Collaborative Physical Layer Authentication in Internet of Things Based on Federated Learning

Abstract

With the booming growth of Internet of Things (IoT), ensuring the secure access of the IoT terminals is paramount importance. Machine learning (ML) based physical layer authentication (PLA) has been proposed as a promising solution for preventing unauthorized access of terminals due to its robust security. However, suffered from limited battery resources, IoT terminals are difficult to bear the ML tasks, which restricts the generality of ML-based PLA. In this paper, we propose a collaborative PLA scheme based on horizontal federated learning (HFL) to release computational pressure on resource-constrained IoT terminals. Particularly, we model the ML-based PLA as a training issue of the classifier, in which the training task is to obtain the weight parameter of the neural network. Then, we design a distributed authentication framework to assign the ML task of authentication to the trusted collaborators. Finally, all the local parameters trained by collaborators are aggregated at the center IoT terminal to obtain a global classifier used for PLA. Simulations show that both miss detection rate and false alarm rate are less than 1%, confirming the effectiveness of the proposed PLA scheme.