Physical layer security in MIMO hybrid FSO-mmWave systems: A learning-based link selection approach

Hybrid Free-Space Optical (FSO) and millimeter Wave (mmWave) systems have emerged as a promising candidate for backhaul networks of 5G and beyond radio access technologies due to the unique complementary properties against different channel and environment conditions. Therefore, in this study, we investigate the transmitter link selection problem for multiple-input multiple-output (MIMO) hybrid FSO-mmWave systems from a physical layer security point of view in the presence of different types of eavesdroppers. In particular, we propose convolutional neural network (CNN)-based link selection schemes to maximize the secrecy performance by activating the antennas and lasers at the transmitter side based on predefined configurations. The impact of fundamental physical layer parameters on the secrecy performance of a hybrid system is examined by taking the availability of channel state information (CSI), channel estimation errors, weather conditions, pointing error in FSO system, link distances, signal-to-noise ratios, path losses into account. In the light of the results, we show that the proposed CNN-based link selection scheme achieves the same performance as the conventional link selection mechanism.