ViT-PLA: A Vision Transformer-Based Physical Layer Authentication Method for Industrial Wireless Networks

In this letter, we propose a Vision Transformer-based Physical Layer Authentication (ViT-PLA) method for industrial wireless networks. To this end, Channel Frequency Response (CFR) samples are organized in dual-channel CFR images, which together with request positions encompass necessary information on the spatial-temporal correlation between CFR samples. Further, we design a novel Deep Neural Network (DNN) model consisting of a ViT and two feedforward neural networks to learn from the well-designed training samples. The implementation of the trained DNN model for online authentication is also discussed. Finally, the effectiveness and the generalizability of ViT-PLA are demonstrated on real industrial datasets.