A Secure and Efficient Federated Learning Framework for Radio Frequency Fingerprint Recognition

With the rapid development of cognitive radio networks, the number of terminal devices has exponentially increased, leading to the generation of a vast amount of privacy-sensitive data, particularly WiFi signals. Radio Frequency (RF) fingerprinting holds a significant advantage in device authenti-cation within the Internet of Things (IoT) domain. To protect data privacy and enable RF fingerprint recognition without transmitting sensitive data, federated learning is necessary. In this paper, we apply hierarchical federated learning (HFL) for RF fingerprint recognition to reduce the communication overhead. Additionally, we conduct research on the vulnerability of federated learning to Byzantine attacks, and we propose a new defense method called Full-Krum to enhance the robustness of federated learning. Experimental results demonstrate the outstanding performance of our algorithm in safeguarding the security of federated learning.