Security Enhancement of AODV Protocol using Fuzzy based Trust Computation in Mobile Ad Hoc Networks

Mobile ad hoc network (MANEt) possess self-configuration, self-control and self-maintenance capabilities. Nodes of MANEt are autonomous routers. Hence, they are vulnerable to security attacks. collaborative attacks such as black hole and wormhole in MANEt are difficult to be detected and prevented. trust based routing decision is an effective approach for security enhancement in MANEt. In this study, trust computingusing fuzzy based max-product composition scheme is applied to compute aggregated trust values to determine malicious nodes and thereby safe route in MANEts. the results show performance improvement ofproposed protocol over AOdV protocol. Network metrics are analysed under different mobility conditions and different positions of black hole nodes.


INtROductION
Wireless local area networks (WLAN) present unique and global way of networking with mobile nodes.WLAN were based on IEEE standards 802.11 a,b,g standards 1 .But, in most of the configurations of WLAN, only the last link is connected with access point thereby acting as wireless.Access point itself may be considered as bottleneck because of its limited range.there are many application areas such as battlefield communication, urban sensing and vehicular networking, where spontaneous communication is needed, WLAN may not be suitable, and hence infrastructure less networks are required.
Mobile ad hoc networks (MANEts) are infrastructure less network.In case of MANEts, the nodes have capabilities ofself-configuration, lack of central control and self-maintenance 2 .collaboration among MANEt nodes is a problematic issue for forwarding packets 2 as the nodes are autonomous routers.MANEt nodes themselves forwards data packets among each other hence are vulnerable to many security attacks.Mobility of the nodes increases its vulnerability towards many collaborative attacks.Black hole and wormhole are two collaborative attacks on MANEt.
Black hole node tries to show that there is shortest path towards destination node just sending Route Reply (RREP) immediately upon receiving Route Request (RREG) 3 .Black hole may become cooperative attack, when two or more nodes participate in the attack.
Worm hole is a cooperative attack, in which two malicious nodes form a fast tunnel between them.thereby, both attackers collude together and fabricate a false route and cheat other nodes 4 .
this problematic issue of collaboration among MANEt nodes can be resolved using trust based computing.
In MANEt, trustworthy routes can be established by eliminating malicious nodes using trust based computing approach 3 .trust can be computed directly by a node for the other node or indirectly, when a node A recommends node B to the node c. trust of a non-neighbouring node can only be computed indirectly.Hence, indirect trust plays a vital role in computing the overall trust of a node, which needs methods of trust propagation and trust aggregation.Hence, our approach is based on enhancing trust propagation and trust aggregation in order to evaluate trust, so that trustworthy routes can be established.trust computing is by default a fuzzy approach, as both of these are probabilistic approach.Hence, we are applying fuzzy approach in the proposed work.this paper proposes fuzzy based trust computation to mitigate black hole attack in MANEts.thus enhancing security of AOdV routing protocol in MANEtsmodel required to secure routing protocol of MANEts.We discuss related work in section II.We propose fuzzy based trust computation model to detect black hole attacks in section III.In section III, we also present several trust relation properties useful for trust based computation in MANEts.Section IV presents results and performance evaluation of simulated fuzzy based trust computation system.Finally conclusion is given in section V.

Related Work
HoanLan et.al. in 5 have demonstrated that the impact of black hole attack is catastrophic and it is malicious node position dependent.If the malicious node is near the source node then it has most damage to the network performance.the network performance is also highly dependent on the speed of mobile node.

Weighted binary Relational Fuzzy Trust Model
trust should be a necessary element of distributed systems which depends on relationship between different entities of the distributed systems 5 .trust is a reliance of one entity to the other.It depends on the first entity that how much it believes in second.An entity may rely fully on the other entity.But in practical scenario this is not possible.therefore, trusts can be modelled as a probabilistic value, which can be denoted as a fractional value between 0 and 1.Hence, trust computation approach is by default a fuzzy approach.
trust can be modelled mathematically as a binary relation on A×A, where A is a set of nodes in MANEt.this binary relation is weighted as the weights of this relation are fractional trust values of one node to the other node.these weights represent the extent to which a node believes in other node.

Trust based Computing
Nodes of MANEt have to trust other nodes of same network.But, all the nodes are not equally trustworthy.Some nodes are selfish, some might be malicious and others might be completely trustworthy.Hence, trusted computation should be used to detect the behaviour of node.trust computation in static networks is straightforward as trust values vary only with the behaviour of the node.After some observations behaviours and trust values are predictable 2 .trust computation in mobile networks is considerably difficult as compared to static networks, as compromised node may move after attack and it will be very difficult to detect such malicious node 2 .Network topology significantly changes within time in a volatile manner.Hence, observations for neighbouring node are difficult.Behaviour of a node is predictable only after enough number of observations.Furthermore, it is difficult to associate a mobile node with its location and gaining observations.MANEts are peer-to peer networks, there is absence of centralized control station and observing the behaviour of node becomes very complicated.

1.
Trust Formulation trust is to be formulated by one entity for the other.this formulation might be guessing an opinion of other entity.Mathematically, trust is probability of trustworthiness of one entity about the other.In case of MANEt, it is required to compute the trust value by one node for the other node.MANEt is an open network; any node can join and leaves the network at any time.When a new node joins the network, the trust with some default value is initialized.For the new node, the default value for the trust will depend upon the application where MANEt is used.the trust will keep on changing over the time based on the feedback obtained from other nodes.trust computation of node about neighbouring node will depend upon certain parameters.these parameters may include, packet delivery ratio of a node, percentage of energy exhaustion of node, percentage of Buffer utilized by the node and number of connection request given by node.

Trust Propagation
A node computes trust for a target node and transmits that trust value to its neighbouring nodes, so that neighbouring nodes save time and resources of recomputing the trust values for the same node

Trust Aggregation
Nodes of MANEt are propagating trust values to their neighbours.Node might get multiple values of trust for any target node.So, aggregation of trust is often needed to be computed.Aggregate value of trust is to be calculated via trust path.Malicious node in between the trust path can change the values of propagated trust.So, multiple paths for the aggregated trust are to be considered.In our earlier research 12 , we have proposed direct trust computation method as depicted in eq (2)   ... (2)   Such that Where Node 'A' calculates trust of node 'B' based on the parameters like packet forwarding ratio of node, energy exhaustion, buffer utilization and number of connection request.Final trust value is calculated based on aggregated trust and direct trust, which is depicted in eq (3).t(i,j)=Aggtrust(i,j)*trust factor + directtrust(i,j)*(1trust factor ) ... (3)   Where 0<trust_factor<1

METHODOLOGY
the proposed protocol trusted Fuzzy Ad hoc on demand distance vector routing protocol (tFAOdV) uses tMPcF approach to calculate aggregated trust.this aggregated trust with direct trust between the nodes are used to calculate trust value as depicted in (3). the trust value derived is used in the proposed protocol tFAOdV.tFAOdV uses trust computation approach, which is based on trust formulation, trust propagation and trust aggregation.Nodes get direct trust values from their neighboring nodes and aggregated trust values from non-neighboring nodes.Afterwards, nodes compute aggregated trust value using trust aggregation.this aggregated trust value is used to detect malicious node.the classification of node is completed using fuzzy discrimination table tab 1.If the trust value drops to fuzzy level of low and very low, then that node will be considered as malicious and discarded from routing.If the aggregated trust value false in the range of medium, high and very high then the node will be used in routing activities.

Performance Analysis Performance Metrics
three performance metrics are assumed for analysis of tFAOdV protocol Packet delivery ratio (PdR), Average End-to-end delay (Etd) and throughput.these network metrics are evaluated for both the protocols AOdV and tFAOdV under varying network conditions.
the simulation was carried out using NS2 as simulator with input parameters as given in table 2.
Various scenario files and network animations files are created by varying the number of nodes, malicious node's position, speed and duration of simulation.Finally, the output in terms of traces is generated for each scenario for both protocols.Awk file is used to evaluate network performance metrics.outperformed AOdV in terms of PdR under this case.Furthermore, tMPcF's PdR under low mobility of 1 m/s is also outperforming stationary node's PdR may be because malicious node may not be in completely contact with victim nodes always, which is possible in stationary node case.Hence, network performance due to malicious nodes may be improved in low mobility case as compared to stationary case.Same behavior is also observed in case of throughput and Etd parameters as seen in fig 6 and 9.   node.Performance under these scenarios shows, that when malicious nodes are near to the source node, performance deteriorates because malicious node is in contact with the source node and drop packet immediately it receives.But, in this case too tMPcF's performance saw some enhancements over AOdV under any position of nodes.

CONCLUSION
MANEt is infrastructure less network and operates in untrusted environment.In the proposed work tFAOdV, we have given solution of black hole attack in an untrusted environment of MANEt using tMPcF approach.Our work uses fuzzy based trust computation approach.Network performance of MANEt varies on conditions such as node mobility and positions of malicious node in the terrain.the proposed work considered different mobility conditions such as stationary network, low, moderate and high mobility nodes of the network.the position of malicious node considered are also different, central position and near source node in the terrain.the performance of MANEt under low mobility is better as compared to stationary nodes, as stationary nodes are more vulnerable.MANEt performance in case of malicious node central position in terrain is better as compared to near the source node, as malicious node near the source may directly attack and thus reduce the performance.Hence, our result shows that the proposed approach, tFAOdV have outperformed AOdV under all conditions in terms of network performance.
Nicolas Mayer, Sasa Radomirovic, "towards a decision model on trust and security risk management", in proceedings of 7 th Australasian Information security conference (AISc 2009), Wellington New Zealand 10. Asad

Fig 4 ,
Fig 4, fig 7 and fig 10 shows PdR, Etd and throughput under high mobility of 5m/s and 10 m/s of nodes.clearly tMPcF outperformed AOdV, but because of high mobility there may be some link breaks occur, so at mobility of 10 m/s performance of both protocols deteriorates.

Fig 5 ,
Fig 5, fig 8 and fig 11 shows PdR, Etd and throughput under different positions of malicious nodes, positions of normal nodes are random.the node's positions are taken as central to the complete terrain area and near the source

Table 2 : Input Parameters For black Hole Attack Under Aodv &Tfaodv
Amir Pirzada and chris Mcdonald, "Establishing trust in Pure Ad hoc Networks", 27 th Australiasian computer sc.conference, 2004