Selfish misbehavior detection in 802.11 based wireless networks: An adaptive approach based on Markov decision process

The open and distributed nature of the IEEE 802.11 based wireless networks provides selfish users the opportunity to to gain an unfair share of the network throughput by manipulating the protocol parameters, say, using a smaller contention window. In this paper, we propose an adaptive approach for real-time detection of such selfish misbehavior. An adaptive detector is necessary in practice, as it needs to deal with different misbehaving scenarios where the number of selfish users and the contention windows exploited by each selfish user are different. In this paper, we first design a basic misbehavior detector based on the non-parametric cumulative sum (CUSUM) test. While the basic detector can be modeled with a Markov chain, we further resort to the Markov decision process (MDP) technique to enhance the basic detector to an adaptive design. In particular, we develop a novel reward function based on which the optimal policy of the MDP can be determined. The optimal policy indicates how the adaptive detector should operate at each state. Another important feature of our detector is that it enables an effective iterative method to detect multiple misbehaving nodes. We present thorough simulation results to confirm the accuracy of our analysis, and demonstrate the efficiency of the adaptive detector compared to a static solution.