Optimal Relay Selection for Wireless Relay Channel with External Eavesdropper: a NN-based Approach

Abstract

In this paper, we exploit the potential benefits of machine learning in enhancing physical layer security in cooperative wireless networks. We focus on the case where multiple relays adopt amplify-and-forward (AF) relaying to forward information from the source to the destination. It is assumed that the global channel state information (CSI) of the legitimate links and wiretap links is available to the source. The optimal relay is selected to improve physical layer security against eavesdropping. By modeling the problem of the relay selection as a multi-class classification problem, a neural network (NN) based scheme is proposed to select the optimal relay which guarantees the perfect secrecy performance of the relay cooperative communication system. Compared with the conventional relay selection scheme, the simulation results show that our proposed scheme not only achieves almost the same secrecy performance, but also has advantage of relatively small feedback overhead. The work presented here provides insights into the new design of relay selection based on machine learning.