A robust MDP approach to secure power control in cognitive radio networks

Power control plays a key role in realizing reliable and spectrum-efficient communications in a cognitive radio network. In this paper we study secure power control schemes for cognitive radios via a robust Markov decision process (MDP) approach. We first use the discrete time Markov chain (DTMC) model to characterize the primary user´s (PU) activities as well as the dynamics of the fading channel. The resulting power control problem can be optimally solved by a dynamic programming (DP) approach. The presence of malicious users, however, necessitates a cooperative spectrum sensing approach that requires extra signal processing and information exchange efforts to identify false spectrum sensing reports. Such a cooperative approach incurs considerable delay in spectrum sensing that may significantly deteriorate the performance of the DP strategy. Furthermore, the false sensing data generated by malicious users may give rise to erroneous estimation of the transition probabilities. Consequently, the solution obtained based on the estimated transition matrices may exhibit poor performance. To cope with these challenges, we propose a framework of power control schemes based on the robust MDP approach that is capable of achieving reasonably good performance when both spectrum sensing delay and estimation errors are present. The tradeoffs between the robustness, the achievable throughput, and the sensing delay are also discussed. Extensive simulation results are presented to demonstrate the performance of the proposed power control strategies.