Trustworthy Event-Information Dissemination in Vehicular ...

Research ArticleTrustworthy Event-Information Dissemination inVehicular Ad Hoc Networks

Rakesh Shrestha and Seung Yeob Nam

Department of Information and Communication Engineering Yeungnam University 280 Daehak-Ro Gyeongsan-siGyeongsangbuk-do 712-749 Republic of Korea

Correspondence should be addressed to Seung Yeob Nam synamynuackr

Received 28 May 2017 Revised 1 October 2017 Accepted 17 October 2017 Published 12 November 2017

Academic Editor Francesco Gringoli

Copyright copy 2017 Rakesh Shrestha and Seung Yeob Nam This is an open access article distributed under the Creative CommonsAttribution License which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

In vehicular networks trustworthiness of exchanged messages is very important since a fake message might incur catastrophicaccidents on the road In this paper we propose a new scheme to disseminate trustworthy event information while mitigatingmessage modification attack and fake message generation attack Our scheme attempts to suppress those attacks by exchanging thetrust level information of adjacent vehicles and using a two-step procedure In the first step each vehicle attempts to determine thetrust level which is referred to as truth-telling probability of adjacent vehicles The truth-telling probability is estimated based onthe average of opinions of adjacent vehicles and we apply a new clustering technique to mitigate the effect of malicious vehicleson this estimation by removing their opinions as outliers Once the truth-telling probability is determined the trustworthiness ofa given message is determined in the second step by applying a modified threshold random walk (TRW) to the opinions of themajority group obtained in the first step We compare our scheme with other schemes using simulation for several scenarios Thesimulation results show that our proposed scheme has a low false decision probability and can efficiently disseminate trustworthyevent information to neighboring vehicles in VANET

1 Introduction

Vehicular networks are expected to be used for trafficcontrol accident avoidance parking management and soon [1] Communication security between vehicles needs tobe addressed carefully due to the safety requirements ofvehicular network applications [2] There is a lot of ongoingresearch on security topics which aims to provide securecommunications and verification of data to thwart maliciousattackers One of the major issues in vehicular ad hocnetwork (VANET) is message trust which can be used tosecure VANET communications It is essential to periodicallyevaluate the trustworthiness of event information based ontrustmetrics Generally trust computation in a static networkis relatively simple because the trust level can be calculatedbased on the behavior of the nodes with sufficient observa-tions [3] However message trust computation in VANETis challenging due to the ephemeral nature of the networktopology

The wireless access in vehicular environments (WAVE)protocol is based on the IEEE 80211p standard and pro-vides the basic radio standard for dedicated short rangecommunication (DSRC) operating in the 59GHz frequencyband [4] Vehicular communications can be achieved inthe infrastructure domain for vehicle-to-infrastructure (V2I)communications or in the ad hoc domain for vehicle-to-vehicle (V2V) communications We mainly focus on V2Vcommunications because road side units (RSUs) [1] may notbe available in some parts of the country during the initialstages of deployment of the vehicular communications infras-tructure Vehicles communicate with other vehicles throughon-board units (OBUs) forming mobile ad hoc networks thatallow communications in a completely distributed manner[5] We note that some event information (eg accidentreports) needs to be disseminated quickly and accuratelywith minimum delay Failure in timely and accurate dis-semination of such time-critical information might lead tocollateral damage to neighboring vehicles

HindawiMobile Information SystemsVolume 2017 Article ID 9050787 16 pageshttpsdoiorg10115520179050787

2 Mobile Information Systems

Some of the issues in vehicular networks include simplerouting problems and application-oriented problems likeSybil attacks and false data dissemination [6]The traditionalreputation systems may not work efficiently in vehicularnetworks [7] Public key infrastructure (PKI) may not beavailable everywhere during the initial stages of vehicularnetwork deployment around a country because some regionsmay not be covered due to deployment costs or budgetissues Generally cryptography-based verification ofmessagetrustworthiness is computationally expensive It can protectagainst a few types of attack from external nodes Howeverit will not protect against malicious nodes in the networkwhich already have the required cryptographic keys andmaynot be suitable for V2V ephemeral network communicationsOur scheme does not use cryptography and centralizedservers and thus it does not have a single point of failureMost VANET models assume that the system is up and run-ning where all vehicles have a certain trust score However itis not easy to know the trustworthiness of vehicles withouthaving had any interaction with those vehicles In highlydistributed vehicular networks vehicles can join and leavea network frequently [8 9] When a new vehicle joins thenetwork for the first time there is no information aboutit One of the challenges faced by VANET is that the trustmodel of the VANET should consider the requirement foranonymity of vehicles The trust model should have minimaloverhead in terms of computation complexity as well asstorage The trust model should be robust to data-centricattacks and be able to detect those attacks [10ndash12] VANETsecurity frameworks should be light scalable reliable andsecure

Our proposed scheme investigates the trustworthinessof event information received from adjacent vehicles whichserves as multiple pieces of evidence We use truth-tellingprobability as a measure for the trustworthiness of a vehicleThe vehicles communicate through safety messages to reportevents such as accident information safety warnings infor-mation on traffic jams weather reports and reports of ice onthe road In our proposed scheme all vehicles are assumedto have a pseudo identity (PID) which is independent ofthe node identity Each vehicle broadcasts an event messageto adjacent vehicles from the time it collects informationabout that event Every vehicle maintains the trust levelof its neighbors in a distributed manner to cope with thepropagation of false information We introduce an enhanced119870-means clustering technique to minimize the effect ofmalicious nodes on trust level calculationWe use a modifiedthreshold random walk algorithm with a single threshold tomake a final decision about the occurrence of an event whilesupporting real-time decision We focus on determining thetrustworthiness of the event information in the receivedmessages by considering reports from neighboring vehiclesdifferently with a truth-telling probability

The main contributions of our work can be summarizedas follows

(i) Our proposed scheme can contribute to dissemina-tion of trustworthy information since each vehicle

makes a decision on the trustworthiness of infor-mation in the received message individually whiledropping packets containing fake information

(ii) Since all the decisions are made based on the infor-mation received from the neighbor vehicles ourproposed scheme can work in an infrastructure-lessenvironment as well

(iii) Our proposed scheme can make a better decisionon the trustworthiness of a given message comparedto a simple voting mechanism since the modifiedthreshold random walk (TRW) can give a higherweight on the opinion of a vehicle whichmakes moretrue statements than false statements

The remainder of this paper is organized as follows InSection 2 we discuss the related work In Section 3 we pro-pose a trustworthy event-information dissemination schemefor VANET In Section 4 we evaluate the performance of theproposed scheme using simulation Section 5 concludes thepaper with future work

2 Related Work

Several trust management systems have been proposed forVANET [13ndash15] Trust management systems evaluate thetrust values of the neighbor nodes to prevent them frominteracting with the malicious nodes The authors in [16]provide a quantitative and systematic review of existing trustmanagement schemes for VANETThey address comprehen-sive trust model concepts problems and solutions related toVANET trust management There are several works on trustmanagement scheme based on infrastructure framework andcryptography techniques Trust management schemes can bedivided into four categories based on the use of infrastructureand cryptographicmeasures such as public key infrastructure(PKI) as shown in Table 1 The first category represents thetrust management techniques based on infrastructure suchas RSU and PKI In the second category the nodes rely oninfrastructure for trust management without using PKI Inthe third category each node handles the issue of messagetrustworthiness based on PKI without using any infrastruc-ture In the fourth category the nodes are fully decentralizedand operate in infrastructure-less environment and they donot depend on PKI

In the first category trust management systems are basedon infrastructure such as RSU and PKI and can be effectivein identifying malicious nodes with some accuracy [17ndash23]However trust management schemes in this category maynot work if infrastructure is not available The trust manage-ment based on PKI is computationally expensive and cannotsecure VANET against insider attack where the maliciousnodes have already acquired the cryptographic keys [5 23]Some researchers used group signatures (GS) techniques [17ndash21] to authenticate message sender and guarantee messageintegrity such as Identity-Based Group Signatures (IBGS) in[18] GSIS in [19] and Identity-based BatchVerification (IBV)in [20] However GS schemes are usually based on PKIandmessage sender authentication cannot prevent legitimatenodes from sending malicious messages

Mobile Information Systems 3

Table 1 Trust management category based on infrastructure and PKI

PKI based Non-PKI based

Infrastructure based DTE [5] ESA [17] IBGS [18] GSIS [19] IBV [20] EGSS[21] iTrust [22] PMBP [23] STRM [13] TRIP [24] RaBTM [25]

Non-infrastructure based ETM [26] MTM [27] LSOT [28] BTM [29] CTID [14] ITM [15] RMCV [30] ERM [31]VARS [32] TMEP [33] CRMS [34]

Trust management schemes in the second categoryrequire infrastructure including roadside unit or centralauthority without using PKI [13 24 25] Schemes such asTrust and Reputation Infrastructure-based Proposal (TRIP)[24] and road side unit (RSU) and Beacon-Based TrustManagement (RaBTM) [25] may not work if infrastruc-ture is not available In the third category the issue ofmessage trustworthiness was investigated based on PKI inan infrastructure-less environment [26ndash29] In [27] theauthors proposed a multidimensional approach for trustmanagement in decentralized VANET environments usingfour different types of roles for different types of vehiclesThe sender needs to authenticate itself to the receiver nodeusing PKI for verifying each otherrsquos role implemented in adistributed manner However VANET faces several issueswhile deploying the PKI scheme due to key distribution ina real world

In order to overcome the limitation of existingapproaches some researchers investigated trust managementwithout using PKI in an infrastructure-less environmentwhich corresponds to the fourth category [14 15 30ndash34]Trust management schemes in this category such as VehicleAd Hoc Reputation System (VARS) [32] are more suitablefor distributed VANET architecture Our proposed schemealso belongs to this category since the decision on thetrustworthiness of event information received from neighbornodes is made in an infrastructure-less environment withoutusing PKI

The existing trust management systems are establishedon specific application domain implementing different trust-based models to enhance intervehicular communicationThe trust-based models can be classified into three maincategories They are entity based data-centric based andhybrid trust models [35] Entity based trust model deals withthe trustworthiness of each node considering the opinions ofthe peer nodes [26ndash28 36] In [24] the authors proposed afuzzy approach for the verification of the trustworthiness ofthe nodes by using feedback from their neighbors Howeverthe trustworthiness of a message may not always agree withthe trustworthiness of the node itselfThus thismodel cannotresolve the issue of message trustworthiness properly

On the other hand in data-centric trust model thetrustworthiness of the reported events from the neighborvehicles is evaluated rather than the trust of the entities orthe node itself [5 30 31 35] In [5] the authors used aBayesian inference decision module to evaluate the receivedevent reports But the inference module uses the prior prob-ability which is not easy to obtain due to dynamic topologyof VANET In [28] the authors proposed a trust modelcalled a Lightweight Self-Organized Trust (LSOT) which

contains trust certificate-based and recommendation-basedtrust evaluations However it did not distinguish betweenthe trust value of a node and that of the reported messageThe trustworthiness of the nodes does not guarantee thetrustworthiness of the message as the trustworthy nodes cansend fake or faulty messages if attackers compromise themIn [30] the authors proposed Real-time Message ContentValidation (RMCV) scheme in an infrastructure-less modeThis scheme assigns a trust score to a received message basedon three metrics that is message content similarity contentconflict and message routing path similarity The messagetrustworthiness is based on the maximum value of finaltrust scores collected from the neighbor nodes However thisscheme does not consider highmobility of the vehicles and itstime complexity is high

Hence a hybrid trust model is introduced that combinesthe entity based and data-centric trust models to evaluatethe trustworthiness of a message [32ndash34] The authors in[29] proposed a hybrid trust management mechanism calledBeacon-based Trust Management (BTM) system which con-structs entity trust from beacon messages and computes datatrust from crosschecking the plausibility of event messagesand beacon messages However their trust model is basedon PKI and digital signature which incurs overhead whilesigning and authenticating each beacon message beforebroadcasting

Thus we attempt to overcome the limitation of theexisting schemes by improving the hybrid trust model formessage trustworthiness As a first step we use initialization-step enhanced 119870-means clustering algorithm (IEKA) forclustering of the vehicles into normal and malicious vehiclegroups to determine the trustworthiness of each neighbornode As a second step we use a modified threshold randomwalk (TRW) algorithm to decide the trustworthiness of agiven message Thus our scheme is based on a hybrid trustmodel AlthoughRMCV is based ondata-centric trustmodelit belongs to the fourth category that is trust manage-ment scheme that requires neither PKI nor infrastructureaccording to the classification in Table 1 Thus we compareour proposed scheme with the RMCV scheme The detailedcomparison and performance evaluation is discussed inSection 4

3 System Model

Each vehicle collects sufficient information to assess thevalidity and correctness of a message Notations explains theparameters and variables used in this paper

When an event occurs on the road a vehicle that is nearthat event sends the safety eventmessage119872119864 to neighboring


Driving direction

Vk

Vj

Vi Event

MF

ME

Figure 1 Trustworthy message dissemination scheme

vehicles Let us suppose that vehicle119881119895 wants to know the trueinformation about the event reported by vehicle119881119894 in Figure 1

The vehicle 119881119895 manages an information pair (pi 120579i) foreach neighbor vehicle 119881119894 where pi is the pseudo identityof the 119894th neighbor vehicle and 120579i is the trust level thatis truth-telling probability of vehicle 119881119894 We assume thatthe transportation authority preloads the pseudo ID ofthe vehicles during vehicle registration and it should berenewed periodically Tomaintain the privacy in VANET thepseudonym should change over time to achieve unlinkabilitythat protects the vehicle from location tracking Only privi-leged authorities are allowed to trace or resolve a pseudonymof the vehicle to a real identity under specific condition [37]The truth-telling probability (120579i) is the ratio of the numberof true event reports propagated by vehicle 119881119894 to the totalnumber of event reports sent by vehicle119881119894 over a specific timeperiod

31 Proposed Trustworthy Information Dissemination SchemeAn outline of the proposed scheme to determine the trust-worthiness of event information in the received messageis shown in Algorithm 1 The vehicle parameters such aspseudo ID (PID) and default trust level are initializedat the beginning All the vehicles periodically broadcastbeacon messages along with status information such asspeed and location to neighboring vehicles If there is noevent triggered then the vehicles will gather informationfrom the neighboring vehicles If a vehicle encounters anyevent by itself then it broadcasts a safety message alongwith the trust levels of neighboring vehicles that it knowsEach vehicle accumulates the trust levels of the neighboringvehicles based on the collected safety messages 119881119895 createsa trust matrix based on the trust level opinions from othervehicles Thus the trust matrix manages the trust levels ofeach neighboring vehicle Sometimes vehicles misbehave by

sending false information due to selfish motives like gettingeasier and faster access to the road or due to faults Toprevent such false information that can corrupt the trustlevel of legitimate vehicles we use a clustering algorithmOur proposed clustering algorithm attempts to separate thetrust level opinions of normal vehicles from the trust levelopinions of malicious vehicles The vehicle will calculate theaggregated trust level of adjacent vehicles belonging to themajority group of normal vehicles from the trust matrixIt will update the trust matrix using the average of trustlevels Then a modified TRW is applied to know whetherthe event has actually occurred or not The modified TRWcan provide better decision on the trustworthiness of anevent information by giving higher weights on the true eventmessages After the trustworthiness of the event informationhas been verified the event message is disseminated to otherneighboring vehicles along with the updated trust levels Ifthe event information contained in the message turns out tobe untrustworthy then the message is dropped

When new vehicles join the VANET they are not likelyto have enough information to infer the trust levels ofneighboring vehicles at the beginning We need a trust levelbootstrapping procedure to assign a default trust level forthis situation [38] The trust level that is the truth-tellingprobability ranges from 0 to 1 If vehicle 119860 does not have anyinformation on vehicle 119861 then the truth-telling probability ofvehicle119861 is set to 05 at vehicle119860We assume that each vehiclesets the truth-telling probability for itself to 1 by default

We mainly deal with two types of messages beaconmessages and safety messages The vehicles use beaconsto periodically broadcast and advertise status informationto neighboring vehicles at intervals of 100ms The senderreports its speed position and so on to neighboring vehicleswith beaconmessages via one-hop communications [39] Onthe other hand safety messages support vehicles on the road


The process is executed by a receiver vehicle upon receiving safety message119901119894119895 Pseudo ID of 119894th neighbor vehicle of 119881119895120579119895 truth-telling probability of 119881119895120579119894119895 estimator of 120579119894 by 119881119895Θ119895 trust level opinion generated by 119881119895119894 estimator for truth-telling probability of 119881119894Input Y = Θ119894 (119894 = 1 2 119899)Output 119894 updated trust matrix

(1) Information gathering from neighbor vehicles(2) If event is triggered then goto step (5)(3) Else goto step (2)(4) If event source is the 119881119895 itself then goto step (13)(5) Else 119881119895 accumulates the trust levels opinions of neighbor vehicles

Θ119895 = ((1199011119895 1205791119895) (1199012119895 1205792119895) (119901119895119895 120579119895119895) (119901119899119895 120579119899119895))(6) 119881119895 generates a trust matrix based on the trust level opinions(7) Use modified clustering algorithm to separate trust level opinions of normal from malicious vehicles(8) Calculate aggregated trust level of adjacent vehicles belonging to majority group from trust matrix

119894 = (1119899)sum119899119895=1 120579119894119895(9) Update the trust matrix(10) Use modified TRW to know if the event has actually occurred or not(11) If we decide that the event message is trustworthy then goto step (13)(12) Else drop the message(13) Broadcast safety message and trust level to neighboring vehicles

Algorithm 1 Determining trustworthiness of event information in the received message

by delivering time-critical information so that proper actioncan be taken to prevent accidents and to save people fromlife-threatening situations Safety messages include differenttypes of events 119864119909 such as road accidents traffic jamsslippery roads road constructions poor visibility due to fogand emergency vehicle warnings Vehicles broadcast a safetymessage to neighboring vehicles when they encounter eventson the road [1] The message payload includes informationabout the vehiclersquos position message sending time directionspeed and road events [19] Each vehicle gathers informationabout the neighboring vehicles within its communicationrange

One advantage of our proposed message disseminationscheme is to avoid a central trusted third party for trust accu-mulation in a distributed vehicular networking environmentWe consider VANET without infrastructure such as RSUsVehicles communicate with each other in V2V mode usingDSRC [40] This allows fast data transmission for criticalsafety applications within a short range of 250m A basicsafety application contains vehicle safety-related informationsuch as speed location and other parameters and thisinformation is broadcast to neighboring vehicles [41ndash43]Let us consider two vehicles 119881119894 and 119881119895 The truth-tellingprobability of 119881119894 depends on whether vehicle 119881119894 is truthfulwhen relaying event information According to Velloso et al[44] themore positive experiences vehicle119881119895 haswith vehicle119881119894 the higher the trust vehicle119881119895 will have towards vehicle119881119894

Let us suppose that vehicle 119881119894 has a pseudo ID 119901119894 andbroadcasts a safety warning message 119872119864 which is definedin (1) when event 119864119909 where 119909 represents an event typeis detected If the vehicle itself detects the event then it

broadcasts the safety message along with the trust levels ofneighboring vehicles If a vehicle receives a safety messagefrom other vehicles it will accumulate the safety messagealong with trust levels from neighboring vehicles Whenvehicle 119881119895 collects event information from vehicle 119881119894 it findsthe type and location of event from the message Let eventmessage 119872119864 be given by

119872119864 = (119901119894 119905 119871119864 119897119894) (1)

where 119901119894 is the pseudo ID of vehicle 119881119894 119905 is the messagegeneration time 119871119864 is the location of event 119864119909 and 119897119894 is thelocation of 119881119894 at time 119905

In addition to this each vehicle periodically broadcastsa beacon message defined as 119872119861 = (119901119894 119905119894 119897119894 119904119894) where 119901119894 isthe pseudo ID of 119881119894 119905119894 is the beacon generation time 119897119894 is thelocation of 119881119894 and 119904119894 is the speed of 119881119894

Let 120579119894 be the trust level that is truth-telling probability ofvehicle 119881119894 Truth-telling probability 120579119894 is defined as the ratioof the number of true events reported by vehicle 119881119894 dividedby the total number of events reported by vehicle 119881119894 over aspecific period of time Let119898 denote the total number of trueevents reported by 119881119894 and let 119899 denote the total number ofevents reported by the vehicle up to the current time Thenthe truth-telling probability is

120579119894 = 119898119899 (2)

A value for 120579119894 approaching 1 indicates reliable behaviorof the corresponding vehicle whereas a value close to zeroindicates a high tendency towards providing false informa-tion [45]


32 Calculation of Trust Level of Neighbor Vehicles When anevent occurs the nearby vehicles broadcast safety messageswith additional data such as pseudo IDs and truth-tellingprobabilities of other vehicles Based on the safety messagesfrom the neighboring vehicles trust matrix [120579119894119895] can beobtained where 120579119894119895 is estimation of 120579119894 by vehicle 119881119895 Thetrust matrix manages the truth-telling probability of eachneighboring vehicle from the viewpoint of other vehicles Weassume that each vehicle sets its own truth-telling probabilityto 1 If the trust matrix is constructed the aggregatedtrust level that is truth-telling probability of vehicle 119881119894 iscalculated from the trust matrix by

119894 = 1119899119899sum119895=1

120579119894119895 (3)

where 119894 is the estimator for the truth-telling probability of119881119894321 Estimation of Truth-Telling Probability Based on theCorrectness ofMessage Information If we can decide whetherspecific event information received from a vehicle is correctthis information can be used to estimate the truth-tellingprobability of the reporting vehicle more accurately Thereliable information about a specific event might be obtainedfrom direct observation of an event spot or announcementfrom a public and reliable group

We explain how the truth-telling probability can beestimated more accurately if we collect more evidence todecide the correctness of messages generated by a givenvehicle We can estimate the truth-telling probability 120579119894defined in (2) based on the correctness of recent119873messagesfrom 119881119894 We introduce a random variable 119883119899 to estimate thenumber of true reports among the recent 119873 reports from 119881119894on arrival of the 119899th report from 119881119894 Then the truth-tellingprobability of 119881119894 can be estimated by 119883119899119873 We attempt toestimate 119883119899 from 119883119899minus1 using the following relation119883119899

= 1 + (1 minus 1

119873)119883119899minus1 when the report 119899 is correct(1 minus 1

119873)119883119899minus1 otherwise(4)

Then we can show that 119883119899119873 approaches the truth-tellingprobability 120579119894 of119881119894 under the assumption that the correctnessof one message is independent of the correctness of othermessages By taking expectation on (4) we can obtain 119864[119883119899]as

119864 [119883119899] = Pr [message 119899 is correct]sdot 119864 [119883119899 | message 119899 is correct]+ Pr [message 119899 is incorrect]sdot 119864 [119883119899 | message 119899 is incorrect]

= 120579119894119864 [1 + (1 minus 1119873)119883119899minus1] + (1 minus 120579119894)

sdot 119864 [(1 minus 1119873)119883119899minus1] = 120579119894 + (1 minus 1

119873)119864 [119883119899minus1] (5)

By solving the recursive relation in (5) we can obtain

119864 [119883119899] = (1 minus 1119873)119899 119864 [1198830] minus 120579119894119873 + 120579119894119873 (6)

Thus regardless of the initial condition on 1198830 we havelim119899rarrinfin119864[119883119899] = 120579119894119873 and lim119899rarrinfin119864[119883119899119873] = 120579119894 from (6)In other words we can say that 119883119899119873 approaches the truth-telling probability 120579119894 asymptotically and we use the estimatorof 119883119899119873 and the relation in (4) to update the truth-tellingprobability of some vehicle whenever we have some evidenceto determine the correctness of a message from that vehicle

33 Clustering Algorithm If there is no evidence to deter-mine the truth of a given message then the truth-tellingprobability of vehicle119881119894 will be calculated using (3) Howevermalicious vehicles can modify the trust levels of neighboringvehicles to mislead vehicles in a vehicular network Thus weneed a clustering algorithm that can separate the trust levelsof normal vehicles from the trust levels of malicious vehiclesIt can reduce the effect of malicious vehicles on the trustlevels of normal vehicles In this subsection we propose a newclustering algorithm to tackle this issue

The main goal of our modified clustering algorithm isoutlier detection Our modified clustering algorithm classi-fies the trust level (truth-telling probability) opinions of thevehicles into two groups one with the trust level opinions ofnormal vehicles and the other with the trust level opinionsof malicious vehicles We will select the majority group andneglect the outliers corresponding to the minority group

Let us assume that an event has occurred on the road andthe vehicles near the event location send eventmessages alongwith trust level opinions to neighbor vehicles The vehicle 119881119895gathers reports about a specific event from neighbor vehiclesand manages the trust level opinions of other vehicles asfollows Each vehicle maintains a sorted vehicle list (SVL)which manages pseudo IDs of all the adjacent vehicles inan ascending order and the vehicle index will be assignedbased on the sequence in the sorted list as shown in Table 2Whenever a vehicle119881119895 needs to disseminate its own trust levelopinion to its neighbors it sends its trust level opinion Θ119895defined as

Θ119895 = ((1199011119895 1205791119895) (1199012119895 1205792119895) (119901119895119895 120579119895119895) (119901119899119895 120579119899119895)) (7)

where 119901119896119895 is the pseudo ID of the 119896th neighbor vehicle of thevehicle 119895 and 120579119895119895 is likely to be set to 1 because every node willtrust itself

If 119881119894 receives trust level opinion Θ119895 then 119881119894 updates itsown SVL by adding the vehicles that are inΘ119895 but are not inthe SVL After updating SVL119881119894 derives Θj1015840 from the receivedΘ119895 as

Θj1015840 = (1j1015840 2j1015840 1198951015840j1015840 1198991015840j1015840) (8)


Table 2 Example of sorted vehicle list (SVL)

Vehicle index Pseudo ID1 11352 20563 20794 21465 3012

where 1198951015840 is the new index of the vehicle 119895 according to itssequence in the updated SVL and 1198991015840 is the total number ofvehicles in the updated SVL When 119901119896119895 in the received Θ119895agrees with the 1198961015840th pseudo ID in the updated SVL 1198961015840j1015840 inΘj1015840 is updated as

1198961015840j1015840 = 120579119896119895 (9)

In this case 1198991015840 is always larger than or equal to 119899 since Θj1015840accommodates all the vehicles in Θ119895 If 1198991015840 gt 119899 then it meansthat there is some pseudo ID that is in the SVL but not inΘ119895If an index 119897 corresponds to such a pseudo ID lj1015840 will be setto 05 since the vehicle 1198951015840 does not know the vehicle 119897 If Θj1015840 isderived then the trust matrix table is updated by adding thetranspose of Θj1015840 as the 1198951015840th column

If each vehicle includes the pseudo IDs and the truth-telling probabilities of all the vehicles that it knows in the trustlevel opinionmessage defined in (7) then the traffic overheaddue to this message can be excessively large However we canreduce the message overhead by omitting trivial informationFor example if 120579119896119895 in Θ119895 is 05 this means that the vehicle119895 does not know the vehicle 119896 since 05 is the default valueused to initialize the truth-telling probability of a new vehicleIn this case the vehicle 119895 need not advertise this probabilitybecause this default value can be easily filled up by theneighbor vehicles according to the trust matrix updating rulementioned above with (7) (8) and (9) The vehicles on theroad are likely to be ignorant of each other in terms of thetrustmatrix table since they need not exchange the trust levelopinions if there is no event Thus we expect that the policyof omitting trivial information can significantly reduce trafficoverhead due to trust level opinion messages

If 119881119895 collects trust level opinions Θ119894 (119894 = 119895) from othervehicles along with event information then 119881119895 can constructtrust matrix Γ defined as

Γ =[[[[[[[

12057911 12057912 120579111989912057921 12057922 1205792119899 d

1205791198991 1205791198992 120579119899119899

]]]]]]]

(10)

We use a simple example to show our proposed clusteringalgorithm illustrated by Table 3 and Figure 2 Table 3 showsan example of trust matrix defined in (10) when 119899 = 3Three columns in Table 3 correspond to points 119860 119861 and

C

A

B

1

08

06

04

02

0

108

0604

020 1

08

06

04

02

0

3

1

2

Figure 2 Clustering example for the proposed clustering algorithm

Table 3 Trust matrix table of vehicle

119881119895 1198811198941 2 3

1 1 07 022 08 1 043 05 05 1

A B C

119862 in Figure 2 respectively The three probability values inthe first column of Table 3 correspond to 12057911 12057921 and 12057931respectively and these values are estimation of 1205791 1205792 and1205793 by 119881119894 This tuple of probability values is represented aspoint 119860 in the three-dimensional space of Figure 2 whereeach axis represents 1205791 1205792 and 1205793 respectively If there is noattacker that is all vehicles tell the truth then all the pointswill be close to each otherThe trust level opinions (Θ119894rsquos) withsimilar characteristics are likely to form the same cluster Ifthere is an attacker that tells a lie then the correspondingpoint will deviate from the majority group and this pointcan be distinguished as an outlier Even if the attacker triesto change or give higher trust levels by using collusion attackit can be detected as an outlier Let us suppose that point119862 represents an attacker that tells a lie by changing thetrust level as shown in Table 3 The clustering algorithm willseparate the trust level opinions into two groups one groupwith normal-vehicle trust levels (ie 119860 and 119861) and the othergroup with malicious vehicle trust levels (ie 119862) Figure 2describes the outcome of one possible clustering algorithmThe final aggregated trust level is calculated based on the trustlevel opinions corresponding to the majority group using (3)The final aggregated trust level based on the majority groupwill be used to update the trust level of the vehicle itselfThe resulting trust level is appended to the message duringmessage propagation


Input Y = Θ119894 (119894 = 1 2 119899)Output C = c119895 (119895 = 1 2)Initialize Calculate unique centroid as initial cluster center

120583 = sum119899119894=1Θ119894|Y|(1) for each cj isin C do(2) cj larr Θ119894 isin Y(3) c1 = argmax

Θ119894isinY120583 minusΘ119894

(4) c2 = argmaxΘ119894isinY

c1 minusΘ119894(5) end(6) While two centroids are not converged do(7) for each Θ119894 isin Y do(8) Assign Θ119894 to nearest centroid(9) cj = argmin

119895

cj minusΘ1198942(10) end(11) Update cluster centroid(12) Calculate new centroid cj as(13) for each cj isin C do(14) c119895 = (1|cj|) sum119894Θ119894isin119888119895 Θ119894(15) end(16) end

Algorithm 2 Proposed modified clustering algorithm

We propose a modified 119870-means clustering algorithmThe main problem with a 119870-means algorithm lies in theinitialization step so we introduce an enhanced 119870-meansclustering technique by modifying the initialization stepwhich is called initialization-step enhanced119870-means cluster-ing algorithm (IEKA) We use the IEKA to cluster the trustlevel opinions while reducing the effect of malicious vehicleson trust levels for other vehicles Our proposed clusteringalgorithm can be described in more detail as follows

After generating a trust matrix IEKA partitions the trustlevel opinions into 119870(le119899) groups C = c1 c2 c119896 Wedesignate Y to be a set of the Θi vectors that is Y =Θ1Θ2 Θ119899 We consider only two clusters for ourscheme Initially we take the mean of all the data points inY to find a unique centroid that is 120583

120583 = sum119899119894=1Θ119894|Y| (11)

We calculate the Euclidean distance between 120583 and eachvector inY Choose the point that has the maximum distancefrom the unique centroid that is the selected point is at thefarthest distance from the unique centroid We consider thispoint as the first centroid c1 for the first cluster

c1 = argmaxΘ119894isinY

1003817100381710038171003817120583 minusΘ1198941003817100381710038171003817 (12)

Similarly we compute the Euclidean distance betweenfirst centroid c1 and the remaining points in Y and select the

point with the maximum distance from c1 Then this pointbecomes the second centroid c2


1003817100381710038171003817c1 minusΘ1198941003817100381710038171003817 (13)

As a next step we run the conventional 119870-means clus-tering algorithm with c1 and c2 being the centroids of twoseparate groups Update centroids c1 and c2 by calculatingthe mean value for each group This gives new centroids c1and c2 and then reassigns each data point to the cluster towhich it is closest We will repeat this process until thosetwo centroids converge The proposed modified clusteringalgorithm is given in Algorithm 2

After clustering usingAlgorithm 2 which is based on (11)(12) and (13) the aggregated trust level of each neighbor vehi-cle is calculated based on the trust level opinions belonging tothe majority group We assume that the number of maliciousvehicles is less than that of normal vehicles The aggregatedtrust level is used for TRWcalculation In the next subsectionwe discuss the decision on the event based on hypothesistesting using TRW

34 Event Decision Based onThreshold RandomWalk (TRW)Sequential hypothesis testing is usually used to determineif a specific hypothesis is true or not based on sequentialobservations [46] Among the sequential hypothesis testingschemes threshold random walk has been used to detectscanners with a minimal number of packet observationswhile guaranteeing false positives and false negatives [47]Since we are interested in determining whether a givenmessage is true or not if true message constitutes one of


Table 4 Aggregated trust level table

Neighbor PID Trust level(truth-telling prob)

Event observations(119883119894)

1199011 1 1198831 = 11986411199012 2 1198832 = 11986411199013 3 1198833 = 1198641sdot sdot sdot sdot sdot sdot sdot sdot sdot119901119899 119899 119883119899 = 1198641

the two hypotheses then threshold random walk mightbe applied to this problem The threshold random walkscheme in [47] uses two thresholds that is one upper boundand one lower bound and the decision is made when alikelihood ratio reaches either threshold However in thisthreshold randomwalk scheme we cannot know the numberof samples required to reach either threshold in advanceThis means that real-time decisions may not be possible ifwe cannot collect a sufficient number of samples in a shortinterval In this paper we use a modified threshold randomwalk scheme to determine the validity of a given event whileresolving the issue of real-time decisionWe resolve this issueby applying threshold random walk with a single thresholdinstead of two thresholds Hereafter we explain the thresholdrandomwalk (TRW) scheme applied to our problem inmoredetail1198641 represents one of the events that can happen on aroad After clustering trust level opinions of neighbor vehiclesusing IEKA each vehicle determines the occurrence of event1198641 based on the aggregated trust level table The aggregatedtrust level table consists of vehicle PIDs aggregated trustlevels and event observations as shown in Table 4

In Table 4119883119894 is the report received from the 119894th neighborvehicle about event 1198641 Event 1198641 represents the occurrenceof an event and 1198641 represents nonoccurrence of event 1198641We need a rule to make a decision about the occurrence ofthe event We assume that119883119894rsquos are independent of each otheramong different vehicles For a given event suppose randomvariable 119884119894 can take only two values (0 and 1) that is

119884119894 = 0 if 119883119894 = 11986411 if 119883119894 = 1198641 (14)

After collecting a sufficient number of reports we wish todetermine whether the event (1198641) has really occurred usingsequential analysis [46]

Let us consider two hypotheses one is null and the otheris an alternate hypothesis (ie 1198670 and 1198671) where 1198670 isthe hypothesis that event 1198641 has occurred and 1198671 is thehypothesis that event 1198641 has not occurred that is 1198641 We alsoassume that conditionals on the hypothesis 119884 | 119867119895 where119895 = 0 1 are independent From the definition of the truth-telling probability and (14) we obtain

Pr (119884119894 = 0 | 1198670) = 120579119894Pr (119884119894 = 1 | 1198670) = 1 minus 120579119894

Pr (119884119894 = 0 | 1198671) = 1 minus 120579119894Pr (119884119894 = 1 | 1198671) = 120579119894

(15)

where Pr(119884 = 119896 | 119867119895) is the conditional probability that theobservation of 119884 given hypothesis 119867119895 is 119896 Then Pr(119884119894 =0 | 1198670) = 120579119894 becomes the truth-telling probability andPr(119884119894 = 1 | 1198670) = 1 minus 120579119894 becomes the lying probability Inorder to make a timely decision we collect report samplesfromneighbor vehicles during an interval of fixed duration119879Let 119873 denote the number of report samples collected duringthis interval Following the approach of Wald [46] we usecollected report samples to calculate the likelihood ratio by

Λ = 119873prod119894=1

Pr (119884119894 | 1198671)Pr (119884119894 | 1198670) (16)

Although the TRW scheme in [47] makes a decisionbased on two thresholds the upper and lower bounds we usea single threshold tomake a decisionwithout the issue of longwaiting time When the threshold is 120578 the decision rule is asfollows

If Λ ge 120578 then accept hypothesis 1198671If Λ lt 120578 then accept hypothesis 1198670

In this paper the threshold 120578will be set to 1 and the truth-telling probability 120579119894 of an unknown vehicle 119894 will be set to05 When a vehicle receives 119873 report messages if the 119873threport has come from a vehicle with no information on thetruth-telling probability Pr(119884119873 | 1198671) = Pr(119884119873 | 1198670) since120579119894 = 1 minus 120579119894 Thus the report from the unknown vehicle willnot affect the likelihood ratio by (15) and (16) Furthermore ifall the report messages are from the vehicles with no historyinformation then the likelihood ratio in (16) becomes 1 andthus it is fair to put 120578 = 1 since it is not easy to make adecision in this case

The advantage of our threshold random walk comparedto a simple voting scheme can be described with a simpleexample as follows Let us consider a case where an event1198641 is true and a vehicle receives 5 report messages Amongthem only two report that 1198641 is true and the other threeclaim that 1198641 did not happen If we make a decision basedon a simple voting then the decision will be 1198641 Howeverif we apply threshold random walk considering the truth-telling probability of each node the decision can be differentas follows If the truth-telling probability of the two nodesclaiming1198641 is 08 and the truth-telling probability of the threenodes claiming 1198641 is 06 then the likelihood ratio defined in(16) becomes

Λ (119883) = 0208 times 02

08 times 0604 times 06

04 times 0604 = 021

lt 1 (= 120578) (17)

Thus we will select the hypothesis 1198670 according to thedecision rule mentioned above since the likelihood ratiocalculated in (17) is less than the threshold 120578 This means


Accident occurred

Accident eventVn

Vm

Vk

Vj

Vj

Vi

Vl

No accident occurred message sent by Vk

(a)

Vn

Vm

Vk

Vj

ViVO

Vl

with traffic data Vk sends false message

direction due to Vk

Vm changes its

(b)

Figure 3 Two types of attack patterns considered in this paper (a) message modification attack and (b) fake message generation attack

the correct decision of 1198641 is made by the proposed thresholdrandom walk This advantage comes from the fact that thelikelihood ratio in (16) gives a higher weight to the opinion ofvehicles with a high truth-telling probability

After the decision on the actual occurrence of the eventis made vehicle 119895 will forward the received message to itsneighboring vehicles (with aggregated trust levels) withinradio range which is denoted by

119872119865 = (119901119895 119905119872119864Θ119895) (18)

where 119901119895 is the PID of vehicle 119895 which forwards the message119905 is the time at which 119872119865 was sent and Θ119895 denotes the trustlevel opinion of vehicle 119895 defined in (7)

35 Attack Model We consider two types of attacks messagemodification attack and fake message generation attack ina VANET environment Figure 3 shows an example of bothmessage modification and fake message generation attackA malicious vehicle might modify warning messages eitherwith malicious intent or due to an error in the communica-tions system In the message modification attack maliciousvehicles can modify message information at any time andfalsify the parameters

In Figure 3(a) when an accident event occurs on theroad the vehicles in an accident or the vehicles which are

close to that accident broadcast the accident event messageAfter vehicle 119881119895 sends an accident report to other vehiclesa malicious vehicle 119881119896 modifies the message and sends themodifiedno-accidentmessage as119872119865 defined in (18) with theintent to affect decisions taken by other vehicles Similarly ina fake message generation attack malicious vehicles generatea false warning message For example in Figure 3(b) [48]a malicious vehicle might send an accident message toneighboring vehicles even when there is no such event onthe road to clear the route it wants to take In this casethe malicious vehicle wants to convince other vehicles thatan event has occurred In this scenario the attacker mayhave already compromised one ormore vehicles and launchesattacks by generating a fakemessage for neighboring vehiclesWe assume that the number of malicious vehicles is less thanthe number of normal vehicles [12] In simulation we varythe number of malicious vehicles from 5 to 50 of overallvehicles to evaluate the performance of our proposed schemein an adversarial environment

4 Performance Evaluation

41 Simulation Setup The performance of our proposedscheme was evaluated through simulation We used theVehicles in Network Simulation (VEINS) framework version


Table 5 Simulation parameters

Parameters ValueNetwork simulation package OMNET++Vehicular traffic generation tool SUMOWireless protocol 80211pSimulation time 300 sScenario UrbanhighwayTransmission range 250m

4a2 [49] which is based on both OMNeT++ version 46 [50]a discrete event-driven network simulator and Simulation ofUrban Mobility (SUMO) version 22 for road traffic simula-tion [51] VEINS connects OMNeT++ and SUMO throughTraffic Control Interface (TraCI) VEINS provides realisticmodels for IEEE 80211p networks It provides OMNeT witha set of application programming interfaces to connect theSUMOplatformand to dynamically access information aboutSUMO simulated objects SUMO allows the creation ofscenarios that include realistic mobility patterns such asvehicle movement and overtaking as well as lane changing

We use the default map of Erlangen Germany from theVEINS framework with the map size of 2500m times 2500mfor our simulation We evaluated our scheme under differenttraffic densities to consider diverse situations When thevehicles reach the edge of the road the vehicles reroutetheir path and can meet other vehicles multiple times duringsimulation The number of vehicles increases linearly withtime from 0 s to 300 s The average vehicle speed changesfrom 40 kmh in an urban scenario to 110 kmh for highwayscenarios The key parameters considered in our simulationare summarized in Table 5

We considered two scenarios (urban and highway) byvarying parameters such as speed vehicle density andpercentage of malicious vehicles as shown in Table 6 Thenumber of malicious vehicles was varied considering themobility of vehicles in a realistic simulation environment byadjusting vehicle densities and vehicle speedsWe assume thatthe normal vehicles and the malicious vehicles are uniformlydistributed on the roads for each ratio of malicious vehicles[52]

42 Simulation Results In this section we analyze thesimulation results based on OMNet++ The traffic densityincreases from free-flow traffic (5 vehicleskm2) to congestedtraffic (100 vehicleskm2) where vehicles can meet multipletimes The simulation scenarios are summarized in Table 6For performance evaluation we have considered false deci-sion probability and message overhead We compared ourscheme with other schemes under different scenarios Inorder to evaluate our proposed scheme we considered themessage modification attack and the fakemessage generationattack one by one while increasing the number of maliciousvehicles from 5 to 50 in both scenarios The positions ofnormal vehicles and the initial distribution of the attackerswere randomly determined We calculated the average falsedecision probability by averaging the simulation results for

0 10 20155 25Total number of messages (N)

0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity

Figure 4 False decision probability versus total number ofmessages(119873)

30 simulation runs A decision is regarded as a false decisionwhen the decision result does not agree with the true statusof the event at the time of the decision In other words a falsedecision probability is the ratio of the number of incorrectdecisions to the total number of decisions

In order to update the truth-telling probability of vehicle119881119894 based on the truth of a given message according to (4) weneed to decide the parameter119873 that is the number of recentmessages from119881119894 that will be considered in this estimation Inorder to decide119873 we run 20 simulations under fake messageattack with 30 of malicious vehicles in a highway scenarioWe calculate the average false decision probability for variousvalues of 119873 and Figure 4 shows the result As the value of 119873increases the false decision probability tends to decreaseThefalse decision probability reaches zero at119873 = 15 and does notchange for larger values of 119873 Thus 119873 is fixed to 15 hereafterbased on this result

In Figure 4 the false decision probability is high for lowervalues of 119873 Let us consider an example to explain worseperformance for lower values of 119873 Let us take an extremecase of 119873 = 1 Then this means that when 119881119895 receives amessage from 119881119894 it decides the truth-telling probability of119881119894 only based on the last message since 119873 = 1 Thus if 119881119895finds that the last message from 119881119894 was false then 119881119895 willthink that the truth-telling probability of 119881119894 is 0 accordingto the updating rule described in (4) On the other hand if119881119895 finds that the last message from 119881119894 was true then 119881119895 willthink that the truth-telling probability of 119881119894 is 1 Thus theestimated truth-telling probability of each vehicle is either0 or 1 However if the truth-telling probability of a givenvehicle is different from 0 or 1 then this updating rule (with119873 = 1) will never find the accurate truth-telling probabilitysince the truth-telling probability is always 0 or 1 accordingto the updating rule Hence the truth-telling probabilitycan be significantly different from the correct truth-tellingprobability for lower values of 119873 especially when 119873 = 1

For our modified TRW scheme we need to determinean optimal value of message collection time 119879 to achieve agood decision accuracy We run several simulations underfake message generation attack with 30 of malicious vehiclein highway scenario We calculated the average false deci-sion probability against the message collection time underthe simulation parameters given in Table 5 In the sparsenetwork we received report message as low as five messages


Table 6 Simulation scenarios

Scenario number Type Total vehicles Malicious vehicle ratio Average speed Std of speed(1) Urban 100 0sim50 40 kmh 476(2) Highway 100 0sim50 110 kmh 752

0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)

Figure 5 False decision probability for various values of messagecollection time (119879)

with report collection time less than 100ms which results inhigh false decision probabilityThe simulation result is shownin Figure 5 As the report collection time interval increasesthe false decision probability decreases and from 800ms thefalse decision probability does not decrease anymore Basedon this we set the value of 119879 to 1 sec and this value will beused for 119879 hereafter

We now compare our proposed scheme with otherschemes RMCV scheme a simple voting scheme and TRW-only scheme RMCV is an information oriented trust modeland the outcome of the scheme is a trustworthiness valueassociated with each receivedmessage In RMCV scheme weconsider the message trustworthiness based on the contentsimilarity The message trustworthiness is likely to increaseas the message contents are similar among different vehiclesIn the TRW-only scheme we used the modified thresholdrandom walk to make a decision about the event in thewarning message without applying our proposed clusteringalgorithm Several voting methods have been proposed toestimate the trustworthiness of each report message [53ndash55]In the simple voting mechanism each vehicle collects a fixednumber of warning messages from the neighboring vehiclesregarding an event and makes a decision by following theopinion of the majority group [55] For the voting schemewe collected 15 messages to make a decision as this was theoptimal number according to our simulation

We compare our proposed scheme with other schemesin terms of false decision probability for various ratios ofmalicious vehicles under the message modification attackin a highway scenario as shown in Figure 6 We can seethat our proposed scheme yields a lower false decisionprobability compared to the other mechanisms even whenthe number ofmalicious vehicles increasesThe simple votingmechanism performs worst among the four schemes Theperformance of the RMCV scheme is close to TRW-onlyscheme when the malicious vehicle ratio is low However itdegrades significantly compared to our proposed scheme asthe malicious vehicle ratio increases Our proposed scheme

0

10

20

30

40

50

False

dec

ision

pro

babi

lity

TRW-only schemeProposed scheme

Voting schemeRCMV scheme

10 15 20 25 30 35 40 45 505Ratio of malicious vehicles ()

Figure 6 Comparison of the proposed scheme with other schemesunder a message modification attack in a highway scenario

25 35 45 503015 4010 205Ratio of malicious vehicles ()

0102030405060

False

dec

ision

pro

babi

lity



Figure 7 Comparison of the proposed scheme with other schemesunder a fake message generation attack in a highway scenario

has a false decision probability of 0 when the ratio ofmalicious vehicles is 30 in highway scenario

We compare our proposed scheme with other schemes interms of false decision probability under a fakemessage attackin a highway scenario as shown in Figure 7 We consider acasewhere the attacker generatesmessages about a fake eventOur proposed scheme yields better performance comparedto the RMCV simple voting and TRW-only schemes witha low false decision probability of less than 10 The falsedecision probability of RMCVand voting scheme exceed 40when the ratio of malicious vehicles increased to 50 InFigure 7 the false decision probability increases as the ratioof malicious vehicles increases with a tendency similar toFigure 6

We compare our proposed scheme with other schemesin terms of false decision probability under a messagemodification attack in an urban scenario in Figure 8 Ourproposed scheme exhibits better performance compared toother schemesThe false decision probability does not exceed





0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Figure 8 Comparison of the proposed scheme with other schemesunder a message modification attack in an urban scenario

2515 2010 30 35 40 45 505Ratio of malicious vehicles ()

0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity



Figure 9 Comparison of the proposed scheme with other schemesunder a fake message generation attack in an urban scenario

10 for our proposed scheme However it reaches 20 forthe TRW-only scheme The RMCV and the simple votingschemes exhibit much higher false decision probabilitiescompared to our proposed scheme

We compare our proposed scheme with other schemesin terms of false decision probability under a fake messageattack in an urban scenario in Figure 9 The false decisionprobability of the proposed scheme increases when thedensity of the malicious vehicles generating the false messageincreases resulting in a false decision probability slightlygreater than 10 In an urban scenario the high density ofvehicles and low speeds help the propagation of false eventmessages generated by attackers Thus the false decisionprobability in this case is slightly higher than that for thehighway scenario In this scenario the RMCV and the simplevoting schemes exhibit higher false decision probabilitiescompared to the proposed scheme with a tendency similarto Figure 8

We now compare our scheme with the RMCV schemein terms of message overhead We considered a situationwhere there is an actual accident without malicious vehiclesIn Figure 10 we present the simulation results of the messageoverhead with respect to the varying density of vehicles persquare kilometer in both urban and highway scenarios The

20 30 4010 60 70 80 90 10050Vehicle density (vehkＧ2)

0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)

RMCV scheme (urban)RMCV scheme (highway)

Proposed scheme (urban)Proposed scheme (highway)

Figure 10 Message overhead for various values of vehicle densitiesunder no malicious vehicle

message overhead is the cost incurred due to the extra mes-sage that is exchanged with neighboring vehicles In termsof message overhead as the vehicle density increases themessage overhead also increases in both scenarios as shownin Figure 10 In the beginning when the vehicle densityis low our proposed scheme has low message overhead asthe scheme does not advertise the default trust level of newneighbor vehicles however the pseudo IDs in the trust levelopinion pair cause some message overhead We can see thatthe message overhead is higher for the RMCV as comparedto our scheme in both scenarios because in their schemethe vehicle nodes send query messages to the neighboringvehicles and then receive response messages regarding theaccident event On the contrary there is no query messagebut only one-way reportmessages are sent in our schemeThemessage overhead in the urban scenario for both schemes isslightly higher than the highway scenario because the speedof the vehicles in the urban scenario is less than that ofhighway scenario and thus each vehicle accumulates moremessages compared to the highway scenario

We also compare our scheme with RMCV in terms ofmessage overhead in the presence ofmalicious vehicles underfake message attack The average vehicle density increasesfrom 1 vehicle per km2 to 100 vehicles per km2 throughoutthe simulation time in urban and highway scenarios We runseveral simulations by increasing the ratio of the maliciousvehicles from 5 to 50 in both scenarios Our scheme col-lects warning messages from neighboring vehicles to detectthe trustworthiness of event information contained in thereceived messages In both schemes the message overheadincreases as the ratio of the malicious vehicles increasesbecause the vehicles accumulate more messages due to thepresence of the malicious vehicles The message overhead fordifferent ratios of malicious vehicles is shown in Figure 11Our scheme has a lower message overhead compared to theRMCV scheme in both scenarios

5 Conclusion and Future Work

In this paper we proposed a trustworthy event-informationdissemination scheme in VANET We determine and dis-seminate only the trustworthy event messages to neighbor


0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)

Ratio of malicious vehicles ()10 15 205 30 35 40 45 5025



Figure 11 Message overhead for various ratios of malicious vehiclesunder fake message generation attack

vehicles We introduced a modified 119870-means clusteringalgorithm to reduce the effect of malicious vehicles onthe trust levels (ie the truth-telling probabilities) of othervehicles In other words the issue of node trustworthiness isresolved through a modified 119870-means clustering algorithmin our proposed scheme In the next step the issue ofmessage trustworthiness is resolved by applying a modifiedTRW to the report messages received from neighbor vehiclesalong with the information on node trustworthiness Wecompared our proposed scheme with RMCV simple votingand TRW-only schemes through simulation The simulationresults show that our proposed scheme has a lower falsedecision probability compared to other schemes as well aslow message overhead compared to the RMCV scheme Thesimulation results also show that our proposed scheme caneffectively cope with message modification attack and fakemessage generation attack as long as the number of benignvehicles is larger than the number of malicious vehiclesOur scheme has an additional advantage that the decisionon the trustworthiness of a given message is made in aninfrastructure-less environment without using PKI

In this paper we assumed that the malicious vehicles areuniformly distributed on the roadsHowever this assumptionmay not be valid if colluding malicious vehicles move as agroup to increase their influence on the nearby vehicles Sucha complicated issuewill be studied inmore detail in our futurework

Notations

119901119894 Pseudo ID of vehicle 119881119894120579119894 Trust level (truth-telling prob) of vehicle 119881119894119864119909 Event of type 119909119872119864 Event message119872119861 Beacon message119872119865 Forwarded message119871119864 Location of event 119864119909119897119894 Location of vehicle 119881119894119904119894 Speed of vehicle 119881119894120579119894119895 Estimation of trust level 120579119894 for vehicle 119894 by vehicle 119895

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This research was supported in part by Basic ScienceResearch Program through the National Research Founda-tion of Korea (NRF) funded by the Ministry of Education(2013R1A1A2012006 and 2015R1D1A1A01058595) and by theMSIT (Ministry of Science and ICT) Korea under the ITRC(Information Technology Research Center) support program(IITP-2017-2016-0-00313) supervised by the IITP (Institutefor Informationamp communications Technology Promotion)

References

[1] H Hartenstein and L Kenneth VANET Vehicular Applicationsand Inter-Networking Technologies Wiley New Jersey NJ USA1st edition 2009

[2] P Papadimitratos L Buttyan THolczer et al ldquoSecure vehicularcommunication systems design and architecturerdquo IEEE Com-munications Magazine vol 46 no 11 pp 100ndash109 2008

[3] K Govindan and P Mohapatra ldquoTrust computations andtrust dynamics in mobile adhoc networks a surveyrdquo IEEECommunications Surveys ampTutorials vol 14 no 2 pp 279ndash2982012

[4] Y L Morgan ldquoNotes on DSRC amp WAVE standards suite itsarchitecture design and characteristicsrdquo IEEECommunicationsSurveys amp Tutorials vol 12 no 4 pp 504ndash518 2010

[5] M Raya P Papadimitratos V D Gligor and J-P HubauxldquoOn data-centric trust establishment in ephemeral ad hocnetworksrdquo in Proceedings of the 27th IEEE CommunicationsSociety Conference on Computer Communications (INFOCOMrsquo08) pp 1912ndash1920 Arizona Ariz USA April 2008

[6] R Shrestha S Djuraev and S Y Nam ldquoSybil attack detectionin vehicular network based on received signal strengthrdquo inProceedings of the 3rd International Conference on ConnectedVehicles and Expo ICCVE rsquo14 pp 745-746 November 2014

[7] J Doucer ldquoThe Sybil Attackrdquo in Proceedings of the IPTPS rsquo01Revised Papers from the First International Workshop on Peer-to-Peer Systems 2002

[8] G Martuscelli A Boukerche and P Bellavista ldquoDiscoveringtraffic congestion along routes of interest using VANETsrdquo inProceedings of the 2013 IEEE Global Communications Confer-ence GLOBECOM rsquo13 IEEEAtlanta GAUSADecember 2013

[9] 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

[10] M Raya P Papadimitratos and J-P Hubaux ldquoSecuring vehicu-lar communicationsrdquo IEEEWireless CommunicationsMagazinevol 13 no 5 pp 8ndash15 2006

[11] Z Li and C T Chigan ldquoOn joint privacy and reputationassurance for vehicular ad hoc networksrdquo IEEE Transactions onMobile Computing vol 13 no 10 pp 2334ndash2344 2014

[12] F Ishmanov S W Kim and S Y Nam ldquoA secure trustestablishment scheme for wireless sensor networksrdquo Sensorsvol 14 no 1 pp 1877ndash1897 2014


[13] N Yang ldquoA similarity based trust and reputation managementframework for vanetsrdquo International Journal of Future Genera-tion Communication and Networking vol 6 no 2 pp 25ndash342013

[14] K Rostamzadeh H Nicanfar N Torabi S Gopalakrishnanand V C M Leung ldquoA context-aware trust-based informationdissemination framework for vehicular networksrdquo IEEE Inter-net of Things Journal vol 2 no 2 pp 121ndash132 2015

[15] R A Shaikh and A S Alzahrani ldquoIntrusion-aware trust modelfor vehicular ad hoc networksrdquo Security and CommunicationNetworks vol 7 no 11 pp 1652ndash1669 2014

[16] S A Soleymani A H Abdullah W H Hassan et al ldquoTrustmanagement in vehicular ad hoc network a systematic reviewrdquoEURASIP Journal onWireless Communications and Networkingvol 2015 no 1 article 146 2015

[17] M Raya A Aziz and J P Hubaux ldquoEfficient secure aggregationin VANETsrdquo in Proceedings of the 3rd ACM InternationalWorkshop on Vehicular Ad Hoc Networks (VANET rsquo06) pp 67ndash75 ACM New York NY USA September 2006

[18] B Qin Q Wu J Domingo-Ferrer and L Zhang ldquoPreservingsecurity and privacy in large-scale VANETsrdquo in Proceedingsof the 13th International Conference Information and commu-nications security (ICICS rsquo11) vol 7043 of Lecture Notes inComputer Science pp 121ndash135 Springer Berlin HeidelbergBerlin Germany 2011

[19] X Lin X Sun P-H Ho and X Shen ldquoGSIS a secure andprivacy-preserving protocol for vehicular communicationsrdquoIEEE Transactions on Vehicular Technology vol 56 no 6 I pp3442ndash3456 2007

[20] C Zhang R Lu X Lin P-H Ho and X Shen ldquoAn efficientidentity-based batch verification scheme for vehicular sensornetworksrdquo in Proceedings of the 27th IEEE CommunicationsSociety Conference on Computer Communications (INFOCOMrsquo08) pp 816ndash824 IEEE INFOCOM Arizona Ariz USA April2008

[21] A Wasef and X Shen ldquoEfficient group signature schemesupporting batch verification for securing vehicular networksrdquoin Proceedings of the IEEE International Conference on Commu-nications (ICC rsquo10) pp 1ndash5 Cape Town South Africa May 2010

[22] H Zhu S Du Z Gao M Dong and Z Cao ldquoA probabilisticmisbehavior detection scheme toward efficient trust establish-ment in delay-tolerant networksrdquo IEEE Transactions on Paralleland Distributed Systems vol 25 no 1 pp 22ndash32 2014

[23] D Tian Y Wang H Liu and X Zhang ldquoA trusted multi-hop broadcasting protocol for vehicular ad hoc networksrdquoin Proceedings of the 2012 1st International Conference onConnected Vehicles and Expo ICCVE rsquo12 pp 18ndash22 December2012

[24] 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

[25] Y-C Wei and Y-M Chen ldquoAn efficient trust management sys-tem for balancing the safety and location privacy in VANETsrdquoin Proceedings of the 11th IEEE International Conference onTrust Security and Privacy in Computing and Communications( TrustCom rsquo12) pp 393ndash400 Liverpool UK June 2012

[26] U F Minhas J Zhang T Tran and R Cohen ldquoTowardsexpanded trust management for agents in vehicular ad-hocnetworksrdquo International Journal of Computational IntelligenceTheory and Practice vol 5 no 1 pp 3ndash15 2010

[27] 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

[28] Z Liu JMa Z Jiang H Zhu andYMiao ldquoLSOT a lightweightself-organized trust model in VANETsrdquo Mobile InformationSystems vol 2016 no 1 Article ID 7628231 pp 1ndash15 2016

[29] Y-M Chen and Y-C Wei ldquoA beacon-based trust managementsystem for enhancing user centric location privacy in VANETsrdquoJournal of Communications and Networks vol 15 no 2 ArticleID 6512239 pp 153ndash163 2013

[30] S Gurung D Lin A Squicciarini and E BertinoldquoInformation-oriented trustworthiness evaluation in vehicularad-hoc networksrdquo Lecture Notes in Computer Science (includingsubseries Lecture Notes in Artificial Intelligence and LectureNotes in Bioinformatics) Preface vol 7873 pp 94ndash108 2013

[31] Q Ding X Li M Jiang and X Zhou ldquoReputation-basedtrust model in vehicular ad-hoc networksrdquo in Proceedings ofthe International Conference on Wireless Communications AndSignal Processing (WCSP) 2010

[32] F Dotzer L Fischer and P Magiera ldquoVARS a vehicle ad-hoc network reputation systemrdquo in Proceedings of the 6th IEEEInternational Symposium on a World of Wireless Mobile andMultimedia Networks WoWMoM rsquo05 pp 454ndash456 June 2005

[33] 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 Conferenceon Information Technology Convergence and Services (ITCS rsquo10)IEEE Cebu Phillippines August 2010

[34] A Patwardhan A Joshi T Finin andY Yesha ldquoA data intensivereputationmanagement scheme for vehicular ad hoc networksrdquoin Proceedings of the 2006 3rd Annual International Conferenceon Mobile and Ubiquitous Systems Networking and ServicesMobiQuitous July 2006

[35] 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 Biopolis Singapore March 2011

[36] PWex J Breuer AHeld T Leinmuller and LDelgrossi ldquoTrustissues for vehicular ad hoc networksrdquo in Proceedings of theIEEE 67th Vehicular Technology Conference-Spring (VTC rsquo08)pp 2800ndash2804 Singapore May 2008

[37] J Petit F Schaub M Feiri and F Kargl ldquoPseudonym schemesin vehicular networks a surveyrdquo IEEE Communications Surveysamp Tutorials vol 17 no 1 pp 228ndash255 2015

[38] Z Malik and A Bouguettaya ldquoReputation bootstrapping fortrust establishment among web servicesrdquo IEEE Internet Com-puting vol 13 no 1 pp 40ndash47 2009

[39] W R J Joo and D S Han ldquoAn enhanced broadcastingscheme for IEEE 80211p according to lane traffic densityrdquo inProceedings of the 20th International Conference on SoftwareTelecommunications and Computer Networks (SoftCOM rsquo12)September 2012

[40] W Fehr Security system design for cooperative vehicleto-vehiclecrash avoidance applications using 59 GHz Dedicated ShortRange Communications (DSRC) wireless communications 2012

[41] H Xiong K Beznosov Z Qin and M Ripeanu ldquoEfficient andspontaneous privacy-preserving protocol for secure vehicularcommunicationrdquo in Proceedings of the 2010 IEEE InternationalConference on Communications ICC 2010 May 2010


[42] F Jimenez J E Naranjo and O Gomez ldquoAutonomousmanoeu-vring systems for collision avoidance on single carriagewayroadsrdquo Sensors vol 12 no 12 pp 16498ndash16521 2012

[43] X Tang D Hong and W Chen ldquoData Acquisition Based onStable Matching of Bipartite Graph in Cooperative VehiclendashInfrastructure Systemsrdquo Sensors vol 17 no 6 p 1327 2017

[44] P B Velloso R P Laufer D D O O Cunha O CM B Duarteand G Pujolle ldquoTrust management in mobile ad hoc networksusing a scalable maturity-based modelrdquo IEEE Transactions onNetwork and ServiceManagement vol 7 no 3 pp 172ndash185 2010

[45] F Bao I Chen M Chang and J Cho ldquoHierarchical trustmanagement forwireless sensor networks and its applications totrust-based routing and intrusion detectionrdquo IEEE Transactionson Network and Service Management vol 9 no 2 pp 169ndash1832012

[46] AWald Sequential Analysis Johnwiley ad Sons NewYork NYUSA 7th edition 1965

[47] J Jung V Paxson A W Berger and H Balakrishnan ldquoFastportscan detection using sequential hypothesis testingrdquo inProceedings of the IEEE Symposium on Security and PrivacyIEEE California Calif USA 2004

[48] G Peter S Zsolt andA Szilard ldquoHighly automatedVehilce Sys-temsrdquo Mechatronics Engineer MSc Curriculum DevelopmentBME MOGI 2014

[49] C Sommer R German and F Dressler ldquoBidirectionally cou-pled network and road traffic simulation for improved IVCanalysisrdquo IEEE Transactions on Mobile Computing vol 10 no1 pp 3ndash15 2011

[50] A Varga and R Hornig ldquoAn overview of the OMNeT++simulation environmentrdquo in Proceedings of the 1st InternationalICST Conference on Simulation Tools and Techniques for Com-munications Networks and Systems (SIMUTools rsquo08) March2008

[51] D Krajzewicz J Erdmann M Behrisch and L Bieker ldquoRecentDevelopment and Applications of SUMO - Simulation ofUrban MObilityrdquo in Proceedings of the Recent Development andApplications of SUMO - Simulation of UrbanMObility vol 5 pp128ndash138 December 2012

[52] W Ben Jaballah M Conti M Mosbah and C E PalazzildquoThe impact of malicious nodes positioning on vehicular alertmessaging systemrdquo Ad Hoc Networks vol 52 pp 3ndash16 2016

[53] A Tajeddine A Kayssi and A Chehab ldquoA privacy-preservingtrust model for VANETsrdquo in Proceedings of the 10th IEEEInternational Conference on Computer and Information Technol-ogy CIT-2010 7th IEEE International Conference on EmbeddedSoftware and Systems ICESS-2010 10th IEEE Int Conf ScalableComputing and Communications ScalCom rsquo10 pp 832ndash837 July2010

[54] J Petit and Z Mammeri ldquoDynamic consensus for securedvehicular ad hoc networksrdquo in Proceedings of the 2011 IEEE7th International Conference onWireless andMobile ComputingNetworking and Communications WiMob rsquo11 pp 1ndash8 October2011

[55] B Ostermaier F Dotzer and M Strassberger ldquoEnhancing thesecurity of local danger warnings in VANETs - A simulativeanalysis of voting schemesrdquo in Proceedings of the 2nd Interna-tional Conference on Availability Reliability and Security ARESrsquo07 pp 422ndash431 April 2007

Submit your manuscripts athttpswwwhindawicom

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

Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia


Biomedical Imaging


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



Some of the issues in vehicular networks include simplerouting problems and application-oriented problems likeSybil attacks and false data dissemination [6]The traditionalreputation systems may not work efficiently in vehicularnetworks [7] Public key infrastructure (PKI) may not beavailable everywhere during the initial stages of vehicularnetwork deployment around a country because some regionsmay not be covered due to deployment costs or budgetissues Generally cryptography-based verification ofmessagetrustworthiness is computationally expensive It can protectagainst a few types of attack from external nodes Howeverit will not protect against malicious nodes in the networkwhich already have the required cryptographic keys andmaynot be suitable for V2V ephemeral network communicationsOur scheme does not use cryptography and centralizedservers and thus it does not have a single point of failureMost VANET models assume that the system is up and run-ning where all vehicles have a certain trust score However itis not easy to know the trustworthiness of vehicles withouthaving had any interaction with those vehicles In highlydistributed vehicular networks vehicles can join and leavea network frequently [8 9] When a new vehicle joins thenetwork for the first time there is no information aboutit One of the challenges faced by VANET is that the trustmodel of the VANET should consider the requirement foranonymity of vehicles The trust model should have minimaloverhead in terms of computation complexity as well asstorage The trust model should be robust to data-centricattacks and be able to detect those attacks [10ndash12] VANETsecurity frameworks should be light scalable reliable andsecure

Our proposed scheme investigates the trustworthinessof event information received from adjacent vehicles whichserves as multiple pieces of evidence We use truth-tellingprobability as a measure for the trustworthiness of a vehicleThe vehicles communicate through safety messages to reportevents such as accident information safety warnings infor-mation on traffic jams weather reports and reports of ice onthe road In our proposed scheme all vehicles are assumedto have a pseudo identity (PID) which is independent ofthe node identity Each vehicle broadcasts an event messageto adjacent vehicles from the time it collects informationabout that event Every vehicle maintains the trust levelof its neighbors in a distributed manner to cope with thepropagation of false information We introduce an enhanced119870-means clustering technique to minimize the effect ofmalicious nodes on trust level calculationWe use a modifiedthreshold random walk algorithm with a single threshold tomake a final decision about the occurrence of an event whilesupporting real-time decision We focus on determining thetrustworthiness of the event information in the receivedmessages by considering reports from neighboring vehiclesdifferently with a truth-telling probability

The main contributions of our work can be summarizedas follows

(i) Our proposed scheme can contribute to dissemina-tion of trustworthy information since each vehicle

makes a decision on the trustworthiness of infor-mation in the received message individually whiledropping packets containing fake information

(ii) Since all the decisions are made based on the infor-mation received from the neighbor vehicles ourproposed scheme can work in an infrastructure-lessenvironment as well

(iii) Our proposed scheme can make a better decisionon the trustworthiness of a given message comparedto a simple voting mechanism since the modifiedthreshold random walk (TRW) can give a higherweight on the opinion of a vehicle whichmakes moretrue statements than false statements

The remainder of this paper is organized as follows InSection 2 we discuss the related work In Section 3 we pro-pose a trustworthy event-information dissemination schemefor VANET In Section 4 we evaluate the performance of theproposed scheme using simulation Section 5 concludes thepaper with future work

2 Related Work

Several trust management systems have been proposed forVANET [13ndash15] Trust management systems evaluate thetrust values of the neighbor nodes to prevent them frominteracting with the malicious nodes The authors in [16]provide a quantitative and systematic review of existing trustmanagement schemes for VANETThey address comprehen-sive trust model concepts problems and solutions related toVANET trust management There are several works on trustmanagement scheme based on infrastructure framework andcryptography techniques Trust management schemes can bedivided into four categories based on the use of infrastructureand cryptographicmeasures such as public key infrastructure(PKI) as shown in Table 1 The first category represents thetrust management techniques based on infrastructure suchas RSU and PKI In the second category the nodes rely oninfrastructure for trust management without using PKI Inthe third category each node handles the issue of messagetrustworthiness based on PKI without using any infrastruc-ture In the fourth category the nodes are fully decentralizedand operate in infrastructure-less environment and they donot depend on PKI

In the first category trust management systems are basedon infrastructure such as RSU and PKI and can be effectivein identifying malicious nodes with some accuracy [17ndash23]However trust management schemes in this category maynot work if infrastructure is not available The trust manage-ment based on PKI is computationally expensive and cannotsecure VANET against insider attack where the maliciousnodes have already acquired the cryptographic keys [5 23]Some researchers used group signatures (GS) techniques [17ndash21] to authenticate message sender and guarantee messageintegrity such as Identity-Based Group Signatures (IBGS) in[18] GSIS in [19] and Identity-based BatchVerification (IBV)in [20] However GS schemes are usually based on PKIandmessage sender authentication cannot prevent legitimatenodes from sending malicious messages













3 System Model




Driving direction

Vk

Vj

Vi Event

MF

ME


















119872119864 = (119901119894 119905 119871119864 119897119894) (1)




120579119894 = 119898119899 (2)




119894 = 1119899119899sum119895=1

120579119894119895 (3)



= 1 + (1 minus 1


119873)119883119899minus1 otherwise(4)





119873)119864 [119883119899minus1] (5)


119864 [119883119899] = (1 minus 1119873)119899 119864 [1198830] minus 120579119894119873 + 120579119894119873 (6)





Θ119895 = ((1199011119895 1205791119895) (1199012119895 1205792119895) (119901119895119895 120579119895119895) (119901119899119895 120579119899119895)) (7)



Θj1015840 = (1j1015840 2j1015840 1198951015840j1015840 1198991015840j1015840) (8)





1198961015840j1015840 = 120579119896119895 (9)




Γ =[[[[[[[

12057911 12057912 120579111989912057921 12057922 1205792119899 d

1205791198991 1205791198992 120579119899119899

]]]]]]]

(10)


C

A

B

1

08

06

04

02

0

108

0604

020 1

08

06

04

02

0

3

1

2



119881119895 1198811198941 2 3

1 1 07 022 08 1 043 05 05 1

A B C





Θ119894isinY120583 minusΘ119894



119895





120583 = sum119899119894=1Θ119894|Y| (11)



1003817100381710038171003817120583 minusΘ1198941003817100381710038171003817 (12)




1003817100381710038171003817c1 minusΘ1198941003817100381710038171003817 (13)











119884119894 = 0 if 119883119894 = 11986411 if 119883119894 = 1198641 (14)



Pr (119884119894 = 0 | 1198670) = 120579119894Pr (119884119894 = 1 | 1198670) = 1 minus 120579119894

Pr (119884119894 = 0 | 1198671) = 1 minus 120579119894Pr (119884119894 = 1 | 1198671) = 120579119894

(15)


Λ = 119873prod119894=1

Pr (119884119894 | 1198671)Pr (119884119894 | 1198670) (16)





Λ (119883) = 0208 times 02


04 times 0604 = 021

lt 1 (= 120578) (17)



Accident occurred

Accident eventVn

Vm

Vk

Vj

Vj

Vi

Vl


(a)

Vn

Vm

Vk

Vj

ViVO

Vl


direction due to Vk

Vm changes its

(b)




119872119865 = (119901119895 119905119872119864Θ119895) (18)















0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity









0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in















3 System Model




Driving direction

Vk

Vj

Vi Event

MF

ME


















119872119864 = (119901119894 119905 119871119864 119897119894) (1)




120579119894 = 119898119899 (2)




119894 = 1119899119899sum119895=1

120579119894119895 (3)



= 1 + (1 minus 1


119873)119883119899minus1 otherwise(4)





119873)119864 [119883119899minus1] (5)


119864 [119883119899] = (1 minus 1119873)119899 119864 [1198830] minus 120579119894119873 + 120579119894119873 (6)





Θ119895 = ((1199011119895 1205791119895) (1199012119895 1205792119895) (119901119895119895 120579119895119895) (119901119899119895 120579119899119895)) (7)



Θj1015840 = (1j1015840 2j1015840 1198951015840j1015840 1198991015840j1015840) (8)





1198961015840j1015840 = 120579119896119895 (9)




Γ =[[[[[[[

12057911 12057912 120579111989912057921 12057922 1205792119899 d

1205791198991 1205791198992 120579119899119899

]]]]]]]

(10)


C

A

B

1

08

06

04

02

0

108

0604

020 1

08

06

04

02

0

3

1

2



119881119895 1198811198941 2 3

1 1 07 022 08 1 043 05 05 1

A B C





Θ119894isinY120583 minusΘ119894



119895





120583 = sum119899119894=1Θ119894|Y| (11)



1003817100381710038171003817120583 minusΘ1198941003817100381710038171003817 (12)




1003817100381710038171003817c1 minusΘ1198941003817100381710038171003817 (13)











119884119894 = 0 if 119883119894 = 11986411 if 119883119894 = 1198641 (14)



Pr (119884119894 = 0 | 1198670) = 120579119894Pr (119884119894 = 1 | 1198670) = 1 minus 120579119894

Pr (119884119894 = 0 | 1198671) = 1 minus 120579119894Pr (119884119894 = 1 | 1198671) = 120579119894

(15)


Λ = 119873prod119894=1

Pr (119884119894 | 1198671)Pr (119884119894 | 1198670) (16)





Λ (119883) = 0208 times 02


04 times 0604 = 021

lt 1 (= 120578) (17)



Accident occurred

Accident eventVn

Vm

Vk

Vj

Vj

Vi

Vl


(a)

Vn

Vm

Vk

Vj

ViVO

Vl


direction due to Vk

Vm changes its

(b)




119872119865 = (119901119895 119905119872119864Θ119895) (18)















0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity









0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in




Driving direction

Vk

Vj

Vi Event

MF

ME


















119872119864 = (119901119894 119905 119871119864 119897119894) (1)




120579119894 = 119898119899 (2)




119894 = 1119899119899sum119895=1

120579119894119895 (3)



= 1 + (1 minus 1


119873)119883119899minus1 otherwise(4)





119873)119864 [119883119899minus1] (5)


119864 [119883119899] = (1 minus 1119873)119899 119864 [1198830] minus 120579119894119873 + 120579119894119873 (6)





Θ119895 = ((1199011119895 1205791119895) (1199012119895 1205792119895) (119901119895119895 120579119895119895) (119901119899119895 120579119899119895)) (7)



Θj1015840 = (1j1015840 2j1015840 1198951015840j1015840 1198991015840j1015840) (8)





1198961015840j1015840 = 120579119896119895 (9)




Γ =[[[[[[[

12057911 12057912 120579111989912057921 12057922 1205792119899 d

1205791198991 1205791198992 120579119899119899

]]]]]]]

(10)


C

A

B

1

08

06

04

02

0

108

0604

020 1

08

06

04

02

0

3

1

2



119881119895 1198811198941 2 3

1 1 07 022 08 1 043 05 05 1

A B C





Θ119894isinY120583 minusΘ119894



119895





120583 = sum119899119894=1Θ119894|Y| (11)



1003817100381710038171003817120583 minusΘ1198941003817100381710038171003817 (12)




1003817100381710038171003817c1 minusΘ1198941003817100381710038171003817 (13)











119884119894 = 0 if 119883119894 = 11986411 if 119883119894 = 1198641 (14)



Pr (119884119894 = 0 | 1198670) = 120579119894Pr (119884119894 = 1 | 1198670) = 1 minus 120579119894

Pr (119884119894 = 0 | 1198671) = 1 minus 120579119894Pr (119884119894 = 1 | 1198671) = 120579119894

(15)


Λ = 119873prod119894=1

Pr (119884119894 | 1198671)Pr (119884119894 | 1198670) (16)





Λ (119883) = 0208 times 02


04 times 0604 = 021

lt 1 (= 120578) (17)



Accident occurred

Accident eventVn

Vm

Vk

Vj

Vj

Vi

Vl


(a)

Vn

Vm

Vk

Vj

ViVO

Vl


direction due to Vk

Vm changes its

(b)




119872119865 = (119901119895 119905119872119864Θ119895) (18)















0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity









0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in













119872119864 = (119901119894 119905 119871119864 119897119894) (1)




120579119894 = 119898119899 (2)




119894 = 1119899119899sum119895=1

120579119894119895 (3)



= 1 + (1 minus 1


119873)119883119899minus1 otherwise(4)





119873)119864 [119883119899minus1] (5)


119864 [119883119899] = (1 minus 1119873)119899 119864 [1198830] minus 120579119894119873 + 120579119894119873 (6)





Θ119895 = ((1199011119895 1205791119895) (1199012119895 1205792119895) (119901119895119895 120579119895119895) (119901119899119895 120579119899119895)) (7)



Θj1015840 = (1j1015840 2j1015840 1198951015840j1015840 1198991015840j1015840) (8)





1198961015840j1015840 = 120579119896119895 (9)




Γ =[[[[[[[

12057911 12057912 120579111989912057921 12057922 1205792119899 d

1205791198991 1205791198992 120579119899119899

]]]]]]]

(10)


C

A

B

1

08

06

04

02

0

108

0604

020 1

08

06

04

02

0

3

1

2



119881119895 1198811198941 2 3

1 1 07 022 08 1 043 05 05 1

A B C





Θ119894isinY120583 minusΘ119894



119895





120583 = sum119899119894=1Θ119894|Y| (11)



1003817100381710038171003817120583 minusΘ1198941003817100381710038171003817 (12)




1003817100381710038171003817c1 minusΘ1198941003817100381710038171003817 (13)











119884119894 = 0 if 119883119894 = 11986411 if 119883119894 = 1198641 (14)



Pr (119884119894 = 0 | 1198670) = 120579119894Pr (119884119894 = 1 | 1198670) = 1 minus 120579119894

Pr (119884119894 = 0 | 1198671) = 1 minus 120579119894Pr (119884119894 = 1 | 1198671) = 120579119894

(15)


Λ = 119873prod119894=1

Pr (119884119894 | 1198671)Pr (119884119894 | 1198670) (16)





Λ (119883) = 0208 times 02


04 times 0604 = 021

lt 1 (= 120578) (17)



Accident occurred

Accident eventVn

Vm

Vk

Vj

Vj

Vi

Vl


(a)

Vn

Vm

Vk

Vj

ViVO

Vl


direction due to Vk

Vm changes its

(b)




119872119865 = (119901119895 119905119872119864Θ119895) (18)















0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity









0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in





119894 = 1119899119899sum119895=1

120579119894119895 (3)



= 1 + (1 minus 1


119873)119883119899minus1 otherwise(4)





119873)119864 [119883119899minus1] (5)


119864 [119883119899] = (1 minus 1119873)119899 119864 [1198830] minus 120579119894119873 + 120579119894119873 (6)





Θ119895 = ((1199011119895 1205791119895) (1199012119895 1205792119895) (119901119895119895 120579119895119895) (119901119899119895 120579119899119895)) (7)



Θj1015840 = (1j1015840 2j1015840 1198951015840j1015840 1198991015840j1015840) (8)





1198961015840j1015840 = 120579119896119895 (9)




Γ =[[[[[[[

12057911 12057912 120579111989912057921 12057922 1205792119899 d

1205791198991 1205791198992 120579119899119899

]]]]]]]

(10)


C

A

B

1

08

06

04

02

0

108

0604

020 1

08

06

04

02

0

3

1

2



119881119895 1198811198941 2 3

1 1 07 022 08 1 043 05 05 1

A B C





Θ119894isinY120583 minusΘ119894



119895





120583 = sum119899119894=1Θ119894|Y| (11)



1003817100381710038171003817120583 minusΘ1198941003817100381710038171003817 (12)




1003817100381710038171003817c1 minusΘ1198941003817100381710038171003817 (13)











119884119894 = 0 if 119883119894 = 11986411 if 119883119894 = 1198641 (14)



Pr (119884119894 = 0 | 1198670) = 120579119894Pr (119884119894 = 1 | 1198670) = 1 minus 120579119894

Pr (119884119894 = 0 | 1198671) = 1 minus 120579119894Pr (119884119894 = 1 | 1198671) = 120579119894

(15)


Λ = 119873prod119894=1

Pr (119884119894 | 1198671)Pr (119884119894 | 1198670) (16)





Λ (119883) = 0208 times 02


04 times 0604 = 021

lt 1 (= 120578) (17)



Accident occurred

Accident eventVn

Vm

Vk

Vj

Vj

Vi

Vl


(a)

Vn

Vm

Vk

Vj

ViVO

Vl


direction due to Vk

Vm changes its

(b)




119872119865 = (119901119895 119905119872119864Θ119895) (18)















0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity









0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in







1198961015840j1015840 = 120579119896119895 (9)




Γ =[[[[[[[

12057911 12057912 120579111989912057921 12057922 1205792119899 d

1205791198991 1205791198992 120579119899119899

]]]]]]]

(10)


C

A

B

1

08

06

04

02

0

108

0604

020 1

08

06

04

02

0

3

1

2



119881119895 1198811198941 2 3

1 1 07 022 08 1 043 05 05 1

A B C





Θ119894isinY120583 minusΘ119894



119895





120583 = sum119899119894=1Θ119894|Y| (11)



1003817100381710038171003817120583 minusΘ1198941003817100381710038171003817 (12)




1003817100381710038171003817c1 minusΘ1198941003817100381710038171003817 (13)











119884119894 = 0 if 119883119894 = 11986411 if 119883119894 = 1198641 (14)



Pr (119884119894 = 0 | 1198670) = 120579119894Pr (119884119894 = 1 | 1198670) = 1 minus 120579119894

Pr (119884119894 = 0 | 1198671) = 1 minus 120579119894Pr (119884119894 = 1 | 1198671) = 120579119894

(15)


Λ = 119873prod119894=1

Pr (119884119894 | 1198671)Pr (119884119894 | 1198670) (16)





Λ (119883) = 0208 times 02


04 times 0604 = 021

lt 1 (= 120578) (17)



Accident occurred

Accident eventVn

Vm

Vk

Vj

Vj

Vi

Vl


(a)

Vn

Vm

Vk

Vj

ViVO

Vl


direction due to Vk

Vm changes its

(b)




119872119865 = (119901119895 119905119872119864Θ119895) (18)















0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity









0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in






Θ119894isinY120583 minusΘ119894



119895





120583 = sum119899119894=1Θ119894|Y| (11)



1003817100381710038171003817120583 minusΘ1198941003817100381710038171003817 (12)




1003817100381710038171003817c1 minusΘ1198941003817100381710038171003817 (13)











119884119894 = 0 if 119883119894 = 11986411 if 119883119894 = 1198641 (14)



Pr (119884119894 = 0 | 1198670) = 120579119894Pr (119884119894 = 1 | 1198670) = 1 minus 120579119894

Pr (119884119894 = 0 | 1198671) = 1 minus 120579119894Pr (119884119894 = 1 | 1198671) = 120579119894

(15)


Λ = 119873prod119894=1

Pr (119884119894 | 1198671)Pr (119884119894 | 1198670) (16)





Λ (119883) = 0208 times 02


04 times 0604 = 021

lt 1 (= 120578) (17)



Accident occurred

Accident eventVn

Vm

Vk

Vj

Vj

Vi

Vl


(a)

Vn

Vm

Vk

Vj

ViVO

Vl


direction due to Vk

Vm changes its

(b)




119872119865 = (119901119895 119905119872119864Θ119895) (18)















0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity









0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in










119884119894 = 0 if 119883119894 = 11986411 if 119883119894 = 1198641 (14)



Pr (119884119894 = 0 | 1198670) = 120579119894Pr (119884119894 = 1 | 1198670) = 1 minus 120579119894

Pr (119884119894 = 0 | 1198671) = 1 minus 120579119894Pr (119884119894 = 1 | 1198671) = 120579119894

(15)


Λ = 119873prod119894=1

Pr (119884119894 | 1198671)Pr (119884119894 | 1198670) (16)





Λ (119883) = 0208 times 02


04 times 0604 = 021

lt 1 (= 120578) (17)



Accident occurred

Accident eventVn

Vm

Vk

Vj

Vj

Vi

Vl


(a)

Vn

Vm

Vk

Vj

ViVO

Vl


direction due to Vk

Vm changes its

(b)




119872119865 = (119901119895 119905119872119864Θ119895) (18)















0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity









0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in




Accident occurred

Accident eventVn

Vm

Vk

Vj

Vj

Vi

Vl


(a)

Vn

Vm

Vk

Vj

ViVO

Vl


direction due to Vk

Vm changes its

(b)




119872119865 = (119901119895 119905119872119864Θ119895) (18)















0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity









0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in











0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity









0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in






0

10

20

30

40

50

False

dec

ision

pro

babi

lity

Modified TRW

200 400 600 800 1000 2000100Time (ms)





0

10

20

30

40

50

False

dec

ision

pro

babi

lity






0102030405060

False

dec

ision

pro

babi

lity











0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in







0

10

20

30

40

50

False

dec

ision

pro

babi

lity



0

10

20

30

40

50

60

False

dec

ision

pro

babi

lity








0100020003000400050006000

Mes

sage

ove

rhea

d (b

ytes

)









0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in




0

1000

2000

3000

4000

5000

6000

Mes

sage

ove

rhea

d (b

ytes

)







Notations




Acknowledgments


References

































































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in























































Advances in

FuzzySystems


Volume 2014












Journal of



Advances in

Multimedia


Biomedical Imaging


Advances in


RoboticsJournal of










Advances in



Trustworthy Event-Information Dissemination in Vehicular ...

Documents

Transcript of Trustworthy Event-Information Dissemination in Vehicular ...