Reinforcement Learning for Deceiving Reactive Jammers in Wireless Networks

Abstract

Conventional anti-jamming method mostly rely on frequency hopping to hide or\nescape from jammer. These approaches are not efficient in terms of bandwidth\nusage and can also result in a high probability of jamming. Different from\nexisting works, in this paper, a novel anti-jamming strategy is proposed based\non the idea of deceiving the jammer into attacking a victim channel while\nmaintaining the communications of legitimate users in safe channels. Since the\njammer's channel information is not known to the users, an optimal channel\nselection scheme and a sub optimal power allocation are proposed using\nreinforcement learning (RL). The performance of the proposed anti-jamming\ntechnique is evaluated by deriving the statistical lower bound of the total\nreceived power (TRP). Analytical results show that, for a given access point,\nover 50 % of the highest achievable TRP, i.e. in the absence of jammers, is\nachieved for the case of a single user and three frequency channels. Moreover,\nthis value increases with the number of users and available channels. The\nobtained results are compared with two existing RL based anti-jamming\ntechniques, and random channel allocation strategy without any jamming attacks.\nSimulation results show that the proposed anti-jamming method outperforms the\ncompared RL based anti-jamming methods and random search method, and yields\nnear optimal achievable TRP.\n