Accuracy and Security-Guaranteed Participant Selection and Beamforming Design for RIS-Assisted Federated Learning

Abstract

Federated learning (FL) has emerged as an effective approach for training neural network models without requiring the sharing of participants' raw data, thereby addressing data privacy concerns. In this paper, we propose a reconfigurable intelligent surface (RIS)-assisted FL framework in the presence of eavesdropping, where partial edge devices are selected to participate in the FL training process. In contrast, the remaining devices serve as cooperative jammers by transmitting jamming signals to disrupt eavesdropping. We aim to minimize the training latency in each FL round by jointly optimizing participant selection, bandwidth allocation, and RIS beamforming design, subject to the convergence accuracy of FL and the secure uploading requirements. To solve the resulting mixed-integer nonlinear programming problem, we propose a twin delayed deep deterministic policy gradient (TD3) algorithm. Simulation results demonstrate that the proposed scheme reduces the FL training latency by approximately 27$\%$ compared to baselines.