In this thesis we consider the downlink of a wireless communication system. In particular, there is a base station transmitting information to multiple legitimate users in the presence of eavesdroppers which may compromise users’ privacy by capturing information sent from the base station. The goal of the thesis is to maximize the privacy degree of all legitimate users while ensuring that the eavesdroppers remain as oblivious as possible. To fulfill this, the base station leverages beamforming and reinforcement learning (RL). A specific objective of this thesis is to develop a practical RL algorithm with low latency and compare its performance against other approaches, e.g., based on convex optimization, which in general can be more time-consuming.

Required knowledge: reinforcement learning, wireless communications, signal processing (desirable)