs malicious nodes maycollude with other nodes to raise their own trust values (ieballot stuffing) or slander their honest competitors (ie badmouthing) [42] which will bring risks to message receiversSo a good trust model for VANETs should be able to detectand filter them out

As we well know in the trust certificate-based trustevaluation the certifiers are strange to the active trustor while


I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


Aver

age a

ccep

tanc

e rat

e (

)


(a) EBT

I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90

Test times intervalsAv

erag

e acc

epta

nce r

ate (

)


(b) LCT

Figure 8 Average acceptation rate variations of three kinds of messages in EBT and LCT models

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(a) Honest

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(b) General

Figure 9 Average acceptation rate comparisons of honest and general messages in three kinds of trust models


0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100

Figure 10 Average acceptation rate comparisons of malicious messages with different PCC values

in the recommendation-based trust evaluation the recom-menders are trustworthy in the perspective of active trustorThus the certifiers have a higher likelihood of colludingwith malicious senders than the recommenders LCT modelmerely consists of the trust certificate-based trust evaluationthus it is intuitively vulnerable to the collusion attack Whileour LSOT model aggregates the trust certificate-based andrecommendation-based trust evaluations it has relativelystrong robustness against the collusion attack

Next we validate the above analysis through two sim-ulations in which the recommenders are assumed to betrustworthy and the certifiers may be collusive at a certainpercentage (eg 0 25 50 75 or 100)

741 Ballot Stuffing In this part we compare the robustnessagainst the ballot stuffing of our LSOT model with that ofLCT model In the ballot stuffing the collusive certifiersprovide favorable trust certificates with high rating values tomaliciousmessages in spite of their bad performance In eachsimulation we vary the Percentage of Collusive Certifiers(PCC) and then calculate the average trust value of malicious

messages in each case respectivelyThe simulation is repeated1000 times and the average results are illustrated in Figure 10

In the ideal case (ie PCC= 0) as shown in Figure 10(a)the variation curves of average trust values of maliciousmessages in two kinds of trust models are very close to eachotherWith the increase of PCC the curve in LCTmodel getssteeper and steeperwhile that in our LSOTmodel rises slowlyso the gap of two curves gradually grows In the extreme case(ie PCC = 100) as shown in Figure 10(e) the gap of twocurves reaches the maximum amount and the average trustvalue ofmaliciousmessage in our LSOTmodel is significantlylower than that in LCT model

As wementioned earlier the lower the average trust valueof malicious messages the better thus the above simulationand analysis results demonstrate that our LSOT model hasstronger robustness against the ballot stuffing than LCTmodel

742 Bad Mouthing In this part we validate the robustnessof our LSOT model against the bad mouthing throughcomparing to LCTmodel In the bad mouthing the collusive


0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100

Figure 11 Average acceptation rate comparisons of honest messages with different PCC values

certifiers provide adverse trust certificates with low ratingvalues to honest messages in spite of their good qualityIn each simulation we vary PCC and compute the averagetrust value of honest messages in each case respectively Thesimulation is also repeated 1000 times and average outputs aredemonstrated in Figure 11

In the ideal case (ie PCC = 0) as shown in Figure 11(a)the variation curve of average trust value of honest messagesin our LSOT model is approximately consistent with that inLCT model With the increase of PCC the curve growth inLCT model becomes slower and slower while that in ourLSOT model is relatively fast thus the gap of two variationcurves progressively grows In the extreme case (ie PCC =100) as shown in Figure 11(e) the gap of two curves is upto the maximum value and the average trust value of honestmessages in our LSOT model is greatly higher than that inLCT model

As mentioned earlier the higher the average trust valueof honest messages the better thus the above simulation andanalysis results illustrate that our LSOT model significantly

outperforms LCT model in terms of the robustness againstthe bad mouthing

8 Conclusion

In this work we have proposed a novel LSOT model inwhich both the supernodes and trusted third parties are notneeded for VANETs in a self-organized way It combinesboth trust certificate-based and recommendation-based trustevaluations thus the evaluation in it can bemade quickly andreaches an excellent performance in a lightweight mannerIn trust certificate-based trust evaluation we have compre-hensively considered three factor weights namely numberweight time decayweight and context weight to ease the col-lusion attack and make the evaluation result more accurateIn recommendation-based trust evaluation we have utilizedthe testing interactionmethod to build andmaintain the trustnetwork and proposed an effectiveMLT algorithm to identifytrustworthy recommenders Moreover we have deployed afully distributed VANET scenario based on the celebrated


Advogato dataset and conducted comprehensive simulationsand analysis The results illustrate that our LSOT modelgreatly overmatches the outstanding EBT and LCTmodels interms of both evaluation performance and robustness againstthe collusion attack

Competing Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by National High Technol-ogy Research and Development Program (863 Program)(no 2015AA016007) Key Program of NSFC Grant (noU1405255) Major Natural Science Foundation of China (no61370078) National Natural Science Foundation of China(no 61502375 no 61303218) Natural Science Basis ResearchPlan in Shaanxi Province of China (no 2016JQ6046) andScience and Technology Project of Shaanxi Province (no2016JM6007)

References

[1] J Zhang ldquoTrust management for VANETs challenges desiredproperties and future directionsrdquo International Journal of Dis-tributed Systems and Technologies vol 3 no 1 pp 48ndash62 2012

[2] R Lu X Lin T H Luan X Liang and X Shen ldquoPseudonymchanging at social spots an effective strategy for location pri-vacy in VANETsrdquo IEEE Transactions on Vehicular Technologyvol 61 no 1 pp 86ndash96 2012

[3] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[4] D Zelikman and M Segal ldquoReducing interferences inVANETsrdquo IEEE Transactions on Intelligent TransportationSystems vol 16 no 3 pp 1582ndash1587 2015

[5] S S Tangade and S S Manvi ldquoA survey on attacks security andtrust management solutions in VANETsrdquo in Proceedings of the4th International Conference on Computing Communicationsand Networking Technologies (ICCCNT rsquo13) pp 1ndash6 IEEETiruchengode India July 2013

[6] S Gillani F Shahzad A Qayyum and R Mehmood ldquoA surveyon security in vehicular ad hoc networksrdquo in CommunicationTechnologies for Vehicles pp 59ndash74 Springer Berlin Germany2013

[7] J Grover M S Gaur and V Laxmi ldquoTrust establishmenttechniques in VANETrdquo in Wireless Networks and SecuritySignals andCommunication Technology pp 273ndash301 SpringerBerlin Germany 2013

[8] S Park B Aslam and C C Zou ldquoLong-term reputation systemfor vehicular networking based on vehiclersquos daily commute rou-tinerdquo inProceedings of the 2011 IEEEConsumer Communicationsand Networking Conference (CCNC rsquo11) pp 436ndash441 Las VegasNev USA January 2011

[9] X Li J Liu X Li and W Sun ldquoRGTE a reputation-basedglobal trust establishment in VANETsrdquo in Proceedings of the5th IEEE International Conference on Intelligent Networking and

Collaborative Systems (INCoS rsquo13) pp 210ndash214 IEEE XirsquoanChina September 2013

[10] T D Huynh N R Jennings and N R Shadbolt ldquoCertifiedreputation how an agent can trust a strangerrdquo in Proceedingsof the 5th ACM International Joint Conference on AutonomousAgents and Multi-Agent Systems (AAMAS rsquo06) pp 1217ndash1224ACM May 2006

[11] Z Huang S Ruj M A Cavenaghi M Stojmenovic and ANayak ldquoA social network approach to trust management inVANETsrdquo Peer-to-Peer Networking and Applications vol 7 no3 pp 229ndash242 2014

[12] Z Liu J Ma Z Jiang and Y Miao ldquoLCT a lightweight cross-domain trust model for the mobile distributed environmentrdquoKSII Transactions on Internet and Information Systems vol 10no 2 pp 914ndash934 2016

[13] F Qu Z Wu F Wang and W Cho ldquoA security and privacyreviewofVANETsrdquo IEEETransactions on Intelligent Transporta-tion Systems vol 16 no 6 pp 2985ndash2996 2015

[14] J Zhang ldquoA survey on trust management for VANETsrdquo in Pro-ceedings of the 25th IEEE International Conference on AdvancedInformation Networking and Applications (AINA rsquo11) pp 105ndash112 Singapore March 2011

[15] J Jakubiak and Y Koucheryavy ldquoState of the art and researchchallenges for VANETsrdquo in Proceedings of the 5th IEEE Con-sumer Communications and Networking Conference (CCNCrsquo08) pp 912ndash916 IEEE Las Vegas Nev USA January 2008

[16] D Wang T Muller Y Liu and J Zhang ldquoTowards robust andeffective trust management for security a surveyrdquo in Proceed-ings of the 13th IEEE International Conference on Trust Securityand Privacy in Computing and Communications (TrustCom rsquo14)pp 511ndash518 IEEE Beijing China September 2014

[17] N J Patel and R H Jhaveri ldquoTrust based approaches for securerouting in VANET a surveyrdquo Procedia Computer Science vol45 pp 592ndash601 2015

[18] A Wu J Ma and S Zhang ldquoRATE a RSU-aided scheme fordata-centric trust establishment in VANETsrdquo in Proceedings ofthe 7th International Conference on Wireless CommunicationsNetworking and Mobile Computing (WiCOM rsquo11) pp 1ndash6 IEEEWuhan China September 2011

[19] F Gomez Marmol and G Martınez Perez ldquoTRIP a trust andreputation infrastructure-based proposal for vehicular ad hocnetworksrdquo Journal of Network and Computer Applications vol35 no 3 pp 934ndash941 2012

[20] N Yang ldquoA similarity based trust and reputation managementframework for VANETsrdquo International Journal of Future Gener-ation Communication and Networking vol 6 no 2 pp 25ndash342013

[21] W Bamberger J Schlittenlacher andKDiepold ldquoA trustmodelfor intervehicular communication based on belief theoryrdquo inProceedings of the 2nd IEEE International Conference on SocialComputing (SocialCom rsquo10) pp 73ndash80 IEEE MinneapolisMinn USA August 2010

[22] XHongDHuangMGerla andZCao ldquoSAT situation-awaretrust architecture for vehicular networksrdquo in Proceedings of the3rd International Workshop on Mobility in the Evolving InternetArchitecture MobiArchrsquo08 pp 31ndash36 USA August 2008

[23] Z Huang S Ruj M Cavenaghi and A Nayak ldquoLimitations oftrust management schemes in VANET and countermeasuresrdquoin Proceedings of the IEEE 22nd International Symposium onPersonal Indoor and Mobile Radio Communications (PIMRCrsquo11) pp 1228ndash1232 IEEE Toronto Canada September 2011


[24] P Wex J Breuer A Held T Leinmuller and L DelgrossildquoTrust issues for vehicular ad hoc networksrdquo in Proceedings ofthe IEEE 67th Vehicular Technology Conference (VTC rsquo08) pp2800ndash2804 Singapore May 2008

[25] U F Minhas J Zhang T Tran and R Cohen ldquoA multifacetedapproach to modeling agent trust for effective communicationin the application of mobile Ad Hoc vehicular networksrdquoIEEE Transactions on Systems Man and Cybernetics Part CApplications and Reviews vol 41 no 3 pp 407ndash420 2011

[26] C Chen J Zhang R Cohen and P-H Ho ldquoA trust mod-eling framework for message propagation and evaluation inVANETsrdquo in Proceedings of the 2nd International Conference onInformation Technology Convergence and Services (ITCS rsquo10) pp1ndash8 August 2010

[27] M Saini A Alelaiwi and A El Saddik ldquoHow close are we torealizing a pragmatic VANET solution A meta-surveyrdquo ACMComputing Surveys vol 48 no 2 article 29 2015

[28] K Zheng Q Zheng P Chatzimisios W Xiang and Y ZhouldquoHeterogeneous vehicular networking a survey on architecturechallenges and solutionsrdquo IEEE Communications Surveys andTutorials vol 17 no 4 pp 2377ndash2396 2015

[29] T Ma J Zhou M Tang et al ldquoSocial network and tag sourcesbased augmenting collaborative recommender systemrdquo IEICETransactions on Information and Systems vol E98D no 4 pp902ndash910 2015

[30] H Li D Liu Y Dai and T H Luan ldquoEngineering searchableencryption of mobile cloud networks when QoE meets QoPrdquoIEEE Wireless Communications vol 22 no 4 pp 74ndash80 2015

[31] A M Shabut K P Dahal S K Bista and I U AwanldquoRecommendation based trust model with an effective defencescheme forMANETsrdquo IEEE Transactions onMobile Computingvol 14 no 10 pp 2101ndash2115 2015

[32] J Jiang G Han F Wang L Shu and M Guizani ldquoAn efficientdistributed trust model for wireless sensor networksrdquo IEEETransactions on Parallel and Distributed Systems vol 26 no 5pp 1228ndash1237 2015

[33] J Shen H W Tan J Wang J W Wang and S Y Lee ldquoAnovel routing protocol providing good transmission reliabilityin underwater sensor networksrdquo Journal of Internet Technologyvol 16 no 1 pp 171ndash178 2015

[34] S Xie and Y Wang ldquoConstruction of tree network with limiteddelivery latency in homogeneous wireless sensor networksrdquoWireless Personal Communications vol 78 no 1 pp 231ndash2462014

[35] B Qureshi G Min and D Kouvatsos ldquoA distributed repu-tation and trust management scheme for mobile peer-to-peernetworksrdquo Computer Communications vol 35 no 5 pp 608ndash618 2012

[36] T D Huynh N R Jennings and N R Shadbolt ldquoAn integratedtrust and reputation model for open multi-agent systemsrdquoAutonomous Agents and Multi-Agent Systems vol 13 no 2 pp119ndash154 2006

[37] T Tran and R Cohen ldquoA reliability modelling based strategyto avoid infinite harm from dishonest sellers in electronicmarketplacesrdquo Journal of Business and Technology Special Issueon Business Agents and the SemanticWeb vol 1 no 1 pp 69ndash762005

[38] Z Liu J Ma Z Jiang Y Miao and C Gao ldquoIRLT integratingreputation and local trust for trustworthy service recommenda-tion in service-oriented social networksrdquo PLoS ONE vol 11 no3 Article ID e0151438 2016

[39] G Liu YWang andMAOrgun ldquoTrust transitivity in complexsocial networksrdquo in Proceedings of the 25th AAAI Conferenceon Artificial Intelligence (AAAI rsquo11) vol 11 pp 1222ndash1229 SanFrancisco Calif USA August 2011

[40] Y A Kim and H S Song ldquoStrategies for predicting local trustbased on trust propagation in social networksrdquo Knowledge-Based Systems vol 24 no 8 pp 1360ndash1371 2011

[41] A Joslashsang and J Golbeck ldquoChallenges for robust trust andreputation systemsrdquo in Proceedings of the 5th InternationalWorkshop on Security and Trust Management (SMT rsquo09) SaintMalo France September 2009

[42] S Liu H Yu C Miao and A C Kot ldquoA fuzzy logic basedreputation model against unfair ratingsrdquo in Proceedings of the12th International Conference on Autonomous Agents andMulti-Agent Systems (AAMAS rsquo13) pp 821ndash828 St Paul Minn USAMay 2013

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Distributed Sensor Networks


Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014


ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014


Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications


Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia


Biomedical Imaging


ArtificialNeural Systems

Advances in


RoboticsJournal of



Computational Intelligence and Neuroscience

Industrial EngineeringJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in



A

B C

a b c d

(a) Infrastructure-based models

a

b c

f

ed

(b) Self-organized models

Figure 1 Classic trust models in VANETs (where AsimC denote CAs and asimf represent vehicles)

trusted and able to satisfy a variety of security needs suchas authentication integrity nonrepudiation and privacyHowever this kind of trust models requires too strongassumption For example in these models the CAs must betotally trusted and online at all the time and every vehiclemust be able to access to the CAs at any time while in realitytheCAsmay break downor even colludewith somemaliciousvehicles to cheat other honest ones and the vehicles may notbe able to connect to the CAswhere the roadside units are notavailable (eg outside the city)

Since the self-organized models are more applicable tothe distributed and highly dynamic environment of VANETsmost of the recent trust models are built in this manner (asshown in Figure 1(b)) [11 20ndash22] In these models the CAsare not guaranteed at all the time and each node evaluatesthe trust value of target node based on the local knowledgeobtained from its past experiences and the recommendationsof neighbor nodes during a short period of time Though afew self-organized trust models have been proposed therestill exist the following drawbacks in them

(a) Due to the high dynamic characteristic VANETsare indeed temporary networks and the connectionsamong nodes are short-lived In most cases a nodewill not interact with the other same nodes more thanonce [24] As a result the self rsquos past experiences areusually not available for trust evaluation

(b) Most of the messages in VANETs are time-critical(eg the reports about traffic jams or accidents) andthe nodes need to evaluate their trust quickly anddecide whether to act on them or not while collectingthe trust recommendations requires large amounts oftime and bandwidth resources [12] which does notconform well to the natures of VANETs

(c) Though trust management can effectively detect themalicious nodes and false messages and promote thenode collaboration the trust model self may becomethe target of attacks such as the notorious collusionattack which is an open problem in the area of trustand reputation system [14] while the existing self-organized trust schemes rarely consider the robustnessagainst the collusion attack

To the best of our knowledge there is no existingdistinguished trust model for VANETs that has overcome allthe above limitationsThis is just the motivation of our workIn this paper we introduce the trust certificate [10 12] andtesting interaction [25 26] and propose a novel LSOT modelfor VANETs The major characteristics and contributions ofour proposed model are summarized as follows

(a) Our LSOT Model Is Built in a Lightweight and FullyDistributed Manner In our proposed model the nodes areself-organized and both the supernodes (eg the nodes withspecial roles) and trusted third parties (eg CAs) are notneeded Moreover as our LSOT model aggregates both trustcertificate-based and recommendation-based trust evalua-tions the evaluations in our model can be made quickly andreach an excellent performance in a lightweight manner

(b) Our LSOT Model Has High Evaluation Performance Todemonstrate the performance of our proposed model wedeploy a VANET scenario based on the noted Advogatodataset (httpkonectuni-koblenzdenetworksadvogato)and conduct a series of simulations and analysis Theresults demonstrate that our proposed model significantlyoutperforms the excellent EBT model [25] and LCT model[12] in terms of the evaluation performance

(c) Our LSOT Model Has Strong Robustness against theCollusion Attack In our LSOT model we adopt the testinginteraction method to build and maintain the trust recom-mendation network and combine trust certificate-based andrecommendation-based trust evaluationsThus our proposedmodel has stronger robustness against the collusion attackthan LCT model which has been verified by the simulationsand analysis

The rest of this paper is organized as follows Section 2includes some related work and its limitations Section 3demonstrates the motivation and general evaluation pro-cedure of our LSOT model and the trust certificate-basedand recommendation-based trust evaluations are detailed inSections 4 and 5 respectively Afterwards Section 6 intro-duces the aggregation evaluation method Comprehensivesimulations and analysis are presented in Section 7 andSection 8 concludes this paper


2 Related Work















A

B C

D

EF


A

C

G

E




B

F

A G

C

E

TC(B A)

TC(F A)

MS(A)

TR(C A G)

TR(E A G)







CertifiersTrustee

Time

TimeTrustor

Time


(a) Provide TCs


Pastinteractions

(d) Provide TRs


Time
























WL (119894 119896)=


(6)












TR (119897 119895 119896)= (ID (119897) ID (119895) ID (119896) RV (119897 119895 119896) DS (119897 119895 119896)) (9)



(a) Initial stage

07

08 08

(b) Forming stage

06

09 08

09

05

09

09

03

07
















(13)























0 50 100 150 200 250 300 350 400 450 5000

01

02

03

04

05

06

07

08

Test times

Aver

age t

rust

valu

e






I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)













I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


Aver

age a

ccep

tanc

e rat

e (

)


(a) EBT

I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


erag

e acc

epta

nce r

ate (

)


(b) LCT


I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(a) Honest

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(b) General



0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100










0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




2 Related Work















A

B C

D

EF


A

C

G

E




B

F

A G

C

E

TC(B A)

TC(F A)

MS(A)

TR(C A G)

TR(E A G)







CertifiersTrustee

Time

TimeTrustor

Time


(a) Provide TCs


Pastinteractions

(d) Provide TRs


Time
























WL (119894 119896)=


(6)












TR (119897 119895 119896)= (ID (119897) ID (119895) ID (119896) RV (119897 119895 119896) DS (119897 119895 119896)) (9)



(a) Initial stage

07

08 08

(b) Forming stage

06

09 08

09

05

09

09

03

07
















(13)























0 50 100 150 200 250 300 350 400 450 5000

01

02

03

04

05

06

07

08

Test times

Aver

age t

rust

valu

e






I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)













I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


Aver

age a

ccep

tanc

e rat

e (

)


(a) EBT

I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


erag

e acc

epta

nce r

ate (

)


(b) LCT


I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(a) Honest

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(b) General



0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100










0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in






A

B C

D

EF


A

C

G

E




B

F

A G

C

E

TC(B A)

TC(F A)

MS(A)

TR(C A G)

TR(E A G)







CertifiersTrustee

Time

TimeTrustor

Time


(a) Provide TCs


Pastinteractions

(d) Provide TRs


Time
























WL (119894 119896)=


(6)












TR (119897 119895 119896)= (ID (119897) ID (119895) ID (119896) RV (119897 119895 119896) DS (119897 119895 119896)) (9)



(a) Initial stage

07

08 08

(b) Forming stage

06

09 08

09

05

09

09

03

07
















(13)























0 50 100 150 200 250 300 350 400 450 5000

01

02

03

04

05

06

07

08

Test times

Aver

age t

rust

valu

e






I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)













I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


Aver

age a

ccep

tanc

e rat

e (

)


(a) EBT

I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


erag

e acc

epta

nce r

ate (

)


(b) LCT


I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(a) Honest

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(b) General



0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100










0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




CertifiersTrustee

Time

TimeTrustor

Time


(a) Provide TCs


Pastinteractions

(d) Provide TRs


Time
























WL (119894 119896)=


(6)












TR (119897 119895 119896)= (ID (119897) ID (119895) ID (119896) RV (119897 119895 119896) DS (119897 119895 119896)) (9)



(a) Initial stage

07

08 08

(b) Forming stage

06

09 08

09

05

09

09

03

07
















(13)























0 50 100 150 200 250 300 350 400 450 5000

01

02

03

04

05

06

07

08

Test times

Aver

age t

rust

valu

e






I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)













I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


Aver

age a

ccep

tanc

e rat

e (

)


(a) EBT

I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


erag

e acc

epta

nce r

ate (

)


(b) LCT


I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(a) Honest

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(b) General



0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100










0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in












WL (119894 119896)=


(6)












TR (119897 119895 119896)= (ID (119897) ID (119895) ID (119896) RV (119897 119895 119896) DS (119897 119895 119896)) (9)



(a) Initial stage

07

08 08

(b) Forming stage

06

09 08

09

05

09

09

03

07
















(13)























0 50 100 150 200 250 300 350 400 450 5000

01

02

03

04

05

06

07

08

Test times

Aver

age t

rust

valu

e






I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)













I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


Aver

age a

ccep

tanc

e rat

e (

)


(a) EBT

I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


erag

e acc

epta

nce r

ate (

)


(b) LCT


I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(a) Honest

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(b) General



0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100










0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




(a) Initial stage

07

08 08

(b) Forming stage

06

09 08

09

05

09

09

03

07
















(13)























0 50 100 150 200 250 300 350 400 450 5000

01

02

03

04

05

06

07

08

Test times

Aver

age t

rust

valu

e






I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)













I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


Aver

age a

ccep

tanc

e rat

e (

)


(a) EBT

I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


erag

e acc

epta

nce r

ate (

)


(b) LCT


I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(a) Honest

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(b) General



0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100










0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in
























0 50 100 150 200 250 300 350 400 450 5000

01

02

03

04

05

06

07

08

Test times

Aver

age t

rust

valu

e






I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)













I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


Aver

age a

ccep

tanc

e rat

e (

)


(a) EBT

I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


erag

e acc

epta

nce r

ate (

)


(b) LCT


I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(a) Honest

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(b) General



0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100










0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)













I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


Aver

age a

ccep

tanc

e rat

e (

)


(a) EBT

I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


erag

e acc

epta

nce r

ate (

)


(b) LCT


I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(a) Honest

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(b) General



0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100










0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


Aver

age a

ccep

tanc

e rat

e (

)


(a) EBT

I1 I2 I3 I4 I50

10

20

30

40

50

60

70

80

90


erag

e acc

epta

nce r

ate (

)


(b) LCT


I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(a) Honest

I1 I2 I3 I4 I50

10

20

30

40

50

60

70


Aver

age a

ccep

tanc

e rat

e (

)

EBTLCTLSOT

(b) General



0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100










0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100










0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(a) 0

0 200 4000

02

04

06

Test timesAv

erag

e tru

st va

lue

LCTLSOT

(b) 25

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(c) 50

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(d) 75

0 200 4000

02

04

06

Test times

Aver

age t

rust

valu

e

LCTLSOT

(e) 100






8 Conclusion




Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





Competing Interests


Acknowledgments


References




















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in






























Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in



Research Article LSOT: A Lightweight Self-Organized Trust Model...

Documents

Transcript of Research Article LSOT: A Lightweight Self-Organized Trust Model...