Machine Learning Techniques for Wormhole Attack Detection Techniques in Wireless Sensor Networks

The wormhole attack in Wireless sensor networks (WSNs) decreases the network performance by dropping the No. of Packets. Many techniques have been proposed to so far reduce the impact of the wormhole attack by detecting and preventing it. But, related work indicates that no technique is perfect for every kind of circumstances of WSNs. Among the existing techniques, Watchdog technique has better performance in preventing the wormhole attack. It utilizes the local knowledge of the next hop node and eavesdrops it. If it gets that spending time of the Packet is more than the given threshold, then it characterizes that node as wormhole attacker. However, this method has several shortcomings that it does not track the link transmission errors, which may be because of congestion in WSNs and also it not offers high mobility for maximum No. of nodes, which eventually decreases the WSNs performance. In order to handle this issue, a new multipoint relay based Watchdog monitoring and prevention technique is proposed in this paper. Proposed technique utilizes the dynamic threshold value to detect the wormhole attacker node, and then clustering and the Watchdog based optimistic path is selected for communicating the Packets. Thus, it reduces the overall Packet dropping, which improves the performance of the WSNs.