Secure Routing Path Using Trust Values for Wireless Sensor Networks
IInternational Journal on Cryptography and Information Security (IJCIS), Vol. 4, No. 2, June 2014DOI:10.5121/ijcis.2014.4203 27 S ECURE R OUTING P ATH U SING T RUST V ALUES F OR W IRELESS S ENSOR N ETWORKS
Dr.S.Rajaram , A. Babu Karuppiah , K. Vinoth Kumar Department of ECE, Velammal College of Engineering and Technology
ABSTRACT
Traditional cryptography-based security mechanisms such as authentication and authorization are noteffective against insider attacks like wormhole, sinkhole, selective forwarding attacks, etc., Trust basedapproaches have been widely used to counter insider attacks in wireless sensor networks. It provides aquantitative way to evaluate the trustworthiness of sensor nodes. An untrustworthy node can wreakconsiderable damage and adversely affect the quality and reliability of data. Therefore, analyzing the trustlevel of a node is important. In this paper we focused about indirect trust mechanism, in which each nodemonitors the forwarding behavior of its neighbors in order to detect any node that behaves selfishly anddoes not forward the packets it receives. For this, we used a link state routing protocol based indirecttrusts which forms the shortest route and finds the best trustworthy route among them by comparing thevalues of all the calculated route trusts as for each route present in the network. And finally, we compareour work with similar routing protocols and show its advantages over them. K EYWORDS
Wireless Sensor Networks (WSNs), Routing, Benevolent Node, Malicious Node, Trust Management
1. I
NTRODUCTION
A Wireless sensor network (WSN) is a collection of millimeter-scale, self-contained, micro-electro-mechanical devices. These tiny devices have sensors, computational processing ability(i.e. CPU power), wireless receiver and transmitter technology and a power supply. WSNs arelimited in power, computational capacity and memory, and may not have global IDs. WSNs havea wide range of applications, ranging from monitoring environments, military zones, sensitiveinstallations and remote data collection and analysis.Trust establishment among nodes is a must to evaluate the trustworthiness of other nodes and isone of the most critical issues in WSNs. Trust is dependent on time; it can increase or decreasewith time based on the available evidence through direct interactions with the node orrecommendations from other trusted nodes. Trust-modeling is mathematical representation ofnode ‟ s opinion of another node in a network. We need mathematical tools to represent trust andreputation, update these continuously. Maintaining a record of the transactions with other nodes,directly as well as indire ctly, from this record a „trust ‟ value will be established [1]. nternational Journal on Cryptography and Information Security (IJCIS), Vol. 4, No. 2, June 2014 28 Trust means everybody is trusted somehow and does not require any authentication (lessoverhead). It tells the degree of reliability. Every node finds the trust of all other nodes, based onprevious experience and recommendations in fulfilling its promises.
2. S
ECURITY I SSUES I N W SNS
There are some successful trust schemes which can perform well such as Trust Management forResilient Geographic Routing (T-RGR), Reputation-based Framework for Sensor Networks(RFSN) and Group-based Trust Management Scheme (GTMS), but several important problemsstill need to be solved.(1) The first problem is, WSNs are made up of resources limited nodes. Energy is critical forWSNs. Nevertheless, few of trust schemes take into consideration the node ‟ s residual energywhich can affect the work of the trust schemes. In general, when a node has a higher trust valuein the sponsor node ‟ s trust table, it may be apt to be selected as a service provider in a trustscheme. Then this can make the selected node consume more energy than other nodes andexhaust the energy prematurely. We learn that taking into consideration the node ‟ s residualenergy will avoid this flaw and prolong the network lifetime and balance the network load.(2) The second problem is, when one node sends reputation request messages to its neighbor toget the monitored node ‟ s indirect trust value, this can cost a large amount of additional energy.Consider this case, in a group of n nodes, taking the RFSN trust model for an example let ‟ scompute the amount of data packets that the nodes need to send out during getting indirect trustinformation. When node A wants to interact with node B, it needs to send out n-2 request packetsand then receive n-2 replying packets, and the total number is 2(n-2). If node A wants to interactwith n-1 nodes, the amount of packets is 2(n-1)(n-2).When all the nodes in the group update alltrust value in their trust tables, so the maximum amount of the packets is 2n(n-1)(n-2). Now, wechoose one node as a header, which is the trust information collection centre. The rest nodes inthe group periodically report reputation messages to the header. The header collects andaggregates all the messages. The total number of the packets sent out in this process is n-1. Whennode A wants to communicate with node B, it just only need to send the reputation requestmessage to the header and receives the reply message from the header. This will cost only 2packets. If node A wants to communicate with n-2 nodes, there are 2(n- 2) packets all together.When all the nodes in the group update all the trust value of their neighbors, so the maximumamount of the packets is 2(n-2)(n-1)+n-1=(2n-3)(n-1). According to the energy consumptionanalysis proposed by H. O. Tan and I. Korpeoglu, once the size of packet and the distance aredecided, the more number of packets are sent out, the more energy will be consumed. So the newmethod can cost less energy than the method described in RFSN, especially in the large-scaleWSNs. At the same time, the header can aggregate the information, so it can reduce the memoryconsumption. And it can be easy to manage the trust of the mobile nodes as well. The newmethod may incur the header to exhaust energy quickly that is not beneficial for the networklifetime and the header will be easy to become the aim of the attacker. Altering the headerproperly and regularly may be a good method of solving this problem in this new method. nternational Journal on Cryptography and Information Security (IJCIS), Vol. 4, No. 2, June 2014 29 The third problem is, when the node integrates the indirect trust value from its neighbor nodesinto the mixed trust value, it needs to pay attention on the reliability of the indirect trust value justreceived to avoid bad mouthing attack. In order to keep the WSNs secure, the node will receiverecommendation information only from those nodes who have higher trust value in its trusttables, exceeding a predefined threshold (e.g. above 0.9). When it integrates all of therecommendation information, it allocates a high weight to the information sent from these nodeswho have higher trust value. The higher trust value is, the higher weight it will be allocated. Butthere will be a flaw. Let us consider such kind of the node when it is selected to be a next routingnode, it can service well, e.g. transmitting the received packet accurately. But when it is asked toprovide the indirect trust information about the monitored node, it will send a false trustinformation to interfere the sponsor node to make accurate decisions. This kind of the node iscalled malicious spies[9]. These nodes can provide requested services and serve well, they arebenign nodes in the sponsor node ‟ s opinion. So the indirect trust information obtained from thosemalicious spies will be received as normal information. In RFSN and GTMS models, they allreceive the introduction information from the nodes who has higher trust value. Obviously theyare vulnerable to bad mouthing attack launched by malicious spies. To solve this defect, we firsttry to distinguish the confidence deposited in a node as a recommender and the trust deposited inthe same node as a service provider [9]. But this method is not enough. We also consider to adoptfiltering scheme that is used to filter interferential recommendation which has large differencewith others in the all information and then punish or award depending on the filtering result.Isolating the malicious spies may consume more energy and resource. Maybe we can solve thisproblem by considering the benefit of the recommended nodes. We realize that the maliciousspies will cost energy and bear the risk of being found as well as getting the benefit when theytake participate in recommending. We can infer that when the cost is greater than income, it willnot send out false indirect information; when the cost is less than income, it will send out falseindirect information. This relationship between cost and income is a game relationship, so usinggame theory to solve the problem will be a good choice.
3. T
RUST M ODELING
A trust model can be defined as the representation of the trustworthiness of each node in theopinion of another node, thus each node associates a trust value with every other node.
Figure 1. Combined Trust Mechanismnternational Journal on Cryptography and Information Security (IJCIS), Vol. 4, No. 2, June 2014 30
As illustrated in Figure, When the direct information can meet certain trust requirement, it usesthe direct value, otherwise it will use the indirect trust value. Three different kinds informationhave their own advantages and disadvantages. Firstly, the direct trust value which is generated bythe sponsor node itself is considered to be completely trustable and is the simple and energysaving. At the same time, it can prevent the bad mouthing attacks. On the other hand, the singledirect trust value can make the information uncompleted and unsymmetrical. Sometimes thedirect trust information can ‟ t accurately reflect the actual situation of the nodes. The indirect trustinformation is formed by the sponsor requiring and integrating. The indirect information alone isvery rare in the WSNs. Omitting the indirect trust information can make the decision madedepending on the trust value not be consistent with the actual state of the network[13]. Secondly,the indirect information can be used to initiate the new node ‟ s trust value[12]. We learn that theindirect information can help to solve the problem of the information uncompleted incurred byadopting the direct information alone. But requiring the indirect information can consume amountof additional of energy. This is critical for the sensor network lifetime and may bring themalicious nodes ‟ bad mouthing attacks. We can infer that the direct information and the indirectinformation can supplement each other.The updating of the trust value has two parts. The first one is the aging. We can give the differentweights to the past trust value and the present trust value depending on the actual requirementsand different destinations. If we give the present trust value a higher weight, it will always keepthe node in the normal state. If we give the past trust value a higher weight, it will prevent thenode ‟ s deception. Because the malicious nodes can work well in a short term to promote trustvalue. The second one is updating periodically. It can keep the routing lines stable and meet thecharacteristics of sensor nodes[14].The sensor nodes need to sleep in a timely manner to saveenergy.
4. R
ELATED W ORK
Trust Management in WSNs is a relatively new concept which is still in nascent stage. Althoughsome research papers are available for computing the trust value of the nodes, a very limited worktowards their inclusion in the data routing path has been reported so far. In ATSR [1], the TotalTrust (TT) information of a node is calculated by combining DT and IT information of the node.Finally the routing decisions are taken based on geographical information as well as TTinformation. The disadvantage is that the nodes needed to calculate DTs and ITs for all othernodes require high processing and storage capabilities. In DTLSRP [2], Total Direct Trust value(TDT) is calculated by geometric mean of all trust metrics. A threshold value (TTH) is fixeddepending on application. The nodes having TDT higher than TTH are considered as benevolentnodes and rest are selfish nodes., Link State Routing protocol is used to find all available paths.The multiplicative Route Trusts are calculated for each path. The highest routing path trust valueis chosen as best routing path. DTLSRP simulations perform better result with respect to otherprotocols such as ATSR, etc. This method allows us to find the shortest path without applyingDijkstra ‟ s algorithm. However it does not includeIndirect Trust (IT). TILSRP [3] is a new algorithm for calculation of both DT and IT ofindividual nodes based on LSR protocol. The selfish nodes are eliminated from the network andRoute Trusts (RTs) are calculated thereby. nternational Journal on Cryptography and Information Security (IJCIS), Vol. 4, No. 2, June 2014 31
5. P
ROPOSED W ORK
The indirect trust (IT) value is important mainly for newly initialized nodes or mobile nodes thathave recently arrived in a new neighborhood. If a node is found in different neighborhoods eachtime it needs to send a new message, then direct interaction provides no trust info. In this case, iftrust information is exchanged between nodes and especially if this exchange happens morefrequently than the data message transmission, the source node has gathered (indirectly) trustinformation before it sends out a new data message. Thus, the exchange of indirect trust is usefulmainly when node mobility has to be supported.An important design option for calculating indirect trust information exchange is the frequency ofthe exchange. The exchange frequency directly affects the speed of trust information. To limit theexchanged information, only positive (or negative) information has been proposed to be shared[5]. However, when only positive information is shared, since nodes learn only from their ownexperience about a malicious node, colluding malicious nodes can extend each other ‟ s survivaltime through false praise reports. If all N one-hop neighbors are asked triggering the generation ofN reputation response messages, the network load would be significantly burdened (increasingcollision probability) and the node resource (memory, processing and energy) consumption wouldalso increase significantly.From the analysis of direct trust mechanism, packet losses are severely occurred. So that weproposed one algorithm for obtaining secured routing path in WSNs. The algorithm is based uponboth the direct and indirect trust exchange information. Both the information gives the securedand reliable path for routing. In this model, every node monitors all one-hop neighbors to checkwhether they forward the messages they receive towards the sink node. The ratio between theactually forwarded and all the transmitted messages provides an indication, whether the selectedneighbor is sincerely cooperating. Values close to1 indicate a benign node, while values close to0 indicate malicious nodes. This information is then used to calculate the routing function, basedon which the next hop node is selected. Suppose we take 23 wireless sensor nodes, to explain this algorithm. Although all the nodes areassumed to be homogeneous and uniform in hardware and routing capabilities initially, howevertheir different parameters vary with time in a pre-defined manner in the example taken. Routingof sensed (or collected) data in this case can be initiated via event or time triggering. The methodof routing, in the present case is developed independently of the rules associated with suchtriggering. Every packet of information to be routed definitely comprises of a source and a sink.They are denoted here by red and yellow colours. Streams of packets are sent from the Source (orN19) to the Destination (or N7) as shown in the figure below. nternational Journal on Cryptography and Information Security (IJCIS), Vol. 4, No. 2, June 2014 32Figure 2. Routes from Source to Destination with node numbers as selected by LSRP
We are using the fundamental routing protocol named Link State Routing Protocol with theexclusion of Dijkstra ‟ s algorithm. This protocol is particularly attractive in the case of WirelessSensor Networks which have limited hardware and software features. The redundancy onapplying Dijkstra ‟ s algorithm here reduces the routing overhead. Its absence in calculation of theshortest path is compensated using our Select the Most Trusted Route (SMTR) algorithm whichchooses the most reliable (or trusted) route. R R R R R R indicates the available routesfrom the source to destination.R : S->N14->N9->N6->N3->DR : S->N9->N6->N3->DR : S->N15->N10->N11->DR : S->N15->N11->DR : S->N15->N16->DR : S->N22->N20->N17->N12->D Algorithm:
For calculating the trust of the node.
Initial condition:
Each node wants to communicate with other nodes in the network.
Input:
Get the source and destination of the network.
Output:
Trust value calculation and communication.
Begin:
Apply Link State Routing Protocol. Get the Node id ‟ s N=1 nternational Journal on Cryptography and Information Security (IJCIS), Vol. 4, No. 2, June 2014 33 Mark the sensor nodes in each of the routes Calculate the direct trust by Calculate the indirect trust by Calculate the total trust by Get the path trust by Select route N If PTN>PTi for all i ≠N Path through Route N ElseN=N+1 and go back to previous step
Notations: DT y (X): Direct trust value of node Y in node XIT i (X): Indirect Trust value of Node X with respect to node iW j : Weightage given to the direct trust with respect to particular neighboring nodes used incalculation of indirect trustTT: Total TrustPT: Path Trust
5. S
IMULATION R ESULTS
To verify the performance of new algorithm, simulation is done through MATLAB. nternational Journal on Cryptography and Information Security (IJCIS), Vol. 4, No. 2, June 2014 34Figure 3. Trusted path for Routing from analysis
There are 50 nodes randomly uniformly distributed in the simulation environment, and the area isa circle with radius of 100 meters. The initial energy of each node is 5J and about every 1 second,each node generates a data packet with packet size 500bit. Here equivalent trust values of 0.5were given in prior to some 30 fully powered and working homogeneous nodes and they sentmessages in proportional to those value. It is clear from the fact that in absence of any receivingor sending nodes the indirect trust value of a node is extremely small. Then the indirect trustvalue is 0. The matlab simulation and performance analysis of this model is given above figure 3.
Fig 4. Performance analysis of TRITS compared with ATSR, DTLSRP and TILSRP
Our proposed method TRITS has better performance as compared to the existing systems ATSR,DTLSRP and TILSRP as shown in above figure 4. nternational Journal on Cryptography and Information Security (IJCIS), Vol. 4, No. 2, June 2014 35
6. C
ONCLUSIONS
Taking routing as main objective, proposed routing mechanism dedicated for wireless sensornetworks. In our proposed method, the new algorithm SRPT (Secure Routing Path using Trustvalues) has better performance as compared to existing systems. Here, in this approach, duringtransmission of packets, if any node in the routing path get fails to transmit the packets. That timeit can automatically chooses the another routing path to transmit the packets to the requireddestination.
7. R
EFERENCES [1] S. S.
Babu, A. Raha, and M.K. Naskar, “A Direct Trust Dependent Link State Routing Protocol UsingRoute Trusts for WSNs (DTLSRP)”, Wireless Sensor Network Journal, USA, pp. 125 -134, April2011.[2] Theodore Zahariadis, Helen Leligou, Panagiotis Karkazis, Panagiotis Trakadas, Ioannis
Papaefstathiou, Charalambos Vangelatos, Lionel Besson, “Design and implementation of a trust - aware routing protocol for large WSNs”, Int. Journal of Network Security & Its Applications, July s, H. C. Leligou, S.Voulotis, S. Maniatis, P. Trakadas, and P. Karkazis, “An energy and trust- aware routing protocol for large wireless sensor networks”, World Scientific and Engineering Academy and Society. Wisconsin, USA, 2009, pp. 216-224.[5] A. Raha,
M.K. Naskar, S.S. Babu, Omar Alfandi and Dieter Hogrefe, “Trust Integrated Link State
Routing Protocol for Wireless Sensor Network (TILSRP), published in Proc. of 5thIEEE ANTS 2011,Dec 2011.[6] H. C. Leligou, P. Karkazis, Th. Zahariadis, T. Velivasaki, Th. Tsiodras, C. Matrakidis, S. Voliotis, “The Impact of Indirect Trust Information Exchange on Network Performance and EnergyConsumption in Wireless Sensor Networks”, ELMAR, 2011 proceedings, pp 153 -156.[7]
Youngho Cho, Gang Qu, Yuanming Wu, “Insider
Threats against Trust Mechanism with Watchdog and Defending Approaches in Wireless Sensor Networks”, Security and privacy workshops (SPW),
Yan (Lindsay) Sun, Zhu Han, and K. J. Ray Liu, “Defense of Trust Manageme nt Vulnerabilities in
Distributed Networks”, IEEE Communications Magazine, Vol 46, Issue 2, 2008, pp.112 -119.[9]
Jianqiao GAO, Julong PAN*, Zhanyi XU, “Towards Trust -based Security Mechanism for Wireless
Sensor Networks”, WICON conf on 2011, pp 1 -5.[10]
Theodore Zahariadis, Helen C. Leligou, Panagiotis Trakadas and Stamatis Voliotis, “MobileNetworks Trust Management in Wireless Sensor Networks”, European Transactions on
Telecommunications, 2010; 21:386-395.nternational Journal on Cryptography and Information Security (IJCIS), Vol. 4, No. 2, June 2014 36
Authors
Dr. S. Rajaram, working as Associate Professor of ECE Department of ThiagarajarCollege of Engineering, Madurai, obtained his B.E., degree Thiagarajar College ofEngineering from Madurai Kamaraj Univeristy, Madurai and M.E. degree from A.C.C.E.TKaraikudi. He was awarded PhD by Madurai Kamaraj University in the field of VLSIDesign. He was awarded PDF from Georgia Institute of Technology, USA in 3D VLSI. Hehas 16 years of experience of teaching and research. His area of research includes VLSIDesign, Network Security, Wireless Networks. He has published many research papers inInternational journals, national and international conferences.
A. Babu Karuppiah , Assistant Professor of ECE Department of Velammal College ofEngineering & Technology, Madurai, obtained his B.E., degree from AdhiparasakthiEngineering College, Madras University and M.E. degree from Mepco Schlenk Engg.College, Sivakasi that is affiliated to Anna University, Chennai. He has 11 years ofTeaching and Research experience. Pursuing PhD in Anna University, Tirunelveli inNetworking. He published many research papers in International journals, National andInternational conferences. His area of research includes Wireless Sensor Networks.