Enhanced Secure Beamforming for IRS-Assisted IoT Communication Using a Generative-Diffusion-Model-Enabled Optimization Approach

Abstract

The spatial correlation between the main and eavesdropping channels impacts the effectiveness of beamforming (BF) technology in securing the Internet of Things (IoT) communication system. This study proposes an IRS-assisted secure BF scheme (IRS-SBS) for a multiuser IoT system under imperfect channel state information (CSI). The security issue, constrained by channel spatial correlation, is formulated as a nonconvex optimization problem with probabilistic constraints. To address this, we introduce a novel actor-critic algorithm combined with a generative diffusion model (AC-GDM). This approach jointly optimizes the base station (BS) precoding matrix and the intelligent reflecting surface (IRS) phase shift matrix, subject to constraints on transmit power and phase shifts. The AC-GDM algorithm utilizes a denoising process to recover the optimal BF solution from the Gaussian noise inherent in the multiuser wireless channel environment. Simulation results demonstrate that the minimum achievable secrecy rate of the IRS-SBS outperforms the artificial noise (AN) scheme and the BF scheme by approximately 1.9576 and 2.3596 bps/Hz, respectively. These results validate the effectiveness of IRS-SBS in significantly mitigating the impact of channel spatial correlation on the security of IoT communication systems.