Achieving Secure Federated Learning Assisted by Covert Communication

Federated learning (FL) is widely utilized in machine learning by exploiting its capabilities of protecting data privacy. However, due to the openness and broadcast characteristics of wireless channels, the parameters of FL is facing serious security risks during the upload process. Considering the protection of covert communication on communication behavior, a secure FL framework with the aid of covert communication is proposed, which realizes the covert transmission of parameters. Specifically, edge server (ES) sends artificial noise to confuse a malicious eavesdropper (Eve) when mobile devices (MDs) transmit information. Thus, we investigate the joint optimization problem of MDs power and jamming power to minimize the total latency in FL for a given covert constraint. Then, a particle swarm optimisation (PSO) algorithm is proposed in order to address the optimization problem. The results indicate that the proposed framework can successfully achieve covert transmission of model update parameters, thereby enhancing the overall security of the system.