Deep learning-based secure beamforming for zeRIS-assisted UAV communications under partial blockage
Abstract
This paper investigates secure unmanned aerial vehicle (UAV) communications assisted by a zero-energy reconfigurable intelligent surface (zeRIS) under partial building blockage and imperfect channel state information (CSI). The zeRIS provides an alternative reflected path when the direct links are attenuated, subject to an energy activation constraint. A deep learning (DL) framework is developed to jointly design the information beamforming vector and artificial-noise matrix at the UAV, while the RIS phase shifts are determined using a low-complexity phase-alignment rule that enhances a selected cascaded channel component toward the legitimate receiver. Simulation results show that the proposed zeRIS-assisted DL schemes outperform the considered No- RIS and generalized eigenvalue decomposition (GEVD) baselines under the simulated blockage conditions. The results further indicate that enlarging the RIS surface relieves the secrecy-rate degradation induced by severe blockage of the direct link.