Intelligent Passive Eavesdropping in Massive MIMO-OFDM Systems via Reinforcement Learning

Massive multiple-input-multiple-output (MIMO) with narrow beam enhances the confidentiality of communication between base station and users, but also increases the difficulty for legal eavesdropping. In this letter, we study the passive eavesdropping scheme in the massive MIMO-OFDM systems by utilizing mobility of the monitor. Our objective is to maximize the average eavesdropping rate under the constraints of energy supply, moving direction and speed by jointly optimizing the receiving beamformers and moving trajectory. Due to the unknown environment knowledge and location of suspicious user, we propose the solution based on concatenated deep Q-network (DQN) to obtain the optimal solution. Simulation results verify the validity of the proposed method.

Automated summary

This letter investigates passive eavesdropping scheme in massive MIMO-OFDM systems by utilizing mobility of the monitor, aiming to maximize the eavesdropping rate by optimizing receiving beamformers and moving trajectory. The proposed solution based on concatenated deep Q-network (DQN) is validated to be effective through simulation results.