Implementation of a Trust Model over OLSR in a Wireless ad hoc Testbed Chandrakant Gaurav #1, Dhivya Chandramouleeswaran *2, Rashda Khanam *3, Revathi Venkataraman $4, M.Pushpalatha $5, Dr. T. Rama Rao &6 # Assistant Systems Engineer, TCS Bangalore, India 1 chandrakant51289@gmail.com * Consultant, SAP Labs India Pvt. Ltd., India 2 gabriella.dhivya@gmail.com 3 rashukhanam@gmail.com $ Assistant Professor (SG), SRM University, India 4 revathivenkat@yahoo.com 5 lathamarudappa@yahoo.co.in & Head, Telecommunication Engineering, SRM University, India 6 ramarao@tce.srmuniv.ac.in Abstract The concept of a customized Optimized Link State Routing (OLSR) protocol with focus on augmented security protecting a node against blackhole and flooding attacks launched by its neighbor nodes in the network by implementing a trust model is described in this paper. This trust model is a computational variable that dictates the behavior of the node to protect itself against malicious attack in a robust and intelligent manner. We have also developed a testbed that evaluates the resultant deportment of the node in real time and generates user friendly analysis reports. Through real time test runs, we have concluded results favoring our proposal in the following network measures of throughput, end to end delay and overhead. The throughput was found to be 70% better in real time ad hoc testbed which has been designed to have 25% blackhole and flooding nodes respectively. The overhead encountered was minimal despite the end to end delay of packets being higher than the traditional OLSR by 30 msec. Our work is also the first ever implementation of a trust model over an ad hoc routing protocol in a real ad hoc testbed. Keywords- OLSR, Trust Model, Blackhole attacks, Flooding attacks. Ι. INTRODUCTION Optimized link state routing protocol is a proactive protocol fetching routing information at the inception of network activity. This leaves the engaging network to be vulnerable to malicious attacks from the neighboring nodes as well as the nodes in range for the duration when no changes in the network topology occur. Blackhole and Flooding are identified to be causing the maximum irregularities in the subnet topology. The periled nodes make use of the unreliable links and dynamic topology of ad hoc networks and introduce inconsistencies in the routing table information exchange. The existing OLSR code has no mechanisms to ward itself against aforementioned attacks and no security modules which realize the proposed solution. To bolster the network against these two attacks, we have fortified the protocol making it anticipate the attacks and help a node take intelligent decisions safeguarding it. This parameter enabling intelligence is named Trust. We evaluate trust as a numerical variable obtained as a mathematical equation known as Normalized Trust Value Evaluation. A node enabling trust incorporation will have the ability to reason about and make security related decisions autonomously through a supplement local gateway. The autonomous decisions are taken based upon the history of trust applied intravenously in the code and externally through the module. Trust information or values are stored in structures of source nodes, to represent historical information on the behavioral patterns of neighboring entities through packet drops and forwards. A value is fixed as a threshold after a series of test runs of the network under conditions explained afterwards in the paper. Any expected deviation to this threshold trust will affect the awareness of the network and lead to predicting dishonest behavior, based on conformational patterns in the available evidence of trust values. This dynamic real time view of trust will result in a more flexible model able to defend the network against harmful consequences leading to better network statistics. The main contribution of this work is Development of robust trust model over OLSR Implementation of robust trust in real time wireless ad-hoc testbed. ΙΙ. PROPOSED TRUST MODEL A. Routing in OLSR OLSR is a pro-active protocol fetching routing information only on demand [6]. The process starts with broadcasting HELLO PACKETS in the network. These messages are used for neighbor sensing and MPR (Multi- Point Relay) calculation. The source records into the routing table all the neighbors based on sequential lowest sequence numbers, taking them as a measure of node distance. ROUTE REQUEST is sent to the next hop neighbors, calculated from the neighbor list. This is a recursive procedure followed by every intermediate in the network. When such an intermediate node receives the request, it forwards the control packet by applying the recursive procedure till it reaches the destination. The final route for transmitting the data packets is committed when a ROUTE REPLY from destination reaches the source. 978-81-920249-7-4/13/$31.00 c 2013 IEEE 46
A node enabling trust incorporation will have the ability to reason about and make security related decisions autonomously through a supplement local gateway. The autonomous decisions are taken based upon the history of trust applied intravenously in the code and externally through the module. Trust information or values are stored in structures of source, to represent historical information on the behavioral patterns of neighboring entities through packet drops and forwards. Any expected deviation to threshold trust will affect the awareness and lead to predicting dishonest behavior, based on patterns in the available evidence of trust values. This dynamic real time view of trust will result in a more flexible model able to represent trust in a manner that captures human intuitions, such that positive outcomes of interactions will preserve trust, while trust erodes with runtime degrading packet statistics in a promiscuous network. The main contribution of this work is Development of robust trust model over OLSR Implementation of robust trust in real time wireless adhoc testbed. ΙΙ. PROPOSED TRUST MODEL A. Routing in OLSR OLSR is a pro-active protocol fetching routing information only on demand [6]. The process starts with broadcasting HELLO PACKETS in the network. These messages are used for neighbor sensing and MPR (Multi-Point Relay) calculation. The source records into the routing table all the neighbors based on sequential lowest sequence numbers, taking them as a measure of node distance. ROUTE REQUEST is sent to the next hop neighbors, calculated from the neighbor list. This is a recursive procedure followed by every intermediate in the network. When such an intermediate node receives the request, it forwards the control packet by applying the recursive procedure till it reaches the destination. The final route for transmitting the data packets is committed when a ROUTE REPLY from destination reaches the source. B. Ad hoc Testbed: Hardware and Software The Operating system used for the real time ad hoc scenario setup is Linux Mint version 10, Kernel 2.6.35.10. The OLSR version being used OLSR-uu-0.9.6 from Uppasala University [7]. The testbed for GUI based monitoring is coded using Python wxgtk-2.8. This testbed consist of two sections Test and Analysis. Test This front end interface is used to give instructions to the volunteers performing the test- runs in the testbed. The node configuration parameters, time synchronization before testrun and the protocol specific parameters are set using this front end interface. At the end of the experiment, the logs from individual machines are uploaded to a central system for further analysis Analysis The logs collected from individual ad hoc nodes are consolidated and aggregation charts are prepared for easy analysis of the following performance metrics like throughput, delay, packet loss, etc. C. Trust Model Evaluation Trust is a float value with its range limited by (-1) and (1). We have implemented the trust values by deriving the calculation parameters through packet forwarding and packets received, since the neighbor node behavior is monitored through these variables only. The trust is calculated for each node as shown in Table 1. T1 T2 TABLE I. TRUST METRICS IN USED IN THE TRUST MODEL Trust metrics Number of data packets successfully forwarded from source to destination Number of data packets received from the neighbor. T1 is estimated by transferring raw data packets using hping3. T2 is estimated by analyzing the tshark log files. The source node prioritizes its neighbors based on computed trust. The aggregate trust of a neighboring node is computed by the Ordered Weighted Averaging (OWA) operator as shown in Equation (1) OWA (T1, T2) = wjtj (1) Where wj are weights associated with each trust metric With wj > 0. The reason behind assigning different weighted trust values is to evaluate different conditions of deduction of malicious nodes. The normalized trust value is computed from equation (2). Normalized trust T = (Ac Amin )(Lmax Lmin ) + Lmin Amax Amin Where- Ac - current aggregated trust Amax - maximum possible value of the aggregated trust Amin - minimum possible value of the aggregated trust Lmax - +1 for the node with highest level of trust Lmin - -1 for the node with lowest level of trust The computed trust are categorized as follows. TRUSTED - T= 1 MALICIOUS - T= -1 UNCALCULATED - T= 0 These observations are used in choosing the right path for data transmission, rerouting whenever a malicious path is detected through trust. 2013 2nd International Conference on Information Management in the Knowledge Economy 47
ΙΙΙ. EXPERIMENTAL SETUP AND PERFORMANCE ANALYSIS The real time adhoc network was established in a indoor environment with four testing nodes, each participating in a data transfer session through OLSR running in the real time testbed. The data rate fixed at 2Mbps. The ad hoc nodes are in a diamond topology as shown in Fig.1 and the approximate distance between communicating nodes is 70 meters (indoor environment). The duration of experimentation was an average of 30 minutes. Any node could act as a malicious node through adaptable customized design in real time testbed front end nodes. The tests were run first in normal scenario and then in presence of a malicious node. The results are reflected in the graph produced in the testbed. A Source B Intermediate C Malicious D Destination Fig 2. Ad-hoc Testbed used at node A (source node) Fig 1. Adhoc Nodes in Diamond Topology The experiment was carried out using a graphical user interface in the form of a testbed developed on wxpython. It is predesigned with all the set of characteristics a node could take upon and which can selected by the user through drop down menus. These include data rates, malicious behavior, number of data packets to be sent for computation of trust values and optional file transfer. The results area on the left of the GUI shows real time statistics of the network like neighbor detection, relative computation variable for trust from neighboring nodes which are tabulated as Trust metrics in Table I. A popup menu informs the user accessing the testbed of that node which of its neighboring nodes have been dropped following its trust value prediction. After running test modules having different parametric combinations, the logs are analyzed in the testbed. Normal OLSR refers to the default OLSR. Blackhole OLSR refers to a blackhole node in the network without trust component. Trusted OLSR is the scenario where the nodes run the trust algorithm in their machines and the network has malicious While normal OLSR has almost 100% throughput as shown in Fig. 3, it is seen that trusted OLSR has 70% aggregate throughput which is better than the 20% of OLSR run in the presence of blackhole without the trusted part. The loss of 30% in trusted OLSR is due to the time taken by the neighboring nodes to judge the malicious behavior of its neighbor. TrustedOLSR has an end-to-end delay more than normalolsr by 30 msec as shown in Fig. 4. This is the overhead associated with the trust computation algorithm. As expected, blackholeolsr has 70% packet loss while normalolsr has around 25% as shown in Fig. 5. TrustedOLSR manages a modest 50% which is better than blackholeolsr. The mechanism of trust includes user input raw packets which deals with packet signatures of each neighbor in the network. The graph in Fig. 6 is derived by deploying different values of intervals, in terms of packets used for testing. As predicted, it was found that lesser packets used yielded in lesser time i.e. 1 packet taking 10 milliseconds, for detected malicious nodes while opposite was observed for larger volumes, 10 packets taking 100 milliseconds. 48 2013 2nd International Conference on Information Management in the Knowledge Economy
Fig 3. Comparison of the throughput under normal, black hole and trusted OLSR. Fig 5. Comparison of packet loss percentage in all three scenarios. Fig 4. Comparison of end-to-end packet delay of the normal and trusted OLSR Fig 6. Trust computation overhead varying the frequency of trust computation 2013 2nd International Conference on Information Management in the Knowledge Economy 49
IV.CONCLUSION OLSR employed MANET is prone to malevolent attacks and our implementation of trust in real time has successfully illuminated the advantageous effects of eliminating blackhole and flooding attacks as proposed and tested. The performance analysis has yielded the result that this implementation has an unorthodox behavior compared to normal OLSR in secured networks. However, in the real time custom designed attack prone environment, the output is significantly better aiding detection, removal of deleterious nodes and subsequent rerouting as required. This encourages us to peruse a generalized approach for securing MANET in other protocols like OLSR in the future. ACKNOWLEDGMENT This work is funded by Defense Research and Development Organization of India (Grant No.:IP/ER/0803748/M/01/1189) REFERENCES [1]. C.Sivaram Murthy and B.S Manoj Ad Hoc Wireless Networks, Pearson Education, Second Edition India, 2001. [2]. Levente Buttyan and Jean-Pierre Hubaux Security and Co-operation in wireless networks, Cambridge University Press. Febraury 2007. [3]. Sanjay Ramaswamy, Huirong Fu, Manohar Sreekantaradhya, John, Dixon and Kendall Nygard, Prevention of Cooperative Black Hole Attack in Wireless Ad Hoc Networks, Department of Computer Science, IACC 258, North Dakota State University, Fargo, ND 58105 [4]. Theodorakopoulos G.and J.S.Baras : On trust models and trust evaluation metrices for ad hoc networks. IEEE Journal on Selected Areas in Communications, vol.24, Issue.2, pp.318-326, 2006 [5]. OLSR Internet draft http://tools.ietf.org/html/draft-jacquet-olsr-molsr-00 50 2013 2nd International Conference on Information Management in the Knowledge Economy