Intelligent Physical Layer Secure Relay Selection for Wireless Cooperative Networks with Multiple Eavesdroppers

The modern and emerging wireless communication technologies are benefiting from relaying techniques, which is already adopted in long term evolution (LTE) Release 10. The relay technology has great potential to improve the communication distance and system capabilities. However, due to the use of short-range frequencies for communication and increasing user density, relay technology requires the integration of machine learning and physical layer security (PLS) to attain its full potential. Therefore, this paper presents the machine learning (ML) algorithms for relay selection (RS) to improve the physical layer security of a dual-hop non-regenerative wireless cooperative network. The secure RS is converted into a prediction and decision-making problem. To the best of our knowledge, this is the first attempt to study the secured RS as a prediction and decision-making problem. We propose two reinforcement-learning techniques, namely the physical layer secured Q-learning algorithm for RS (PQL-RS) and the modified physical layer secured Q-learning algorithm for RS (MPQL-RS) for different scenarios. Markov decision process (MDP) is used with the introduced algorithms to select a single relay from a group of relays and multiple eavesdroppers. RS without requiring the prior data and network model knowledge are the unique advantages of the proposed algorithms. The introduced algorithms are compared to the different RS schemes and simulation results show that the MPQL-RS can deliver the best performance and save the convergence time. Moreover, the secrecy performance of different modulation techniques is also analyzed for the proposed algorithms. It is shown that the secrecy of the system improves by 7% on average for every double in modulation order.