Deep Reinforcement Learning-Based Secure Standalone Intelligent Reflecting Surface Operation

Abstract

In this paper, we investigate secure wireless commu-nication in an intelligent reflecting surface (IRS)-assisted system where the IRS is used to secure the communication of one legitimate receiver in presence of an eavesdropper. We assume that the IRS is standalone, i.e. the passive beamforming of the IRS is carried out completely on its own. Thus, we design an IRS with several passive elements and only two RF chains that can obtain a partial channel state information (CSI) among each node and the IRS. The partial CSI is then mapped into full CSI by using the correlation information between the channels of different IRS elements. We develop a deep reinforcement learning (DRL)-based framework using the deep deterministic policy gradient (DDPG) algorithm to obtain the IRS beamforming vector resulting in maximizing the secrecy rate. Numerical results demonstrate the ability of this technique to secure the wireless communication system.