GAN Augmentation-Based Continuous Authentication for Vehicular Digital Twin

In this paper, to secure the communication between autonomous vehicle and its digital representative in the vehicular digital twin system, we propose a GAN augmentation-based continuous authentication scheme. Specifically, in the proposed scheme, we first introduce a data augmentation technique based Generative Adversarial Network (GAN) that provides the augmentation of raw data from vehicle sensors. We then present the efficient authentication: 1) We train a Convolutional Neural Network (CNN) using raw and augmented data; 2) Deep features are extracted through a combination of Principal Component Analysis (PCA) and CNN; 3) We train the OC-SVM classifier during the registration to ensure the legality of vehicle in the authentication phase. Performance evaluations via extensive simulations demonstrate the efficiency and effectiveness of the proposed scheme in terms of GAN loss and accuracy.