Publication

Abstract : This paper focuses on the security issues of the cognitive radio system. Since Cognitive Radio devices are more vulnerable to internal and external attackers, they are more likely to be engaged in malevolent activity. The majority of the Sniffer Channel Assignment (SCA) strategies for Cognitive Radio Networks use optimization-based techniques and rely on prior knowledge about Secondary User (SU) behaviors. Learning-based approaches have recently been developed to ease this restriction, but the theoretical foundation for SCA learning is currently weak. A non-stochastic bandit problem is suggested that tries to maximize the overall volume of the captured SU traffic. Furthermore, the proposed model takes into account the intrinsic flaw in wireless capturing, or incomplete monitoring. This work suggests two online learning algorithms, and the corresponding regret performances are consistently sub linear. The numerical evaluation demonstrates that the suggested algorithms significantly beat the current SCA techniques in the amount of effectively captured SU traffic, in addition to their robust regret capabilities. Without being aware of the SU actions beforehand, the suggested approach has a good theoretical performance guarantee in terms of the volume of traffic that is effectively caught.