Secure Multiuser Communications with Stacked Intelligent Metasurfaces using Quantum Reinforcement Learning

In this paper, we investigate a secure multiple-input single-output communication system empowered by stacked intelligent metasurfaces, comprising a base station, multiple users, and an eavesdropper. The base station leverages stacked intelligent metasurfaces to enhance physical layer security. Specifically, we consider a practical scenario where the BS has imperfect knowledge of the eavesdropper’s channel, introducing uncertainty into the system. Under this condition, the goal is to jointly optimize power allocation and stacked intelligent metasurfaces phase shifts to maximize the average secrecy rate across users. This optimization problem is highly non-convex and complicated further due to the dynamic wireless environment and channel state information uncertainty. To tackle these challenges, we propose a quantum reinforcement learning approach that learns optimal power and phase configurations. Simulation results demonstrate that the proposed QRL-based method significantly outperforms conventional baselines in terms of both adaptability and secrecy performance.