Physical Layer Authentication Based on Channel Variance Normalization and Difference Exponential Amplification

The open nature of wireless media makes networks susceptible to spoofing attacks. In comparison to the conventional cryptographic-based authentication mechanism, Physical Layer Authentication (PLA) employs the intrinsic and distinctive internal security attributes of the physical layer to authenticate the sender, offering the benefits of robust security, minimal complexity, and broad compatibility. In this paper, we propose a hypothesis testing statistic based on channel variance normalization for identity authentication, utilizing two consecutive Channel State Information (CSI). To improve the authentication accuracy, we apply exponential amplification to the statistic mentioned above, aiming to maximize the difference in channel characteristics between legitimate and illegitimate links. Subsequently, we theoretically analyze the performance of the proposed scheme and derive closed-form expressions for four test probabilities: false alarm rate, authentication rate, detection rate, and miss detection rate. Additionally, we evaluate the performance of the scheme under the Neyman-Pearson (NP) criterion and the minimum error (ME) probability criterion, and give the threshold expressions under the two criteria respectively. Finally, we conducted simulation experiments on the proposed scheme in a typical Multiple-Input Multiple-Output (MIMO) scenario. The simulation results demonstrate the effectiveness of the proposed PLA scheme. After appropriate exponential amplification of the channel difference, the overall authentication performance is superior to that without amplification. The simulation results also prove that improving the signal-to-noise ratio, feature dimension, channel correlation coefficient, and reducing the power gain ratio of illegal and legitimate transmitters are helpful to improve the authentication reliability.