Reinforcement Learning based Security Policy to Mitigate Wormhole, Blackhole and Grayhole Attacks in MANET

This research study applies reinforcement learning to the problem of preventing wormhole, blackhole and grayhole attacks on mobile ad hoc networks (MANETs). The flexible actor-critic architecture is utilized here to implement reinforcement learning. Utilization of wireless sensor networks in the development of a variety of communication infrastructures is increasing. Multiple types of intrusions can be launched against wireless network sensors. Due to these sensor devices, a wireless network is susceptible to a denial of service attack. One is a wormhole, blackhole and grayhole attack, in which two malicious sensor nodes are connected via a low-latency link to disrupt the normal routing of the network. Researchers are increasingly attracted to machine learning techniques due to their ability to uncover previously unknown hazards. Our primary objective is to identify these intricate patterns and construct a secure mobile ad hoc network by prioritizing security measures such as detecting and blocking malicious nodes. Simulation results demonstrate the efficient mitigation of wormhole, blackhole and grayhole attack.