DRL-Based Secure Beamforming Design for RIS-Assisted ISAC System: A CRB Perspective

In this paper, we investigate the secure beamforming design for the reconfigurable intelligent surface (RIS) assisted integrated sensing and communications (ISAC) system. We consider a challenging scenario where the target act as a potential eavesdropper and the line-of-sight path between the target and the base station (BS) is blocked by the obstacles. Aiming at maximizing the secrecy rate of the system, we consider the joint design of transmit beamforming and RIS phase shift subject to the Cramér-Rao bound (CRB) constraint on the direction-of-arrival estimation of the eavesdropper. To solve this joint optimization problem, we propose an algorithm based on deep deterministic policy gradient (DDPG), by introducing a modified reward mechanism to satisfy the CRB constraint. Compared with the traditional algorithms, DDPG is simple and versatile to various environments. Numerical results demonstrate the effectiveness of the proposed algorithm and represent the trade-off relationship between communications and sensing.