Intelligent Resource Allocation for IRS-Assisted Sensing-Enhanced Secure Communication CRNs

Cognitive radio networks (CRNs) have demonstrated great benefits in improving spectrum efficiency. However, due to the exploitative character of wireless access, CRNs confront serious security issues. In this paper, intelligent reflection surface (IRS)-assisted secure communication cognitive radio networks (CRNs) are investigated, where IRS is exploited to enhance secure communication and improve spectrum efficiency. To maximize the secrecy rate of the secondary network, the subchannel assignment, reflection coefficients of the IRS, transmit beamforming of the secondary base station (SBS) and the sensing time are jointly optimized. An intelligent resource allocation algorithm using multiple deep Q networks (MDQN) algorithm and deep deterministic policy gradient (DDPG) algorithm is designed to address the tough non-convex problem with coupled variables. Simulation results show that our proposed intelligent approach achieves better secure performance than benchmark schemes, and our proposed intelligent resource allocation algorithm behaves with perfect convergence.