Enhanced Secure Transmission Against Intelligent Attacks

Abstract

In this paper, we proposed an enhanced secure scheme for the wireless communication system threatened by an intelligent attacker, which can work in eavesdropping, jamming, and spoofing modes. The conventional secure scheme is to apply Q-learning-based algorithm to reach a Nash equilibrium (NE) in the framework of a zero-sum game between the transmitter and attacker, which, however, requires the number of antennas at the transmitter to be much larger than that at the attacker. To overcome this limitation, we first consider the scenario where the attacker can flexibly increase the number of antennas in order to increase the attack rate. By adaptively setting the number of antennas at the transmitter and the legitimate receiver equal to that at the attacker, we then apply the beamforming at the transmitter to suppress the eavesdropping and use the filtering at the receiver to prevent the jamming and spoofing. By incorporating the beamforming and filtering, the benefits of the attacker in this game are efficiently restricted. Furthermore, the Q-learning-based power control strategy is used to reach a new NE. The simulation results have been demonstrated to show that the proposed scheme can suppress the intelligent attack efficiently, which outperforms the conventional scheme in the secrecy performance.

Keywords

• Reinforcement learning
• model-free
• beamforming
• zero-sum game