Cognitive Radio Jamming Mitigation using Markov Decision Process and Reinforcement Learning

The Cognitive radio technology is a promising solution to the imbalance between scarcity and under utilization of the spectrum. However, this technology is susceptible to both classical and advanced jamming attacks which can prevent it from the efficient exploitation of the free frequency bands. In this paper, we explain how a cognitive radio can exploit its ability of dynamic spectrum access and its learning capabilities to avoid jammed channels. We start by the definition of jamming attacks in cognitive radio networks and we give a review of its potential countermeasures. Then, we model the cognitive radio behavior in the suspicious environment as a markov decision process. To solve this optimization problem, we implement the Q-learning algorithm in order to learn the jammer strategy and to pro-actively avoid jammed channels. We present the limits of this algorithm in cognitive radio context and we propose a modified version to speed up learning a safe strategy. The effectiveness of this modified algorithm is evaluated by simulations and compared to the original Q-learning algorithm.