Physical Layer Authentication Based on CNN-GAN

The rapid growth of the Internet of Things (loT) has created significant opportunities for the future of telecommunications. Research on physical layer authentication with channel features holds promise to improve wireless network security for loT devices. In this paper, we exploit Channel State Information (CSI) phase extracted from the receiver with multiple antennas to enhance the channel spatial diversity. Then Convolutional Neural Network and Generative Adversarial Network (CNN-GAN) model is designed for physical layer authentication. Outlier score from model is introduced to authenticate device identity. Finally, we carry out extensive experimental evaluations about the authentication performance and robustness of the proposed scheme by data collected from Wi-Fi signal based on IEEE 802.11n protocol. The results have successfully demonstrated that the proposed scheme exhibits outstanding performance.