an efficient dynamic trust evaluation model for wireless

Research ArticleAn Efficient Dynamic Trust Evaluation Model forWireless Sensor Networks

Zhengwang Ye12 TaoWen13 Zhenyu Liu13 Xiaoying Song1 and Chongguo Fu13

1School of Computer Science and Engineering Northeastern University Shenyang China2Department of Network Information Center Tonghua Normal University Tonghua China3School of Computer Science and Technology Dalian Neusoft University of Information Dalian China

Correspondence should be addressed to Zhengwang Ye zhengwang119126com

Received 7 April 2017 Accepted 25 September 2017 Published 24 October 2017

Academic Editor Jaime Lloret

Copyright copy 2017 Zhengwang Ye et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Trust evaluation is an effective method to detect malicious nodes and ensure security in wireless sensor networks (WSNs) Inthis paper an efficient dynamic trust evaluation model (DTEM) for WSNs is proposed which implements accurate efficient anddynamic trust evaluation by dynamically adjusting the weights of direct trust and indirect trust and the parameters of the updatemechanism To achieve accurate trust evaluation the direct trust is calculated considering multitrust including communicationtrust data trust and energy trust with the punishment factor and regulating function The indirect trust is evaluated conditionallyby the trusted recommendations from a third party Moreover the integrated trust is measured by assigning dynamic weights fordirect trust and indirect trust and combining them Finally we propose an updatemechanismby a slidingwindowbased on inducedordered weighted averaging operator to enhance flexibility We can dynamically adapt the parameters and the interactive historywindows number according to the actual needs of the network to realize dynamic update of direct trust value Simulation resultsindicate that the proposed dynamic trust model is an efficient dynamic and attack-resistant trust evaluationmodel Compared withexisting approaches the proposed dynamic trust model performs better in defending multiple malicious attacks

1 Introduction

Nowadays technology development in the fields of micro-electromechanical system (MEMS) andwireless communica-tion has facilitated the extensive distribution ofWSNsWSNsare composed of a large number of sensor nodes In generalsensor nodes are reliable accurate flexible inexpensive andeasy to deploy Some areas and industries that are subject toenvironmental constraints rely on WSNs for data collectionand monitoring [1] They are widely used in many applica-tions such as emergency response [2] healthcare monitoring[3] military agriculture [4] environmental monitoring andsmart power grid [5] However due to the characteristicsof working environments (usually deployed in remote andunattended) and the way of wireless communication WSNsare prone to sudden accidents failures and suffer fromattacks of malicious nodes Once a node is compromised theavailability and integrity of the network can be destroyed In

addition it is difficult to predict themalicious attacks Hencenetwork security is a vital issue which needs to be addressedto guarantee correct operation of the whole network

Recently in the security field of wireless network agreat deal of research [6ndash8] has been carried out commonlyusing cryptography authentication and hash functions toimprove the security of network Undoubtedly the presentachievements have greatly promoted related research inimproving security of network especially the confidentiallyintegrity authentication availability and no-repudiation ofdata in the network But in the security field of WSNs theabove traditional security mechanisms such as cryptographyand authentication are not mostly suitable for processingcapability constrained and energy limited WSNs due to thecomplexity and huge computing memory [9] Furthermorethe traditional security mechanisms are widely and availablyused to deal with external attacks but cannot solve insider ornode misbehavior attacks effectively which are caused by the

HindawiJournal of SensorsVolume 2017 Article ID 7864671 16 pageshttpsdoiorg10115520177864671

2 Journal of Sensors

captured nodes [10] In pursuit of the security of WSNs trustand reputation mechanisms have proven to be more resilientagainst insider or node misbehavior attacks [10 11]

Trust in the field of wireless communication networksmay be defined as the degree of belief on the future behaviorof other nodes which is based on past experience andobservations of the nodesrsquo action [11] So we give thedefinition of trust in WSNs as follows node 119860rsquos trust innode 119861 describes the belief or expectation or assuranceof sincerity competence and integrity of node 119861rsquos futureactionbehavior [12] The basic idea of trust based scheme isto quantify trust to describe the trustworthiness reliabilityor competence of individual nodes [5] Trust managementsystem can be implemented in various applications forsecurity management such as secure protocol [8] secure dataaggregation [13] trusted routing [14] and intrusion detectionsystem [15] In recent years lots of state-of-the-art models[5 12ndash31] have been proposed in this field Undoubtedly thepresent achievements have greatly promoted related researchin improving security of WSNs Even so trust evaluationin WSNs is still a challenging issue Some limitations areexhibited which need more attention to be solved

Considerable research has been done on modeling andmanaging trust and reputation in WSNs Many currentstudies [18ndash20] have been done for trust establishmentjust only based on the communication interaction recordsbetween nodes without considering the data consistency sothey cannot be against attacks on data While other studies[25ndash28] combine multifactors to calculate the trust valuethe multitrust sums up in weighted manner to computethe integrated trust But the weights are obtained by expertopinion method or average weight method The results ofthe prediction are subjective which affect the scientific andflexibility of the trust decision In addition trust evaluationis a dynamic phenomenon and changes with time andenvironment condition In many current trust models [2324] the trust value is updated by a sliding time windowusing forgetting or aging mechanism But the number ofsliding windows is defined by expert opinion method Oncethe number of the sliding time windows is confirmed it isdifficult to change It makes the trust models unable to adaptto the dynamic changes of the network environment whichaffects the accuracy of the result To our knowledge thereis no literature that can dynamically adjust the number ofthe sliding time windows and the parameters to achieve adynamic update mechanism Moreover some existing trustmodels [18 23 27] rarely consider the influence of theenergy consumption Due to these reasons there is a growingdemand for adequate provision of an efficient dynamic trustevaluation model for WSNs it can achieve accurate trustevaluation dynamically according to the environment andrequirements and can realize the identification and defenseof various types of malicious attack

In this paper an efficient dynamic trust evaluation model(DTEM) for WSNs is proposed that aims to address theabove problems In the proposed trust model the trust valueis calculated considering multitrust factors it can achieveaccurate trust evaluation Moreover DTEM can dynamicallyadjust the weights of direct trust and indirect trust It

reflects the dynamic adaptability of the trust computing Italso can dynamically adjust the parameters of the updatemechanism to update the trust value to meet the actualneeds of the network environmentTheDTEM can be againstvarious types of malicious attack and can be configured andeffectively applied to different environments with differentrequirementsThemajor contributions of this paper are listedas follows

(1) To improve the accuracy of trust evaluation againstattacks on data the trust value is calculated con-sidering direct trust and indirect trust The directtrust is calculated considering multitrust includingcommunication trust data trust and energy trustwith the punishment factor and regulating function tomeet the following character ldquotrust is hard to acquireand easy to loserdquo The indirect trust is evaluatedconditionally by the trusted recommendations froma third party

(2) To ensure that the trust model makes a decisionmore scientifically dynamically and adaptively wedefine a dynamic balance weight factor functionwhich is changed dynamically with the number ofcommunication interactions The adaptive dynamicbalance weight factor dynamically adjusts the weightof the direct trust and indirect trust

(3) To make sure that the proposed trust model canbe configured and effectively applied to differentenvironmentswith different requirements and againston-off attack we give an update mechanism bya sliding time window based on induced orderedweighted averaging operator (IOWA) to enhance flex-ibility We can dynamically adapt the parameters andthe interactive history windows number to changethe weight sequence to update the trust value toadapt with different environments and requirementsMeanwhile on-off attacks can be handled efficiently

At last compared to the existing trust models (RFSN [18]and BTMS [24]) simulation results show that the proposedtrust model has remarkable enhancements in the accuracy oftrust decision and has a better capability to capture dynamicmalicious nodes behaviors

The remainder of this paper is organized as followsSection 2 gives an overview of related works Network modeland attack model are described in Section 3 Section 4 givesthe overview and process of the DTEM Section 5 discussestrust model and trust evaluation mechanism in DTEM InSection 6 the experiment is made under simulative envi-ronments and the performance of the DTEM is evaluatedSection 7 concludes this paper

2 Related Works

In most current trust model researches focus on sensor radiocommunication behaviors Sensor nodes build node trustmodel through wireless radio transaction with neighboringnodes Ganeriwal and Srivastava [18] first proposed a repu-tation based framework for sensor networks (RFSN) where

Journal of Sensors 3

nodes used reputation to evaluate otherrsquos trustworthinessThe framework uses watchdog mechanism to monitor com-munication behavior of neighboring nodes and representsnode reputation distribution using Beta distribution Thenthe trust value is figured out according to the statisticalexpectation of the probability reputation distribution Thetrust framework is good robustness and very classic Butthe recommendation trust is not considered it cannot resistvarious internal attacks In [19] an agent-based trust modelwas proposed in WSNs (ATSN) agent node was used tomonitor behaviors of sensor nodes and classify the behaviorsinto good or bad ones Agent nodes count all the numberof good behaviors and malicious behaviors respectively andsave the results into a three-tuple ATSN scheme uses agentwhich can save the computational resources and energyconsumption However in ATSN only the direct trust valueis calculated while the recommendation trust is ignoredIn addition the updating process of the trust value is notconsidered Shaikh et al [20] proposed a new lightweightgroup-based trustmanagement scheme (GTMS) for clusteredWSNs The trust value is obtained through the communi-cation behavior of neighboring nodes It works on trust atthree levels the node level the cluster head level and thebase station level The model establishes trust mechanismfrom the above aspects to resist the attack of maliciousnodes respectively GTMS can effectively resist the attacks ofmalicious nodes and it does not require large data storageand complex computations However only observing thenumber of successful and unsuccessful interactions cannotreflect soundest trust value Song et al [21] proposed adynamic trust evaluation method based on multifactor Thenodesrsquo trustworthiness is measured by combining direct trustwith indirect trust dynamically Besides both the involvedclassification standard and dynamic weight assignment aredependent on the interaction times between nodes which areput forward under the background of Hoeffdingrsquos Inequalityin Probability Theory The simulation results show that thismethod is sensitive to multiple attacks But the updating pro-cess of the trust value is not considered Li et al [22] proposeda lightweight and dependable trust system for the clusteredWSNs The trust decision-making scheme is proposed basedon the nodesrsquo roles in clustered WSNs They improve systemefficiency by canceling feedback between cluster membersor between cluster heads The trust scheme also definesa self-adaptive weighted method for trust aggregation atcluster head level This approach surpasses the limitationsof traditional weighting methods for trust factors in whichweights are assigned subjectively In [23] He et al definedan attack-resistant and lightweight trustmanagement schemecalled ReTrust This system is oriented to medical sensornetwork and based on hierarchical architecture comprisedof master nodes and sensor nodes ReTrust uses sliding timewindow and aging factor to identify and eliminate the on-off attack Bad-mouthing attack is avoided by eliminatingoutliers after collecting recommendations It is resistant tobad-mouthing and on-off attacks However the drawbackof this scheme is that master nodes must have abundantstorage and energy Feng et al [24] proposed a credibleBayesian-based trustmanagement scheme (BTMS)The trust

management scheme takes the direct and indirect trust intoaccountThe direct trust is calculated by a modified Bayesianequation with punishment factor and updated by a slidingwindow using an adaptive forgetting factor Moreover theindirect trust computation is invoked from a third partyBTMS performs better in resisting attacks

While in above-mentioned trust models data security isneglected in existing trust models many of them focus on thetrust of data as the main work In [25] Zhan et al proposeda resilient trust model with a focus on data integrity andsensor node trust for hierarchical WSNs The sensor nodecurrent trust level is evaluated through the past history andrecent risk And then it employs Gaussian model to rate dataintegrity in a fine-grained style The model is proven to beresilient against faulty data and malicious data manipulationBut the energy consumption on node is not considered In[26] a wireless sensor network based on multiangle trustof node was proposed The method considers the sensingdata and the nodersquos energy in the factors of trust assessmentthe integrated trust value is calculated through the averageweight of the communication trust energy trust and datatrust It is more reliable and effective against Dos attackand data forgery attack But the trust update mechanism isignored Jiang et al [27] proposed an efficient distributedtrust model (EDTM) for wireless sensor networks In EDTMthe direct trust and recommendation trust are selectivelycalculated according to the number of packets received bysensor node The direct trust value is calculated throughthe average weight of the communication trust energytrust and data trust In addition trust reliability and trustfamiliarity are defined to improve recommendation accuracyEDTM can evaluate trustworthiness of sensor nodes moreprecisely and identify the malicious nodes more effectivelyIn [28] a consensus-aware sociopsychological trust modelfor WSNs was proposed The trust model uses the conceptof consensus and consistency in understanding the behaviorof the sensor nodes for detecting fraudulent nodes in WSNThe factors of ability benevolence and integrity are usedfor the computation of trust The approach can deal evenin the presence of higher number of fraudulent nodes thanbenevolent nodes It is more reliable and effective againstattacks on data in WSNs But communication faults thatdelay the rate at which packets are sent are not considered

Recently several techniques are used in computing thetrust of sensor nodes such as the fuzzy logic approach theBayesian network approach the game theoretic approachswarm intelligence and the cloud method In [29] Feng etal first established various trust factors depending on thecommunication behaviors to evaluate the trustworthinessof sensor nodes The direct and indirect trust are obtainedthrough calculating weighted average of trust factors Mean-while the fuzzy set method is applied to measure how muchthe trust value of node belongs to each trust degree Andthen the evidence difference is calculated between the directand indirect trust which links the revised D-S evidencecombination rule to finally synthesize integrated trust valueof nodes Zhang et al [30] proposed a trust evaluationmethod for clustered wireless sensor networks based oncloud model The method considers multifactors including


communication factor message factor and energy factor toget factor trust cloud And the trust cloud is calculated byassigning weights for each factor trust cloud and combiningthem The final trust cloud is measured by synthesizing therecommendation trust cloud and immediate trust cloud andis converted to trust grade by trust cloud decision-makingThe method can detect malicious nodes according to differ-ent secure requirements under different WSNs applicationswhich provides a safe running environment for differentapplications Shen et al [31] studied the trust decision andits dynamics that played a key role in stabilizing the wholenetwork using evolutionary game theory The evolutionarygame theory is used for the area of trust evolution in WSNsIt sets up a WSNs trust game concerning the dynamics oftrust evolution during sensor nodersquos decision-making Whensensor nodes are making their decisions to select action trustor mistrust a WSNs trust game is created to reflect theirutilities It can find out the conditions that will lead sensornodes to choose action trust as their final behavior to ensureWSNsrsquo security and stability

Our work is partly motivated by those related worksabove however there are some distinctions compared withthem In our trust model the trust value is calculated consid-ering multitrust factors including communication trust datatrust and energy trust Not like the works in [18ndash20] the trustvalues are just only based on the communication interactionrecords so they cannot be against attacks on data Whileother studies [25ndash28] combine multifactors to calculate thetrust value they do not consider the following characterldquotrust is hard to acquire and easy to loserdquo In our proposedtrust model we add the punishment factor and regulatingfunction to realize the punishment and adjustment of trustvalue against bad-mouth attack and collusion attack Thenin most trust models such as in [18 25ndash28] the direct trustand indirect trust sum up in weighted manner to computethe integrated trust But the weights are obtained by expertopinionmethod or average weight method in [25ndash28] In ourproposed trustmodel we define an adaptive dynamic balanceweight function to dynamically adjust the weight of the directtrust and indirect trust Although the works in [21 22] havegiven various computation methods of the dynamic balanceweight the trust decision of our proposed trustmodel ismorescientific and flexible Furthermore themost important thingis the dynamic of the trust evaluationmodel Inmany currenttrust models [23 24] the trust value is updated by a slidingtime window using forgetting or aging mechanism againston-off attack and other malicious attacks but the number ofsliding time windows is predefined Once the number of thesliding time windows is confirmed it is difficult to adapt tothe dynamic changes of the network In our proposed trustmodel we centrally focus on setting up a dynamic updatemechanism To our knowledge there is no literature that candynamically adjust the number of the sliding time windowsand the parameters to achieve a dynamic update mechanismIn our proposed efficient dynamic trust model an updatemechanism is defined by a sliding time window based oninduced ordered weighted averaging operator (IOWA) toenhance flexibility We can dynamically adapt the parametersand the interactive history windows number to change the

weight sequence to meet the actual needs of the networkBased on the above analysis the proposed trust model canbe dynamically adjusted according to the environment andrequirements to achieve accurate trust evaluation and canrealize the identification and defense of various types ofmalicious attack It has a powerful capability of the trustestimation for WSNs

3 Network Model and Attack Model

31 Network Model In this paper we consider a scenario inwhich all the sensor nodes are randomly deployed in a two-dimensional space Nodes are neither added nor removedfrom network after deployment It assumes that sensor nodeshave the same capabilities of computing communicatingand storing initial energy level and communication rangeonly when the two nodes enter the communication range ofeach other to start communication Based on these assump-tions WSNs can be abstracted as a graph 119866 = (119881 119864) where119881 is the set of all nodes and 119864 is the set of all edges Each edge119890(119894 119895) isin 119864 denotes that the two nodes are located within eachotherrsquos transmission range Each node keeps a list of neigh-boring nodes which stores their unique ID communicationinformation and trust relationship It assumes that neitherthe source nor the destination is malicious Malicious nodesdo not collude themselves and all communication links arebidirectional And the communication channel is secure

32 Attack Model for WSNs With the open and remotedeployment environment WSNs are generally susceptible tovarious internal attacks including black hole attack worm-hole attack Sybil attack and grey-hole attack The attackbehavior of those malicious nodes shows diversity [6] suchas discarding routing packets injecting large amounts ofredundant information and error information maliciouslymodifying data packets and providing unreal recommen-dation trusted data information According to the attackbehavior and target of malicious attack we divide the variousattacks into three categories attacks on routing protocols inWSNs [32 33] attacks on communications data or messagesand attacks on trust models in WSNs [34] The target of thefirst kind of malicious attack is the routing protocol Themalicious attack behavior of this type discards all the routingpackets or drops part of the routing packets which makesthe data packets unable to be forwarded properly betweennodes This type of attack includes black hole attack grey-hole attack and wormhole attack Due to the vulnerabilityof the wireless communication channel the second typeof malicious attack is that the malicious nodes can easilycapture transmitting data information through a wirelesslink The target of this type of malicious attack is communi-cations data or messages The transmitting data can be easilyconducted with eavesdropping forgery and tamper Thistype of attack includes Dos attack and message tamperingattack The third type of malicious attack is a special kindof attack whose target is the trust management This typeof malicious attack can destroy the trust model by providingfalse information This type of attack includes on-off attackconflicting behavior selfish attack and bad-mouthing attack


Recommendation nodesNeighbor nodes

Node

Watchdog mechanism

Thecommunicationtrust

Dataconsistency

Theenergytrust

Direct trust

Trustedrecommendationnodes

Indirect trust

Dynamic weighting function

Update the trust value based on IOWA

Integrated trust value

End

Is it a trusted recommendation

node

Yes

Timer timeout

Figure 1 The process of the proposed trust model

As we know trust management system can deal with mostof the existing attacks and improve the security of thenetwork However it is difficult to detect these maliciousnodes completely by conventional trust modelThe proposedDTEM can evaluate trustworthiness of sensor nodes moreprecisely and identify the different malicious nodes moreeffectively

4 Overview of the Efficient Dynamic TrustEvaluation Model

To efficiently compute the trust values on sensor nodes wefirst need a clear understanding of the trust model processThe DTEM runs at the middleware of every sensor nodeEvery node maintains the trust value about other nodesthere is no central repository for storing trust values ofevery entity in the system The process of DTEM is shownin Figure 1 the direction of the arrow represents the flowof information between them And the specific process isas follows

(1) Information gathering we use watchdog mechanismto monitor neighboring nodesrsquo activities periodicallyas RFSN [18] to collect evidence information Theavailable neighbor nodesrsquo information is stored in therouting table of the node

(2) Trust value calculation in the proposed trust modelswhen a subject node wants to obtain the trust valueof an object it first checks its recorded list of neigh-bor nodes The direct trust and recommendationare used to evaluate the trustworthiness of sensornodes based on the recorded list The direct trust isdirectly calculated based on the communication dataconsistency and energy However due to maliciousattacks using only direct trust to evaluate sensornodes is not accurate Thus the recommendationfrom other sensor nodes is needed to improve thetrust evaluation Due to the dynamic behavior ofWSNs and the impact of some special maliciousattacks like on-off attack the calculation of trust valueshould be based upon history interaction records andupdated periodically

(3) Trust value integration to ensure that the trust modelmakes a decisionmore scientifically dynamically andadaptively we define a dynamic balance weight factorto realize the integration trust value of direct trust andindirect trust

Trust evaluation process is a complex process The infor-mation on a sensor nodersquos prior behavior is one of themost important aspects in trust model Hence every nodemaintains the trust value about other nodes in routing table


Direct trust moduleCalculate the communication trust based on data consistency

Calculate the energy trust

Calculate direct trust value

Integrated trust module

Establish a dynamic trust weightfunction

Calculate integrated trust valuebased on direct trust and indirecttrust

Establish dynamic weight based on IOWA

Trust update module

Update trust values with a time sliding window

Indirect trust module

Calculate indirect trust value

Select trustedrecommendation nodes

The joint neighbors between the evaluation node and the evaluated node

Figure 2 The relationship among the four modules

the process of the trust model is periodic The evaluationresults of trust values depend on historical records of eachnode

5 The Efficient Dynamic TrustEvaluation Model

In this section we present the composition of the proposedefficient dynamic trust evaluation model and the calculationprocedure of trust in detail

51The Composition of the Efficient Dynamic Trust EvaluationModel The efficient dynamic trust model proposed in thispaper consists of the following four modules direct trustmodule indirect trust module integrated trust module andtrust update module The direct trust is calculated basedon the Beta trust model The direct trust value of sensornode is calculated by taking communication trust datatrust and energy trust into account The indirect trust isevaluated based on the trusted recommendations from athird party And then the integrated trust is calculated byassigning adaptive dynamic balance weights for direct trustand indirect trust and combining them Finally we give anupdate mechanism by a sliding time window based on IOWAto complete the updating of direct trust value accordingto historical interaction records The relationship among

the four modules is shown in Figure 2 And the specificimplementation process is as follows

52 The Calculation of Direct Trust As we know the evalu-ation of trust in WSNs is a complex process which includesa lot of factors In the proposed trust model the direct trustvalue is calculated by taking communication trust data trustand energy trust into account The communication trustmeasures whether sensor nodes can cooperatively performthe intended protocol The data trust measures the trust ofdata created and manipulated by sensor node which canassess the fault tolerance and consistency of data The energytrust measures whether a node has enough residual energy tocomplete new communications and data processing tasksWeconsider the communication behavior and data consistencyto establish a trust environment against errors and attacksand combine the energy factor to prevent the low competitivenodes from participating in the network operation to ensurenetwork security and reliable operation Next we give thedetailed calculation procedure of direct trust

521 The Communication Trust Based on Data ConsistencyWe assume that a node can monitor neighboring nodesrsquoactivities within its communication range using watchdogmechanism [18] For example a node can monitor its neigh-borsrsquo transmissions and in this way we can detect whether


the node is forwarding or dropping the packets In mostcurrent trust models they define trust as the probabilitythat node 119894 holds on node 119895 to perform as expected [24]Assume that Beta distribution is employed as the prior dis-tribution of interactions among sensor nodes They monitorthe communication interaction record information based onwatchdog mechanism to count the number of well behaviorsor malicious behaviors between nodes to calculate the trustvalue The trust model at every node uses Beta probabilitydensity function [35] to evaluate expected probability of wellbehavior of neighboring nodes because trust modeling prob-lem is characterized by uncertainty And the Beta probabilitydensity function can be represented as

119891 (119909 | 120574 120573) = 1119861 (120574 120573)119909

120574minus1 (1 minus 119909)120573minus1 (1)

where 0 le 119909 le 1 120574 gt 0 120573 gt 0 and 120574 and 120573 are two indexedparameters

If the number of successful (well behavior) outcomes isrepresented by 119886 and 119887 represents the number of unsuccessful(malicious behavior) outcomes between nodes the probabil-ity for outcomes can be obtained by setting the values for 120574and 120573 as follows

120574 = 119886 + 1120573 = 119887 + 1 (2)

The probability expectation value for Beta probabilitydensity function is defined as

119864 (119909) = 120574120574 + 120573 =

119886 + 1119886 + 119887 + 2 (3)

Equation (3) can be used for the computation of the prob-ability expectation of successful outcomes between nodesThe probability expectation value is defined as the trust value

Based on the above discussion we assume that the wayof future interaction is the same as that of the previousone the probability expectation function can be representedby the mathematical expectation of Beta distribution as thecommunication trust In our proposed model the commu-nication trust denoted by 119879119862119894119895(119905) is derived from the directobservations of node 119894 on node 119895 at time 119905 which is definedas

119879119862119894119895 (119905) = 119878 + 1119878 + 119865 + 2 (1 minus

119865119882)(1 minus

1119878 + 120575) (4)

where 119878 and 119865 denote the total amount of successful andunsuccessful interactions between nodes 119894 and 119895 during 119905respectively But they are different from the above discussionthat just considered whether the node is forwarding or drop-ping the packets In (4) the nodersquos successful or unsuccessfulinteraction is accessed by communication behavior and dataconsistency together The expression (1 minus 119865119882) [24] is calledpunishment factor where 119882 is the total number of effectinteraction records The punishment factor punishes trustvalue by the dynamic change of the number of maliciousbehaviors between nodes The expression (1 minus 1(119878 + 120575)) is

called regulating function which approaches 1 rapidly withthe increasing of the number of successful interactions where120575 is a positive constant that can be tuned to control the speedof reaching 1 It would take longer time for a node to increaseits trust value for another node The detailed realization ofeach part is as follows

In (4) successful or unsuccessful communicationbetween nodes is judged by the communication behaviorsand data consistency togetherThe successful communicationmeans that a node not only forwards the packets to its nexthop neighbor but also requires forwarding the packetsreliably in its true form Otherwise it will be considered asunsuccessful communication The evaluated node forwardsthe packets to its next hop successfully based on thewatchdogmonitored [18] The data consistency based on the detectionalgorithm is as follows

Following the idea introduced in [36] the data trustaffects the trust of the sensor nodes that created and manip-ulated the data The data packets have spatial correlation inWSNs that is the packets sent among neighboring nodes arealways similar in the same area And the difference amongnodes is reduced to zero In our proposed paper we usethe detection algorithms [37] to assort abnormal and honestdata in the network The detection algorithm compares alocalized threshold 120582119894(119905) to the difference produced by eachnode data and its neighborsThenode 119894makes ameasurementindependently and transfers the measurement data value119909119894(119905) to its neighboring node 119895 Then node 119894 compares thedata value 119909119894(119905) to its neighborrsquos node 119895rsquos value 119909119895(119905) It isdenoted as normal if |119909119895(119905) minus 119909119894(119905)| lt 120582119894(119905) is satisfiedand else if |119909119895(119905) minus 119909119894(119905)| ge 120582119894(119905) is satisfied it is denotedas abnormal According to commutation behavior and dataconsistency we can get 119878 and 119865 which denote the totalamount of successful and unsuccessful interactions betweennodes during 119905 respectively

Considering that the malicious node injection false datahas a large deviation from the sensing data of authentic nodeswe can detect the attacker by comparing the threshold withthe difference between the neighbors and the node 119894 Theequation of the threshold 120582119894(119905) of each node 119894 as follows

120582119894 (119905) = 110038161003816100381610038161198731198941003816100381610038161003816 sum119895isin119873119894100381610038161003816100381610038161003816100381610038161003816119909119895 (119905) minus

119909119894 (119905) + sum119895isin119873119894 119909119895 (119905)10038161003816100381610038161198731198941003816100381610038161003816 + 1100381610038161003816100381610038161003816100381610038161003816 (5)

where 119873119894 represents the neighbor set of node 119894 The numberof element in119873119894 is denoted by |119873119894|

In (4) the punishment function shows strict punishmentto trust value according to the number of malicious behav-iors Once the number of misbehavior behaviors increasesthe trust value drops rapidly It reflects the following trustcharacter ldquotrust is easy to loserdquo It can quickly and accuratelyidentify the malicious behavior

We choose the regulating function in (4) instead of alinear function since such a function would approach veryslowly to 1 with the increasing of successful interactionswhich is similar to literature [20] According to the regulatingfunction it would take longer time for a node to increaseonersquos trust value This design can effectively restrain thetrust value rapidly increasing with the sudden increasing


number of successful communication interactions It reflectsthe following trust character ldquotrust is hard to acquirerdquo It canrestrain the collusion or on-off attack

522 The Energy Trust The energy trust is introducedmainly to complete the assessment on the performance of anode itself By introducing the energy factor we can effec-tively avoid the low competitiveness of nodes participating innetwork operation When the energy consumption of a nodeis lower than a certain energy threshold 119864th the node canonly complete simple basic operation to prolong the networklifetime and balance energy consumption betweennodesTheenergy trust is defined as

119879119864119895 (119905) = 0 if 119864res lt 119864th1 else

(6)

where 119864res represents the residual energy of a node 119864threpresents the energy threshold

523 The Direct Trust Based on the communication trust119879119862119894119895(119905) and the energy trust 119879119864119895(119905) we can obtain the directtrust between two neighbor nodes as

119879119863119894119895 (119905) = 119879119862119894119895 (119905) if 119879119864119895 (119905) = 105 lowast 119879119862119894119895 (119905) else 119879119864119895 (119905) = 0

(7)

53The Calculation of Indirect Trust Similar tomost existingrelated works we also consider the indirect trust to evaluatethe trust value in the proposed trust model The indirecttrust value is evaluated by the recommendations from a thirdparty The recommendations are composed of the commonneighboring node 119896 of node 119894 and node 119895 symbolized as119873119896As we know trust has the property of transitivity In the mostexisting trust models the trust value 119879V119894119895 from node 119894 to node119895 is calculated by the recommendation 119873V (119873V isin 119873119896) and isnotated as

119879V119894119895 = 119879119894V times 119879V119895 (8)

where 119879V119894119895 is the trust value of node 119894 to node V to node 119895 119879119894V isthe direct trust value of node 119894 to the common neighbor node119873V119879V119895 is the direct trust value of the common neighbor node119873V to 119895 But not all the recommendations are reliable How todetect and get rid ofmalicious recommendations is importantsince it has great impact on the calculation of trust In orderto judge the credibility of recommendations to calculateindirect trust more accurately it needs to detect dishonestrecommendations and exclude thembefore recommendationaggregation In our proposed model we only choose therecommendation whose trustworthiness is higher than thespecified trust threshold 120579 (0 le 120579 le 1) Suppose thereare 119896 recommendations and their direct trust values held bynode 119894 are notated as 1198791198941 1198791198942 119879119894119899 119879119894(119896minus1) 119879119894119896 If 119879119894119899 ge120579 (119899 = 1 2 119896) the recommendation from119873119899 is acceptedOtherwise it will be totally neglected

Due to the above discussion we can get the trustedrecommendations In our trust model we assign differentweights to the selected recommendations by their directtrust Intuitively we should give more weight to the selectedrecommendation from recommenders with high reputationHence we allocate weights based on the trust degree of therecommenders to avoid individual preference The followingapproach is taken to calculate the weight 120603119899 of the selectedrecommendation119873119899

120603119899 = 119879119894119899sum119902119899=1 119879119894119899 119899 = 1 2 119902 (9)

where 119879119894119899 is the direct trust value of node 119894 to the commonneighbor node 119873119899 and 119902 is the number of the selectedrecommendations And 0 le 120603119899 le 1 and sum119902119899=1 120603119899 = 1

Finally the indirect trust value denoted by 119879119868119894119895(119905) isobtained

119879119868119894119895 (119905) =119902

sum119899=1

120603119899 lowast 119879119899119894119895 119899 = 1 2 119902 (10)

where 120603119899 is the weight of 119879119899119894119895 119879119899119894119895 is obtained using (8) whichrepresents the trust value from node 119894 to node 119895 calculated bythe recommendation11987311989954 The Integration of Trust Value The integrated trust value119879119894119895(119905) which node 119894 holds about node 119895 at time 119905 is establishedvia the following formula

119879119894119895 (119905) = 120593119879119863119894119895 (119905) + (1 minus 120593) 119879119868119894119895 (119905) (11)

where 120593 is the balance weight factor which is referred to asself-confidence factor and 120593 isin [0 1] The value of 120593 is thedegree of recognition of the direct trust and indirect trust of119894 to 119895 120593 is defined using a dynamic function the functionintroduced in the literature [38] is given as

120593 = 119891 (119896) = 12 +1120587arctan(10 times

119896 minus COMth119873 ) (12)

where the balance weight factor 120593 is changed dynamicallywith the number of communication interactions 119896 to defectcaused by allotting weights subjectively 119873 is the maximuminteraction between 119894 and 119895 The specific value is determinedaccording to the network environment and the specificrequirement COMth is the threshold of communicationinteractionsWhen the communication interactions betweenevaluation node and evaluated node are higher than thethreshold COMth the weight factor becomes much moreand the integrated trust value is more dependent on thedirect trust value Otherwise the communication interac-tions between neighbor nodes are too small it is difficult todecide whether the evaluated node is good or bad and theintegrated trust value is more dependent on the indirect trustvalue We can dynamically adjust the importance of directtrust and indirect trust according to the dynamic weightfactor function

In the process of the integrated trust quantitative cal-culation we introduce a dynamic balance weight factor to


Time slot 1 Time slot m + 2

1 2 3 m m + 1 m + 2

middot middot middot


T1 T2 Tmminus1 Tm

T2 T3 Tm Tm+1

T3 T4 Tm+1 Tm+2

Figure 3 The sliding time window

solve the weight problem of direct trust and indirect trustThe balance weight is changed dynamically with the numberof communication interactions which reflects the dynamicadaptability of the trust computing

55 The Update of the Direct Trust In our proposed modelwe use a sliding time window to update the trust value [20]which can reflect the extent of variation of the trust value ina particular time interval

The sliding time window is used to update the directtrust value It consists of several time slots each slot is acycle time Only interactive records within the sliding timewindows are valid We define the length of sliding window as119898 which reflects evaluatorrsquos emphasis on historical recordsAs Figure 3 shows each sliding time window is divided into119898 slots from left to rightThe update process of the trust valuecan be shown as 1198791 1198792 119879119898minus1 119879119898 1198792 1198793 119879119898 119879119898+11198793 1198794 119879119898+1 119879119898+2 However it is intuitive that old histor-ical record has less contribution and new historical recordhas more influence on trust decision We give an updatemechanism using a sliding time window based on inducedordered weighted averaging operator (IOWA) [39] to solvethe problem of trust update We use the IOWA operator toobtain the weight of each time window which can give moreaccurate evaluation of the direct trust in DTEM

IOWA is defined as follows assume ⟨V1 1198861⟩ ⟨V2 1198862⟩ ⟨V119898 119886119898⟩ is two-dimensional array for119898 and119891119882 (⟨V1 1198861⟩ ⟨V2 1198862⟩ ⟨V119898 119886119898⟩)

=119898

sum119894=1

119908lowast119894 119886V-index(119894)(13)

The function 119891119882 is called the 119898-dimensional inducedordered weighted averaging operator generated by V1V2 V119898 abbreviated as IOWA operator and V-index(119894) isthe index of the V1 V2 V119894 V119898 in the order from large tosmall ones119882 = (119908lowast1 119908lowast2 119908lowast119898)119879 is the ordered weightedvector of IOWA and satisfies sum119898119894=1 119908lowast119894 = 1 0 le 119908lowast119894 le 1 119894 =1 2 119898

From (13) we can know that the value of 1198861 1198862 119886119898corresponds to the order in which the induced values ofV1 V2 V119898 in descending sorting are ordered weightedaverages the weight coefficient 119908lowast119894 has nothing to do withthe size and position of 119886119894 but is related to the location of theinduced value V119894 Hence the model can be used to sort the

historical trust value according to the time of occurrence andthe sliding time window is used as the induced value of theIOWA operator and the IOWA operator is completely usedto update the trust value

In accordance with the occurrence time the trust evalua-tion sequence of nodes 119894 on the node 119895 is expressed as follows119879 = 1198791 1198791 119879119905 119879119898 0 le 119879119905 le 1 1 le 119905 le 119898 inwhich 119879 is the trust evaluation sequence based on the slidingtime window Each value corresponds to a child windowand the time parameter is added to each of the 119879 elements⟨1199051 1198791⟩ ⟨1199052 1198792⟩ ⟨119905119898 119879119898⟩ and the two-dimensional trustsequence is defined as the induced value based on the timesliding window The trust update depends on both real-timeand historical windows to complete the updating operation Itmeans that we know ⟨1199051 1198791⟩ ⟨1199052 1198792⟩ ⟨119905119898 119879119898⟩ obtaining⟨119905119898+1 119879119898+1⟩ and this is what IOWA can solveThe definitioncan be expressed as

119879119898+1 = 119891119882 (⟨1199051 1198791⟩ ⟨1199052 1198792⟩ ⟨119905119898 119879119898⟩)

=119898

sum119894=1

119908lowast119894 119879V-index(119894)(14)

where 119908lowast119894 represents the importance of the trust value ofthe 119894th child window and sum119898119894=1 119908lowast119894 = 1 0 le 119908lowast119894 le1 119894 = 1 2 119898 So the ordered weighted vector 119882lowast =(119908lowast1 119908lowast2 119908lowast119898)119879 is 119879rsquos weight the key problem of theupdating mechanism of trust model is how to find theclassification weight of each window

Let 119882lowast = (119908lowast1 119908lowast2 119908lowast119898)119879 be the ordered weightedvector of any IOWA operator The perfect method of theweighted vector is based on the maximum degree of disper-sion [35] which can make full use of all the data informationunder given andor degree [40] We can get the weight vectoras follows

120572 = 119874119903119899119890119904119904 (119882lowast) = 1119898 minus 1

119898

sum119894=1

(119898 minus 119894) 119908lowast119894 (15)

119908lowast119895 = 119898minus1radic119908lowast1 119898minus119895119908lowast119898119895minus1 2 le 119895 le 119898 (16)

119908lowast1 [(119898 minus 1) 120572 + 1 minus 119898119908lowast1 ]119898= [(119898 minus 1) 120572]119898minus1 [((119898 minus 1) 120572 minus 119898)119908lowast1 + 1]

(17)

119908lowast119898 = ((119898 minus 1) 120572 minus 119898)119908lowast1 + 1

(119898 minus 1) 120572 + 1 minus 119898119908lowast1 (18)

In addition the relationship of trust has time decay In thetrust sequence of ⟨1199051 1198791⟩ ⟨1199052 1198792⟩ ⟨119905119898 119879119898⟩ the elementof ⟨119905119898 119879119898⟩ has the greatest influence on ⟨119905119898+1 119879119898+1⟩ and thelonger the time the smaller the influence on trust decisionIn (18) we can know that the calculation of the weightcoefficient vector is mainly determined by two parametersthe parameters 120572 and the number of interactive historywindows 119898 According to the literature [41] we know thatwhen 120572 isin [05 1] the distribution of the weight coefficientssatisfies the characteristic of time decay The correspondingweight sequences in different 119898 and 120572 are shown in Table 1


Table1Th

eweightsequenceindifferent

mand120572

120572=05

120572=06

120572=07

120572=08

120572=09

120572=1

119898=3

033330333303333

023840323304384

01540

0292105540

00819

0236305540

00263

014

7408263

000010

119898=4

025025025025

016710213302722

03474

00984

016

4702756

04614

004500106502520

05965

00103

0043401821

07641

00000010

119898=5

0202020202

0127801566019

20

0235302884

00706

010

86016

72

0257403962

00290

0059901240

0256505307

00051

00175006

02

0206807105

0000000010


Table 2 Simulation parameters

Parameter ValueInitial energyJ 05Initial trust value 05Packet lengthbit 2000dm 37Number of behaviors in each time unit 10Trust estimation periods 10Simulation times 1000120579 05119898 4120572 07

It can be seen from Table 1 that the value of the parameter120572 reflects the forgetting degree of the history interactiveexperience in trust model When 120572 is larger the historicalexperience is easier to forget And if 120572 = 1 the previoushistory is completely forgotten So we can dynamically adjust120572 to meet the different needs of the trust model The value ofparameter119898 responds to the number of windows of historicalexperience When 119898 is larger the trust value is obtainedmore accurately But it requires more energy consumptionand storage space So the determined parameter 119898 requiresa combination of the actual demand and considers therestriction of WSNs in accordance with the requirementsdefinition

6 Simulation Results and Analysis

Our experiments are performed using Matlab to analyzethe performance of the proposed algorithm similar to theliterature [25 29] The concrete simulation scene is setto be 100m times 100m with 50 randomly deployed nodesSome parameters vary with the scenes and the purposes ofexperiment and will be explained in detail The other defaultsimulation parameters that we have chosen are summarizedin Table 2

In this section the simulations can be divided into twoparts First we analyze the performance of the DTEM whichincludes the effect of dynamicweight value on integrated trustvalue and the direct trust value update mechanismThen wecompare DTEMwith the existing trust models typical RFSN[18] and BTMS [24]The results demonstrate that the DTEMhas a powerful capability of the trust estimation

61The Performance of the DTEM In this section we analyzethe dynamic performance of the DTEM

611 The Effect of DynamicWeights on Integrated Trust ValueThe value of 120593 is the degree of recognition of the direct trustand indirect trust of 119894 to 119895 The relationship between thedynamic weight of 120593 and the number of interaction times isshown in Figure 4 As shown in Figure 4 in the early stage oftrust measurement when the number of interactions is less

1 minus

The interaction times120010008006004002000

00

01

02

03

04

05

06

07

08

09

10

The v

alue

of w

eigh

t (

)

Figure 4 The dynamic weights of the direct values 120593 and 1 minus 120593

than COMth the trust calculation is much more dependenton the indirect trust valueWith the increasing of the numberof interactions when the number of interactions is greaterthan COMth the node 119894 is more willing to believe their directinteractive experience and the weight 120593 of direct trust willbecome much larger Hence the weight factor 120593 is dynam-ically changed with the interaction times And the value ofCOMth can be adjusted according to the actual demand ofthe network to achieve a more reasonable distribution of theweights between direct trust and indirect trust In this paperwe give119873 = 1200 andCOMth = 1198733 And giving120593= 01 0509 and the dynamic value the effect of 120593 on the integratedtrust value is shown in Figure 5 respectively As shown inFigure 5 at the beginning the interactions between nodesare very few the effect of dynamic weight 120593 on integratedtrust value is relatively close to 120593 = 01 That notes whenthe number of interactions between nodes is less the trustvalue is more dependent on the indirect trust and with theincreasing of interaction times the effect of dynamic weight120593 on integrated trust value slowly trends towards 120593 = 09That means with the increasing of the number of interactionsbetween nodes the integrated trust value metric measure ismore dependent on direct trust The results obtained are inagreement with the previous theoretical analysis The weightfactor120593 can be dynamically adjusted according to the numberof the interactions between nodes which can better ensurethe accuracy of trust measurement

612 The Effect of Update Mechanism on Direct Trust ValueFrom the analysis in Table 1 we can see that the IOWAoperator is determined by119898 and 120572 In this section the effectof different 119898 and 120572 on the direct trust value is realizedFigure 6 shows the effect of the different 120572 on the update trustvalue when119898 = 4 As shown in Figure 6 with the increasingof 120572 the update trust values are much closer to the trust valuewithout update which shows that the greater 120572 is the less


The interaction rounds100806040200

050

055

060

065

070

075

080

The i

nteg

ratio

n tr

ust v

alue

= 01

= 05

= 09

Dynamic

Figure 5 The effect of dynamic weights on integrated trust value


045

050

055

060

065

070

075

The d

irect

trus

t val

ue

m = 4 = 06

m = 4 = 07

m = 4 = 08

m = 4 = 09

The direct trust value

Figure 6 The effect of the parameter 120572 under119898 = 4

dependent the update trust value is on historical experienceFigure 7 shows the effect of the different119898 on the update trustvalue when 120572 = 07 As shown in Figure 7 with the increasingof 119898 the updated trust value is more accurate which showsthat the update trust value ismore dependent on the historicalexperience According to dynamic adjusting of 119898 and 120572 wecan effectively control the impact of historical interactions ontrust value and enhance the accuracy of trust value

62 Comparison of DTEM RFSN and BTMS In this sectionwe compare DTEMwith the existing trust models RFSN andBTMSThe former is one of the earliest classical trust schemes

045

050

055

060

065

070

075

The d

irect

trus

t val

ue

302520151050

The interaction rounds

= 07 m = 3

= 07 m = 4

= 07 m = 5


Figure 7 The effect of the parameter119898 under 120572 = 07

forWSNs the latter is one of the representative classical trustschemes

621The Trust Evaluation In this section we assess the inte-grated trust of normal node andmalicious node respectivelyIt is assumed that a normal node always chooses to cooperateand a malicious node always chooses not to cooperate Thetarget of this kind of attack is the routing protocol Asdepicted in Figure 8 the integrated trust increases with theincreasing of successful interactions among normal nodesanddecreaseswith the increasing of unsuccessful interactionsamong malicious nodes in RFSN BTMS and DTEM Onthe one hand we can see intuitively that for the integratedtrust between normal nodes the trust value increases fasterthan the other two algorithms in RFSN In BTMS the trustvalue increases more slowly than the other two algorithmsbecause of the effect of the punishing factor at the beginningIn our proposed model in DTEM the trust value changeswith the increasing number of the interaction rounds At thebeginning the trust value increases faster than BTMS andmore slowly than theRFSNAfter a few rounds the increasingof the trust value becomes the slowest On the other hand forthe integrated trust between malicious nodes in DTEM theintegrated trust value decreases fastest in all the algorithmsFrom the above analysis we can get that the DTEM reflectsthe following character ldquotrust is hard to acquire and easy toloserdquo Having compared RFSN BTMS and DTEM DTEMevaluates the trust more accurately among normal nodes Itreflects nodesrsquo commutation behavior changing acutely andhas more sensitive changing of the malicious actions whichcan effectively identify routing attacks

622 The Data Attack We analyze the efficacy of DTEMagainst faulty data and malicious data manipulation Thetarget of this type of malicious attack is communications dataor messages We generate a few common types of faults and


00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e

The interaction rounds60503010 20 400

RFSN normal nodeBTMS normal nodeDTEM normal node

RFSN malicious nodeBTMS malicious nodeDTEM malicious node

Figure 8 Comparison of the trust value of the normal node andmalicious node

fake data attacks together against the normal data Firstlywe generate random data at randomly selected sensor nodesbut it does not affect the data communication This meansthat although the sensor node sends false data it can beconsidered as successful communication Running RFSNBTMS and DTEM in this scene the result is shown inFigure 9 As shown in Figure 9 in BTMS and RFSN thetrust value does not make any change It is shown that thesetwo algorithms do not consider the consistency of the datathey just only consider communication behaviors betweennodes to calculate the trust value In DTEM the trust valuedecreases sharply and thenwith the increasing of the numberof interaction rounds the trust value increases but the trustvalue is always lower than 05 The result indicates thatthe resilience of DTEM is very good which can fast andaccurately identify the faulty data manipulation of maliciousnode It ismore sensitive in order to identify data informationattacks compared with RFSN and BTMS

623 The On-Off Attack In this section we analyze theefficacy of DTEM against the on-off attack This type ofmalicious attack is a special kind of attack whose targetis the trust management The on-off attack malicious nodealternates its behavior from malicious to normal and fromnormal to malicious so it remains undetected while causingdamage [42] In this paper we suppose that an on-off attackerbehaves well in the first 30 interaction rounds to build upgood reputation but behaves badly in the next 30 roundsAfter that it behaves well continuouslyThe result is shown inFigure 10 It is not difficult to see that the trust value increasesin the first 30 interaction rounds and themalicious node doesnothing or only performs well But in the next 30 rounds thetrust value drops when the malicious node launches attacksHaving compared RFSN BTMS and DTEM the DTEMcan acutely reflect nodesrsquo changing and sensitively detect

00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM

The interaction rounds1008060402010 30 50 70 900

Figure 9 Comparison of the trust value under data attack

RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue

Figure 10 Comparison of the direct trust value under on-off attack

on-off malicious attack And more importantly an on-offmalicious node recovers its trust valuemore slowly andmuchlongerWe can come to a conclusion that DTEMoutperformsRFSN and BTMS against on-off attack Meanwhile thesimulation result again verifies the character ldquotrust is hardto acquire and easy to loserdquo in DTEM This is becausethe trusted recommendation node selection mechanism anddynamic update mechanism are added to the process of trustevaluation which makes the evaluation of trust betweennodes more objective and accurate It can effectively identifytrust model attacks such as on-off attack and bad-mouthattack


Table 3 Comparison of state-of-the-art trust model

Trust model RFSN [18] GTMS [20] NBBTE [29] BTMS [24] DTEM (ours)Estimation method Probabilistic Weight Fuzzy logic Probabilistic Probabilistic

Direct trust module Transmission factorsdata factors

Transmissionfactors

Received packetsrate successfullysending packetsrate and so on

Transmission factorsTransmission factorsdata factors energy

factors

Indirect trust module Recommendationnodes

Recommendationnodes

Recommendationnodes

Trustedrecommendation

nodes


nodes

Integrated trust module Probability Weighted D-S evidence Self-confidence factor Dynamic weightingfunction

Trust update module Aging times times Sliding window Adjustable slidingwindow

Black hole attack radic radic radic radic radicAttack on information times radic times times radicOn-off attack times times radic radic radic

RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20

010 30 400 5020The proportion of malicious nodes ()

Figure 11 Comparison of the detection rate

624 The Detection Rate In this section the simulatedmalicious attacks are selective forwarding attack on-offattack conflicting behavior attack data forgery attack anddata tampering attack We vary the percentage of maliciousnodes from 10 to 50 percent with a 10 percent incrementAs shown in Figure 11 which gives the detection rate indifferent trust model we can see that the performance of theDTEM is better than RFSN and BTMS RFSN and BTMSare vulnerable against data forgery attack and data tamperingattack So with the increasing of the number of maliciousnodes the detection rate decreases rapidly but the DTEMkeeps high detection rate Hence the DTEM is an efficienttrust evaluation model which can identify different kinds ofmalicious nodes and can be dynamically adjusted accordingto the specific requirements of the network

In order to better illustrate the operation mechanism andperformance of DTEM Table 3 shows the comparison ofstate of the art in terms of trust estimation method directtrust factors indirect trust module integrated trust moduletrust update module and considered attacks Through theabove proof and analysis of the experiment we can knowthat DTEM is an efficient dynamic trust evaluationmodel forWSNs It can effectively identify various malicious attacks

7 Conclusion

In this paper we propose an efficient dynamic trust evalu-ation model for WSNs It includes the direct trust modulewith multiple trust factors the trusted recommendationtrust module the dynamic trust integration module andthe adjustable trust update module In the course of thecalculation of direct trust we take the communication trustdata consistency and energy trust into account it can achieveaccurate trust evaluation against routing attacks and datainformation attacks The punishment factor and regulatingfunction are introduced based on the character ldquotrust is hardto acquire and easy to loserdquo The calculation of indirect trustis invoked conditionally in order to enhance the accuracyof trust value which can be against the trust model attackssuch as bad-mouth attack Moreover in the process ofthe integrated trust quantitative calculation we define adynamic balance weight factor function to overcome thedefect caused by allotting weights subjectively Afterwardswe give the update mechanism based on IOWA to enhanceflexibility During this process we can dynamically adapt theparameters to change the weight sequence to meet the actualneeds of the network The proposed dynamic trust modelenables dynamic accurate and objective evaluation of trustbetween nodes based on the behavior of nodes

We have performed several tests to validate the proposedtrust model Simulation results indicate that DTEM is anefficient dynamic and attack-resistant trust evaluationmodelIt can dynamically evaluate the reputation among nodes


based on the communication behavior data consistency andenergy consumption of nodes And having compared RFSNBTMS and DTEM DTEM is of great help in defendingagainst routing attacks data information attacks and trustmodel attacks It can effectively identify various maliciousattacks

The traditional security mechanisms (cryptographyauthentication etc) are widely used to deal with externalattacks Trust model is a useful complement to the traditionalsecurity mechanism which can solve insider or nodemisbehavior attacks Hence trust model is importantto providing security service for upper layer networkapplication in WSNs such as secure routing and secure datafusion In our future work we would like to focus on theapplication of trust model in routing and data fusion forWSNs

Conflicts of Interest

The authors declare no conflicts of interest

Acknowledgments

This research is supported by the National Natural ScienceFoundation of China (61772101 61170169 61170168 and61602075)

References

[1] R D Gomes M O Adissi T A da Silva A C Filho MA Spohn and F A Belo ldquoApplication of wireless sensor net-works technology for inductionmotor monitoring in industrialenvironmentsrdquo in Intelligent Environmental Sensing vol 13 ofSmart Sensors Measurement and Instrumentation pp 227ndash277Springer International Publishing Cham 2015

[2] M M Alam D Ben Arbia and E Ben Hamida ldquoWearablewireless sensor networks for emergency response in publicsafety networksrdquo Wireless Public Safety Networks pp 63ndash942016

[3] M Bhuiyan G Wang J Wu et al ldquoDependable structuralhealth monitoring using wireless sensor networksrdquo IEEE Trans-actions on Dependable and Secure Computing pp 1ndash47 2015

[4] T Ojha S Misra and N S Raghuwanshi ldquoWireless sensornetworks for agriculture the state-of-the-art in practice andfuture challengesrdquoComputers and Electronics in Agriculture vol118 pp 66ndash84 2015

[5] G Han J Jiang L Shu J Niu and H-C Chao ldquoManagementand applications of trust in wireless sensor networks a surveyrdquoJournal of Computer and System Sciences vol 80 no 3 pp 602ndash617 2014

[6] H Modares A Moravejosharieh R Salleh and J LloretldquoSecurity overview of wireless sensor networkrdquo Life ScienceJournal vol 10 no 2 pp 1627ndash1632 2013

[7] R Lacuesta J Lloret M Garcia and L Penalver ldquoTwo secureand energy-saving spontaneous ad-hoc protocol for wirelessmesh client networksrdquo Journal of Network and Computer Appli-cations vol 34 no 2 pp 492ndash505 2011

[8] R Lacuesta J Lloret M Garcia and L Penalver ldquoA secureprotocol for spontaneous wireless Ad Hoc networks creationrdquo

IEEE Transactions on Parallel and Distributed Systems vol 24no 4 pp 629ndash641 2013

[9] J Cordasco and S Wetzel ldquoCryptographic versus trust-basedmethods for MANET routing securityrdquo Electronic Notes inTheoretical Computer Science vol 197 no 2 pp 131ndash140 2008

[10] M Momani and S Challa ldquoSurvey of trust models in differentnetwork domainsrdquo International Journal of Ad hoc Sensor ampUbiquitous Computing vol 1 no 3 pp 1ndash19 2010

[11] A Boukerch L Xu and K EL-Khatib ldquoTrust-based security forwireless ad hoc and sensor networksrdquo Computer Communica-tions vol 30 no 11-12 pp 2413ndash2427 2007

[12] F Ishmanov A S Malik S W Kim and B Begalov ldquoTrustmanagement system in wireless sensor networks design con-siderations and research challengesrdquo Transactions on EmergingTelecommunications Technologies vol 26 no 2 pp 107ndash1302015

[13] Y Liu C-X Liu and Q-A Zeng ldquoImproved trust managementbased on the strength of ties for secure data aggregation inwireless sensor networksrdquo Telecommunication Systems vol 62no 2 pp 319ndash325 2016

[14] X Anita M A Bhagyaveni and J M L Manickam ldquoCollabo-rative lightweight trust management scheme for wireless sensornetworksrdquoWireless Personal Communications vol 80 no 1 pp117ndash140 2015

[15] G Rajeshkumar and K R Valluvan ldquoAn Energy Aware TrustBased Intrusion Detection System with Adaptive Acknowl-edgement for Wireless Sensor Networkrdquo Wireless PersonalCommunications vol 94 no 4 pp 1993ndash2007 2016

[16] Y L Sun ZHan andK J R Liu ldquoDefense of trustmanagementvulnerabilities in distributed networksrdquo IEEE CommunicationsMagazine vol 46 no 2 pp 112ndash119 2008

[17] F Ishmanov S W Kim and S Y Nam ldquoA robust trustestablishment scheme for wireless sensor networksrdquo Sensorsvol 15 no 3 pp 7040ndash7061 2015

[18] S Ganeriwal and M B Srivastava ldquoReputation-based frame-work for high integrity sensor networksrdquo in Proceedings of the2nd ACMWorkshop on Security of Ad Hoc and Sensor Networks(SASN rsquo04) pp 66ndash77 October 2004

[19] HChenHWuX Zhou andCGao ldquoAgent-based trustmodelin wireless sensor networksrdquo in Proceedings of the 8th ACISInternational Conference on Software Engineering ArtificialIntelligence Networking and ParallelDistributed Computing(SNPD rsquo07) pp 119ndash124 IEEE Qingdao China August 2007

[20] R A ShaikhH Jameel B J drsquoAuriol H Lee S Lee andY SongldquoGroup-based trust management scheme for clustered wirelesssensor networksrdquo IEEE Transactions on Parallel and DistributedSystems vol 20 no 11 pp 1698ndash1712 2009

[21] J Song X Li J Hu G Xu and Z Feng ldquoDynamic trustevaluation of wireless sensor networks based on multi-factorrdquoin Proceedings of the 14th IEEE International Conference onTrust Security and Privacy in Computing and CommunicationsTrustCom 2015 pp 33ndash40 fin August 2015

[22] X Li F Zhou and J Du ldquoLDTS a lightweight and dependabletrust system for clustered wireless sensor networksrdquo IEEETransactions on Information Forensics and Security vol 8 no6 pp 924ndash935 2013

[23] D He C Chen S Chan J Bu and A V Vasilakos ldquoReTrustattack-resistant and lightweight trust management for medicalsensor networksrdquo IEEE Transactions on Information Technologyin Biomedicine vol 16 no 4 pp 623ndash632 2012


[24] R Feng X Han Q Liu and N Yu ldquoA credible bayesian-based trust management scheme for wireless sensor networksrdquoInternational Journal of Distributed Sensor Networks vol 2015Article ID 678926 2015

[25] G Zhan W Shi and J Deng ldquoSensorTrust a resilient trustmodel for wireless sensing systemsrdquo Pervasive and MobileComputing vol 7 no 4 pp 509ndash522 2011

[26] H-H Dong Y-J Guo Z-Q Yu and C Hao ldquoA wireless sensornetworks based on multi-angle trust of noderdquo in Proceedingsof the International Forum on Information Technology andApplications (IFITA rsquo09) vol 1 pp 28ndash31 May 2009

[27] J Jiang G Han F Wang L Shu and M Guizani ldquoAn efficientdistributed trust model for wireless sensor networksrdquo IEEETransactions on Parallel and Distributed Systems 2014

[28] H Rathore V Badarla and S Shit ldquoConsensus-aware sociopsy-chological trust model for wireless sensor networksrdquo ACMTransactions on Sensor Networks vol 12 no 3 article no 212016

[29] R Feng X Xu X Zhou and J Wan ldquoA trust evaluationalgorithm forwireless sensor networks based on node behaviorsand D-S evidence theoryrdquo Sensors vol 11 no 2 pp 1345ndash13602011

[30] T Zhang L Yan and Y Yang ldquoTrust evaluation method forclustered wireless sensor networks based on cloud modelrdquoWireless Networks pp 1ndash21 2016

[31] S Shen L Huang E Fan K Hu J Liu and Q Cao ldquoTrustdynamics in WSNs an evolutionary game-theoretic approachrdquoJournal of Sensors vol 2016 Article ID 4254701 10 pages 2016

[32] J-W Ho M Wright and S K Das ldquoDistributed detection ofmobile malicious node attacks in wireless sensor networksrdquo AdHoc Networks vol 10 no 3 pp 512ndash523 2012

[33] C Karlof and D Wagner ldquoSecure routing in wireless sensornetworks attacks and countermeasuresrdquo inProceedings of the 1stIEEE International Workshop on Sensor Network Protocols andApplications pp 113ndash127 May 2003

[34] Y L Sun Z Han W Yu and K J R Liu ldquoA trust evaluationframework in distributed networks Vulnerability analysis anddefense against attacksrdquo in Proceedings of the INFOCOM 200625th IEEE International Conference on Computer Communica-tions esp April 2006

[35] A Joslashsang and R Ismail ldquoThe Beta Reputation Systemrdquo inProceedings of the 15th Bled Electronic Commerce Conference(Bled EC) pp 324ndash337 Slovenia 2002

[36] E Elnahrawy and B Nath ldquoCleaning and querying noisysensorsrdquo in Proceedings of the the 2nd ACM internationalconference p 78 San Diego CA USA September 2003

[37] S Mi H Han C Chen J Yan and X Guan ldquoA secure schemefor distributed consensus estimation against data falsification inheterogeneouswireless sensor networksrdquo Sensors vol 16 no 22016

[38] B Zhao H Jing-Sha E Y Zhang X et al ldquoAnalysis of multi-factor in trust evaluation of open networkrdquo Journal of ShandongUniversity vol 49 no 9 pp 103ndash108 2014

[39] R R Yager ldquoInduced aggregation operatorsrdquo Fuzzy Sets andSystems vol 137 no 1 pp 59ndash69 2003

[40] R Fuller and P Majlender ldquoAn analytic approach for obtainingmaximal entropy OWA operator weightsrdquo Fuzzy Sets andSystems vol 124 no 1 pp 53ndash57 2001

[41] X-Y Li and X-L Gui ldquoCognitive model of dynamic trustforecastingrdquo Ruan Jian Xue BaoJournal of Software vol 21 no1 pp 163ndash176 2010

[42] L F Perrone and S C Nelson ldquoA study of on-off attackmodels for wireless ad hoc networksrdquo in Proceedings of the 20061st Workshop on Operator-Assisted (Wireless-Mesh) CommunityNetworks OpComm 2006 deu September 2006

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of


International Journal of

RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal of

Volume 201

Submit your manuscripts athttpswwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



captured nodes [10] In pursuit of the security of WSNs trustand reputation mechanisms have proven to be more resilientagainst insider or node misbehavior attacks [10 11]

Trust in the field of wireless communication networksmay be defined as the degree of belief on the future behaviorof other nodes which is based on past experience andobservations of the nodesrsquo action [11] So we give thedefinition of trust in WSNs as follows node 119860rsquos trust innode 119861 describes the belief or expectation or assuranceof sincerity competence and integrity of node 119861rsquos futureactionbehavior [12] The basic idea of trust based scheme isto quantify trust to describe the trustworthiness reliabilityor competence of individual nodes [5] Trust managementsystem can be implemented in various applications forsecurity management such as secure protocol [8] secure dataaggregation [13] trusted routing [14] and intrusion detectionsystem [15] In recent years lots of state-of-the-art models[5 12ndash31] have been proposed in this field Undoubtedly thepresent achievements have greatly promoted related researchin improving security of WSNs Even so trust evaluationin WSNs is still a challenging issue Some limitations areexhibited which need more attention to be solved

Considerable research has been done on modeling andmanaging trust and reputation in WSNs Many currentstudies [18ndash20] have been done for trust establishmentjust only based on the communication interaction recordsbetween nodes without considering the data consistency sothey cannot be against attacks on data While other studies[25ndash28] combine multifactors to calculate the trust valuethe multitrust sums up in weighted manner to computethe integrated trust But the weights are obtained by expertopinion method or average weight method The results ofthe prediction are subjective which affect the scientific andflexibility of the trust decision In addition trust evaluationis a dynamic phenomenon and changes with time andenvironment condition In many current trust models [2324] the trust value is updated by a sliding time windowusing forgetting or aging mechanism But the number ofsliding windows is defined by expert opinion method Oncethe number of the sliding time windows is confirmed it isdifficult to change It makes the trust models unable to adaptto the dynamic changes of the network environment whichaffects the accuracy of the result To our knowledge thereis no literature that can dynamically adjust the number ofthe sliding time windows and the parameters to achieve adynamic update mechanism Moreover some existing trustmodels [18 23 27] rarely consider the influence of theenergy consumption Due to these reasons there is a growingdemand for adequate provision of an efficient dynamic trustevaluation model for WSNs it can achieve accurate trustevaluation dynamically according to the environment andrequirements and can realize the identification and defenseof various types of malicious attack

In this paper an efficient dynamic trust evaluation model(DTEM) for WSNs is proposed that aims to address theabove problems In the proposed trust model the trust valueis calculated considering multitrust factors it can achieveaccurate trust evaluation Moreover DTEM can dynamicallyadjust the weights of direct trust and indirect trust It

reflects the dynamic adaptability of the trust computing Italso can dynamically adjust the parameters of the updatemechanism to update the trust value to meet the actualneeds of the network environmentTheDTEM can be againstvarious types of malicious attack and can be configured andeffectively applied to different environments with differentrequirementsThemajor contributions of this paper are listedas follows

(1) To improve the accuracy of trust evaluation againstattacks on data the trust value is calculated con-sidering direct trust and indirect trust The directtrust is calculated considering multitrust includingcommunication trust data trust and energy trustwith the punishment factor and regulating function tomeet the following character ldquotrust is hard to acquireand easy to loserdquo The indirect trust is evaluatedconditionally by the trusted recommendations froma third party

(2) To ensure that the trust model makes a decisionmore scientifically dynamically and adaptively wedefine a dynamic balance weight factor functionwhich is changed dynamically with the number ofcommunication interactions The adaptive dynamicbalance weight factor dynamically adjusts the weightof the direct trust and indirect trust

(3) To make sure that the proposed trust model canbe configured and effectively applied to differentenvironmentswith different requirements and againston-off attack we give an update mechanism bya sliding time window based on induced orderedweighted averaging operator (IOWA) to enhance flex-ibility We can dynamically adapt the parameters andthe interactive history windows number to changethe weight sequence to update the trust value toadapt with different environments and requirementsMeanwhile on-off attacks can be handled efficiently

At last compared to the existing trust models (RFSN [18]and BTMS [24]) simulation results show that the proposedtrust model has remarkable enhancements in the accuracy oftrust decision and has a better capability to capture dynamicmalicious nodes behaviors

The remainder of this paper is organized as followsSection 2 gives an overview of related works Network modeland attack model are described in Section 3 Section 4 givesthe overview and process of the DTEM Section 5 discussestrust model and trust evaluation mechanism in DTEM InSection 6 the experiment is made under simulative envi-ronments and the performance of the DTEM is evaluatedSection 7 concludes this paper

2 Related Works

In most current trust model researches focus on sensor radiocommunication behaviors Sensor nodes build node trustmodel through wireless radio transaction with neighboringnodes Ganeriwal and Srivastava [18] first proposed a repu-tation based framework for sensor networks (RFSN) where















Node

Watchdog mechanism


Dataconsistency

Theenergytrust

Direct trust


Indirect trust




End


node

Yes

Timer timeout

















Trust update module
















119891 (119909 | 120574 120573) = 1119861 (120574 120573)119909




120574 = 119886 + 1120573 = 119887 + 1 (2)


119864 (119909) = 120574120574 + 120573 =

119886 + 1119886 + 119887 + 2 (3)



119879119862119894119895 (119905) = 119878 + 1119878 + 119865 + 2 (1 minus

119865119882)(1 minus

1119878 + 120575) (4)






120582119894 (119905) = 110038161003816100381610038161198731198941003816100381610038161003816 sum119895isin119873119894100381610038161003816100381610038161003816100381610038161003816119909119895 (119905) minus

119909119894 (119905) + sum119895isin119873119894 119909119895 (119905)10038161003816100381610038161198731198941003816100381610038161003816 + 1100381610038161003816100381610038161003816100381610038161003816 (5)







119879119864119895 (119905) = 0 if 119864res lt 119864th1 else

(6)



119879119863119894119895 (119905) = 119879119862119894119895 (119905) if 119879119864119895 (119905) = 105 lowast 119879119862119894119895 (119905) else 119879119864119895 (119905) = 0

(7)


119879V119894119895 = 119879119894V times 119879V119895 (8)



120603119899 = 119879119894119899sum119902119899=1 119879119894119899 119899 = 1 2 119902 (9)



119879119868119894119895 (119905) =119902

sum119899=1

120603119899 lowast 119879119899119894119895 119899 = 1 2 119902 (10)


119879119894119895 (119905) = 120593119879119863119894119895 (119905) + (1 minus 120593) 119879119868119894119895 (119905) (11)


120593 = 119891 (119896) = 12 +1120587arctan(10 times

119896 minus COMth119873 ) (12)





1 2 3 m m + 1 m + 2



T1 T2 Tmminus1 Tm

T2 T3 Tm Tm+1

T3 T4 Tm+1 Tm+2






=119898

sum119894=1

119908lowast119894 119886V-index(119894)(13)





119879119898+1 = 119891119882 (⟨1199051 1198791⟩ ⟨1199052 1198792⟩ ⟨119905119898 119879119898⟩)

=119898

sum119894=1

119908lowast119894 119879V-index(119894)(14)



120572 = 119874119903119899119890119904119904 (119882lowast) = 1119898 minus 1

119898

sum119894=1

(119898 minus 119894) 119908lowast119894 (15)



(17)





Table1Th


mand120572

120572=05

120572=06

120572=07

120572=08

120572=09

120572=1

119898=3

033330333303333

023840323304384

01540

0292105540

00819

0236305540

00263

014

7408263

000010

119898=4

025025025025

016710213302722

03474

00984

016

4702756

04614

004500106502520

05965

00103

0043401821

07641

00000010

119898=5

0202020202

0127801566019

20

0235302884

00706

010

86016

72

0257403962

00290

0059901240

0256505307

00051

00175006

02

0206807105

0000000010










1 minus


00

01

02

03

04

05

06

07

08

09

10

The v

alue

of w

eigh

t (

)






050

055

060

065

070

075

080

The i

nteg

ratio

n tr

ust v

alue

= 01

= 05

= 09

Dynamic



045

050

055

060

065

070

075

The d

irect

trus

t val

ue

m = 4 = 06

m = 4 = 07

m = 4 = 08

m = 4 = 09





045

050

055

060

065

070

075

The d

irect

trus

t val

ue

302520151050


= 07 m = 3

= 07 m = 4

= 07 m = 5







00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e







00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM



RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue







Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation























Node

Watchdog mechanism


Dataconsistency

Theenergytrust

Direct trust


Indirect trust




End


node

Yes

Timer timeout

















Trust update module
















119891 (119909 | 120574 120573) = 1119861 (120574 120573)119909




120574 = 119886 + 1120573 = 119887 + 1 (2)


119864 (119909) = 120574120574 + 120573 =

119886 + 1119886 + 119887 + 2 (3)



119879119862119894119895 (119905) = 119878 + 1119878 + 119865 + 2 (1 minus

119865119882)(1 minus

1119878 + 120575) (4)






120582119894 (119905) = 110038161003816100381610038161198731198941003816100381610038161003816 sum119895isin119873119894100381610038161003816100381610038161003816100381610038161003816119909119895 (119905) minus

119909119894 (119905) + sum119895isin119873119894 119909119895 (119905)10038161003816100381610038161198731198941003816100381610038161003816 + 1100381610038161003816100381610038161003816100381610038161003816 (5)







119879119864119895 (119905) = 0 if 119864res lt 119864th1 else

(6)



119879119863119894119895 (119905) = 119879119862119894119895 (119905) if 119879119864119895 (119905) = 105 lowast 119879119862119894119895 (119905) else 119879119864119895 (119905) = 0

(7)


119879V119894119895 = 119879119894V times 119879V119895 (8)



120603119899 = 119879119894119899sum119902119899=1 119879119894119899 119899 = 1 2 119902 (9)



119879119868119894119895 (119905) =119902

sum119899=1

120603119899 lowast 119879119899119894119895 119899 = 1 2 119902 (10)


119879119894119895 (119905) = 120593119879119863119894119895 (119905) + (1 minus 120593) 119879119868119894119895 (119905) (11)


120593 = 119891 (119896) = 12 +1120587arctan(10 times

119896 minus COMth119873 ) (12)





1 2 3 m m + 1 m + 2



T1 T2 Tmminus1 Tm

T2 T3 Tm Tm+1

T3 T4 Tm+1 Tm+2






=119898

sum119894=1

119908lowast119894 119886V-index(119894)(13)





119879119898+1 = 119891119882 (⟨1199051 1198791⟩ ⟨1199052 1198792⟩ ⟨119905119898 119879119898⟩)

=119898

sum119894=1

119908lowast119894 119879V-index(119894)(14)



120572 = 119874119903119899119890119904119904 (119882lowast) = 1119898 minus 1

119898

sum119894=1

(119898 minus 119894) 119908lowast119894 (15)



(17)





Table1Th


mand120572

120572=05

120572=06

120572=07

120572=08

120572=09

120572=1

119898=3

033330333303333

023840323304384

01540

0292105540

00819

0236305540

00263

014

7408263

000010

119898=4

025025025025

016710213302722

03474

00984

016

4702756

04614

004500106502520

05965

00103

0043401821

07641

00000010

119898=5

0202020202

0127801566019

20

0235302884

00706

010

86016

72

0257403962

00290

0059901240

0256505307

00051

00175006

02

0206807105

0000000010










1 minus


00

01

02

03

04

05

06

07

08

09

10

The v

alue

of w

eigh

t (

)






050

055

060

065

070

075

080

The i

nteg

ratio

n tr

ust v

alue

= 01

= 05

= 09

Dynamic



045

050

055

060

065

070

075

The d

irect

trus

t val

ue

m = 4 = 06

m = 4 = 07

m = 4 = 08

m = 4 = 09





045

050

055

060

065

070

075

The d

irect

trus

t val

ue

302520151050


= 07 m = 3

= 07 m = 4

= 07 m = 5







00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e







00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM



RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue







Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation

















Trust update module
















119891 (119909 | 120574 120573) = 1119861 (120574 120573)119909




120574 = 119886 + 1120573 = 119887 + 1 (2)


119864 (119909) = 120574120574 + 120573 =

119886 + 1119886 + 119887 + 2 (3)



119879119862119894119895 (119905) = 119878 + 1119878 + 119865 + 2 (1 minus

119865119882)(1 minus

1119878 + 120575) (4)






120582119894 (119905) = 110038161003816100381610038161198731198941003816100381610038161003816 sum119895isin119873119894100381610038161003816100381610038161003816100381610038161003816119909119895 (119905) minus

119909119894 (119905) + sum119895isin119873119894 119909119895 (119905)10038161003816100381610038161198731198941003816100381610038161003816 + 1100381610038161003816100381610038161003816100381610038161003816 (5)







119879119864119895 (119905) = 0 if 119864res lt 119864th1 else

(6)



119879119863119894119895 (119905) = 119879119862119894119895 (119905) if 119879119864119895 (119905) = 105 lowast 119879119862119894119895 (119905) else 119879119864119895 (119905) = 0

(7)


119879V119894119895 = 119879119894V times 119879V119895 (8)



120603119899 = 119879119894119899sum119902119899=1 119879119894119899 119899 = 1 2 119902 (9)



119879119868119894119895 (119905) =119902

sum119899=1

120603119899 lowast 119879119899119894119895 119899 = 1 2 119902 (10)


119879119894119895 (119905) = 120593119879119863119894119895 (119905) + (1 minus 120593) 119879119868119894119895 (119905) (11)


120593 = 119891 (119896) = 12 +1120587arctan(10 times

119896 minus COMth119873 ) (12)





1 2 3 m m + 1 m + 2



T1 T2 Tmminus1 Tm

T2 T3 Tm Tm+1

T3 T4 Tm+1 Tm+2






=119898

sum119894=1

119908lowast119894 119886V-index(119894)(13)





119879119898+1 = 119891119882 (⟨1199051 1198791⟩ ⟨1199052 1198792⟩ ⟨119905119898 119879119898⟩)

=119898

sum119894=1

119908lowast119894 119879V-index(119894)(14)



120572 = 119874119903119899119890119904119904 (119882lowast) = 1119898 minus 1

119898

sum119894=1

(119898 minus 119894) 119908lowast119894 (15)



(17)





Table1Th


mand120572

120572=05

120572=06

120572=07

120572=08

120572=09

120572=1

119898=3

033330333303333

023840323304384

01540

0292105540

00819

0236305540

00263

014

7408263

000010

119898=4

025025025025

016710213302722

03474

00984

016

4702756

04614

004500106502520

05965

00103

0043401821

07641

00000010

119898=5

0202020202

0127801566019

20

0235302884

00706

010

86016

72

0257403962

00290

0059901240

0256505307

00051

00175006

02

0206807105

0000000010










1 minus


00

01

02

03

04

05

06

07

08

09

10

The v

alue

of w

eigh

t (

)






050

055

060

065

070

075

080

The i

nteg

ratio

n tr

ust v

alue

= 01

= 05

= 09

Dynamic



045

050

055

060

065

070

075

The d

irect

trus

t val

ue

m = 4 = 06

m = 4 = 07

m = 4 = 08

m = 4 = 09





045

050

055

060

065

070

075

The d

irect

trus

t val

ue

302520151050


= 07 m = 3

= 07 m = 4

= 07 m = 5







00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e







00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM



RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue







Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











119891 (119909 | 120574 120573) = 1119861 (120574 120573)119909




120574 = 119886 + 1120573 = 119887 + 1 (2)


119864 (119909) = 120574120574 + 120573 =

119886 + 1119886 + 119887 + 2 (3)



119879119862119894119895 (119905) = 119878 + 1119878 + 119865 + 2 (1 minus

119865119882)(1 minus

1119878 + 120575) (4)






120582119894 (119905) = 110038161003816100381610038161198731198941003816100381610038161003816 sum119895isin119873119894100381610038161003816100381610038161003816100381610038161003816119909119895 (119905) minus

119909119894 (119905) + sum119895isin119873119894 119909119895 (119905)10038161003816100381610038161198731198941003816100381610038161003816 + 1100381610038161003816100381610038161003816100381610038161003816 (5)







119879119864119895 (119905) = 0 if 119864res lt 119864th1 else

(6)



119879119863119894119895 (119905) = 119879119862119894119895 (119905) if 119879119864119895 (119905) = 105 lowast 119879119862119894119895 (119905) else 119879119864119895 (119905) = 0

(7)


119879V119894119895 = 119879119894V times 119879V119895 (8)



120603119899 = 119879119894119899sum119902119899=1 119879119894119899 119899 = 1 2 119902 (9)



119879119868119894119895 (119905) =119902

sum119899=1

120603119899 lowast 119879119899119894119895 119899 = 1 2 119902 (10)


119879119894119895 (119905) = 120593119879119863119894119895 (119905) + (1 minus 120593) 119879119868119894119895 (119905) (11)


120593 = 119891 (119896) = 12 +1120587arctan(10 times

119896 minus COMth119873 ) (12)





1 2 3 m m + 1 m + 2



T1 T2 Tmminus1 Tm

T2 T3 Tm Tm+1

T3 T4 Tm+1 Tm+2






=119898

sum119894=1

119908lowast119894 119886V-index(119894)(13)





119879119898+1 = 119891119882 (⟨1199051 1198791⟩ ⟨1199052 1198792⟩ ⟨119905119898 119879119898⟩)

=119898

sum119894=1

119908lowast119894 119879V-index(119894)(14)



120572 = 119874119903119899119890119904119904 (119882lowast) = 1119898 minus 1

119898

sum119894=1

(119898 minus 119894) 119908lowast119894 (15)



(17)





Table1Th


mand120572

120572=05

120572=06

120572=07

120572=08

120572=09

120572=1

119898=3

033330333303333

023840323304384

01540

0292105540

00819

0236305540

00263

014

7408263

000010

119898=4

025025025025

016710213302722

03474

00984

016

4702756

04614

004500106502520

05965

00103

0043401821

07641

00000010

119898=5

0202020202

0127801566019

20

0235302884

00706

010

86016

72

0257403962

00290

0059901240

0256505307

00051

00175006

02

0206807105

0000000010










1 minus


00

01

02

03

04

05

06

07

08

09

10

The v

alue

of w

eigh

t (

)






050

055

060

065

070

075

080

The i

nteg

ratio

n tr

ust v

alue

= 01

= 05

= 09

Dynamic



045

050

055

060

065

070

075

The d

irect

trus

t val

ue

m = 4 = 06

m = 4 = 07

m = 4 = 08

m = 4 = 09





045

050

055

060

065

070

075

The d

irect

trus

t val

ue

302520151050


= 07 m = 3

= 07 m = 4

= 07 m = 5







00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e







00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM



RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue







Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












119879119864119895 (119905) = 0 if 119864res lt 119864th1 else

(6)



119879119863119894119895 (119905) = 119879119862119894119895 (119905) if 119879119864119895 (119905) = 105 lowast 119879119862119894119895 (119905) else 119879119864119895 (119905) = 0

(7)


119879V119894119895 = 119879119894V times 119879V119895 (8)



120603119899 = 119879119894119899sum119902119899=1 119879119894119899 119899 = 1 2 119902 (9)



119879119868119894119895 (119905) =119902

sum119899=1

120603119899 lowast 119879119899119894119895 119899 = 1 2 119902 (10)


119879119894119895 (119905) = 120593119879119863119894119895 (119905) + (1 minus 120593) 119879119868119894119895 (119905) (11)


120593 = 119891 (119896) = 12 +1120587arctan(10 times

119896 minus COMth119873 ) (12)





1 2 3 m m + 1 m + 2



T1 T2 Tmminus1 Tm

T2 T3 Tm Tm+1

T3 T4 Tm+1 Tm+2






=119898

sum119894=1

119908lowast119894 119886V-index(119894)(13)





119879119898+1 = 119891119882 (⟨1199051 1198791⟩ ⟨1199052 1198792⟩ ⟨119905119898 119879119898⟩)

=119898

sum119894=1

119908lowast119894 119879V-index(119894)(14)



120572 = 119874119903119899119890119904119904 (119882lowast) = 1119898 minus 1

119898

sum119894=1

(119898 minus 119894) 119908lowast119894 (15)



(17)





Table1Th


mand120572

120572=05

120572=06

120572=07

120572=08

120572=09

120572=1

119898=3

033330333303333

023840323304384

01540

0292105540

00819

0236305540

00263

014

7408263

000010

119898=4

025025025025

016710213302722

03474

00984

016

4702756

04614

004500106502520

05965

00103

0043401821

07641

00000010

119898=5

0202020202

0127801566019

20

0235302884

00706

010

86016

72

0257403962

00290

0059901240

0256505307

00051

00175006

02

0206807105

0000000010










1 minus


00

01

02

03

04

05

06

07

08

09

10

The v

alue

of w

eigh

t (

)






050

055

060

065

070

075

080

The i

nteg

ratio

n tr

ust v

alue

= 01

= 05

= 09

Dynamic



045

050

055

060

065

070

075

The d

irect

trus

t val

ue

m = 4 = 06

m = 4 = 07

m = 4 = 08

m = 4 = 09





045

050

055

060

065

070

075

The d

irect

trus

t val

ue

302520151050


= 07 m = 3

= 07 m = 4

= 07 m = 5







00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e







00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM



RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue







Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











1 2 3 m m + 1 m + 2



T1 T2 Tmminus1 Tm

T2 T3 Tm Tm+1

T3 T4 Tm+1 Tm+2






=119898

sum119894=1

119908lowast119894 119886V-index(119894)(13)





119879119898+1 = 119891119882 (⟨1199051 1198791⟩ ⟨1199052 1198792⟩ ⟨119905119898 119879119898⟩)

=119898

sum119894=1

119908lowast119894 119879V-index(119894)(14)



120572 = 119874119903119899119890119904119904 (119882lowast) = 1119898 minus 1

119898

sum119894=1

(119898 minus 119894) 119908lowast119894 (15)



(17)





Table1Th


mand120572

120572=05

120572=06

120572=07

120572=08

120572=09

120572=1

119898=3

033330333303333

023840323304384

01540

0292105540

00819

0236305540

00263

014

7408263

000010

119898=4

025025025025

016710213302722

03474

00984

016

4702756

04614

004500106502520

05965

00103

0043401821

07641

00000010

119898=5

0202020202

0127801566019

20

0235302884

00706

010

86016

72

0257403962

00290

0059901240

0256505307

00051

00175006

02

0206807105

0000000010










1 minus


00

01

02

03

04

05

06

07

08

09

10

The v

alue

of w

eigh

t (

)






050

055

060

065

070

075

080

The i

nteg

ratio

n tr

ust v

alue

= 01

= 05

= 09

Dynamic



045

050

055

060

065

070

075

The d

irect

trus

t val

ue

m = 4 = 06

m = 4 = 07

m = 4 = 08

m = 4 = 09





045

050

055

060

065

070

075

The d

irect

trus

t val

ue

302520151050


= 07 m = 3

= 07 m = 4

= 07 m = 5







00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e







00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM



RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue







Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Table1Th


mand120572

120572=05

120572=06

120572=07

120572=08

120572=09

120572=1

119898=3

033330333303333

023840323304384

01540

0292105540

00819

0236305540

00263

014

7408263

000010

119898=4

025025025025

016710213302722

03474

00984

016

4702756

04614

004500106502520

05965

00103

0043401821

07641

00000010

119898=5

0202020202

0127801566019

20

0235302884

00706

010

86016

72

0257403962

00290

0059901240

0256505307

00051

00175006

02

0206807105

0000000010










1 minus


00

01

02

03

04

05

06

07

08

09

10

The v

alue

of w

eigh

t (

)






050

055

060

065

070

075

080

The i

nteg

ratio

n tr

ust v

alue

= 01

= 05

= 09

Dynamic



045

050

055

060

065

070

075

The d

irect

trus

t val

ue

m = 4 = 06

m = 4 = 07

m = 4 = 08

m = 4 = 09





045

050

055

060

065

070

075

The d

irect

trus

t val

ue

302520151050


= 07 m = 3

= 07 m = 4

= 07 m = 5







00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e







00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM



RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue







Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation


















1 minus


00

01

02

03

04

05

06

07

08

09

10

The v

alue

of w

eigh

t (

)






050

055

060

065

070

075

080

The i

nteg

ratio

n tr

ust v

alue

= 01

= 05

= 09

Dynamic



045

050

055

060

065

070

075

The d

irect

trus

t val

ue

m = 4 = 06

m = 4 = 07

m = 4 = 08

m = 4 = 09





045

050

055

060

065

070

075

The d

irect

trus

t val

ue

302520151050


= 07 m = 3

= 07 m = 4

= 07 m = 5







00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e







00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM



RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue







Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











050

055

060

065

070

075

080

The i

nteg

ratio

n tr

ust v

alue

= 01

= 05

= 09

Dynamic



045

050

055

060

065

070

075

The d

irect

trus

t val

ue

m = 4 = 06

m = 4 = 07

m = 4 = 08

m = 4 = 09





045

050

055

060

065

070

075

The d

irect

trus

t val

ue

302520151050


= 07 m = 3

= 07 m = 4

= 07 m = 5







00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e







00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM



RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue







Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










00

01

02

03

04

05

06

07

08

09

The t

rust

valu

e







00

01

02

03

04

05

06

07

08

09

10

The t

rust

valu

e

RFSNBTMSDTEM



RFSNBTMSDTEM


00

01

02

03

04

05

06

07

08

09

10

The d

irect

trus

t val

ue







Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













Transmissionfactors



factors


Recommendationnodes

Recommendationnodes


nodes


nodes




RFSNBTMSDTEM

The d

etec

tion

rate

()

100

80

60

40

20





7 Conclusion








Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation














Acknowledgments


References













































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation





























RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









an efficient dynamic trust evaluation model for wireless

Documents