The article discusses AR-GAN, a Generative Adversarial Network-based defense method against adversarial attacks on the Traffic Sign Classification System of Autonomous Vehicles. The AR-GAN classification system includes a generator that denoises an image by reconstruction, and a classifier that classifies the reconstructed image. It was tested under various adversarial attacks and demonstrated high resilience, especially against white-box attacks, where it outperformed benchmark defense methods. The study emphasizes the importance of robust defense methods for AVs as compromised traffic sign information can be hazardous.
Publication date: 26 Jan 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2401.14232
