Latency Minimization for RIS Empowered Covert Federated Learning

We investigate in this letter the critical issue of latency minimization in the reconfigurable intelligent surface (RIS) empowered covert federated learning (FL) with the aggregation quality and privacy guarantee. Specifically, we propose to deploy a RIS to facilitate the two-way parameter transmissions of the FL system, and present a tradeoff between the training efficiency and covertness constraint by formulating a joint wireless resource optimization problem to optimize the uplink transmission design, the shared RIS phase configuration, and the downlink broadcast beamforming. To seek the solutions, we carefully invoke an alternating optimization strategy to convert the variable-coupled problem into several tractable subproblems through multiple equivalent transformations. Simulation results reveal that the proposed approach substantially surpasses baseline methods.