AR-GAN: Generative Adversarial Network-Based Defense Method Against Adversarial Attacks on the Traffic Sign Classification System of Autonomous Vehicles

The study focuses on the development of an attack-resilient Generative Adversarial Network (AR-GAN) for traffic sign classification in autonomous vehicles (AVs). AVs rely on Deep Neural Network (DNN)-based systems to recognize traffic signs, however these systems are vulnerable to adversarial attacks. The AR-GAN system includes a generator that denoises an image by reconstruction and a classifier that classifies the reconstructed image. The AR-GAN was tested under various adversarial attacks and outperformed other benchmark adversarial defense methods, especially in white-box attacks where the attackers possess full knowledge of the classifier.

Publication date: 26 Jan 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2401.14232