Covert ISAC With SIMs: A Deep Reinforcement Learning Approach Empowered by Generative Diffusion Model

This paper proposes a novel covert integrated sensing and communication (ISAC) system assisted by stacked intelligent metasurfaces (SIM) at both the transmitter and receiver. By deploying a SIM at the base station (BS) and a corresponding SIM at each user end, efficient beamforming in the electromagnetic wave domain is achieved. This architecture eliminates the reliance on traditional digital beamforming techniques and high-resolution digital-to-analog converters. Under the constraints of transmit power budget, radar sensing, and covert communication requirements, we maximize the sum rate by jointly optimizing the transmit power allocation and the SIM phase shifts. To address the non-convex and highly dynamic optimization problem, we introduce an innovative approach that integrates a generative diffusion model into the deep deterministic policy gradient (DDPG) algorithm, and further elaborate on the potential advantages of applying SIM technology in wireless networks. Simulation results demonstrate that the proposed method outperforms benchmarks in deep reinforcement learning (DRL) tests, validating the effectiveness of the SIM-enabled beamforming design scheme.