The Vulnerability of Autonomous Driving to Typographic Attacks: Transferability and Realizability

This article reviews and compares two major types of adversarial attacks against neural networks: gradient-based methods (like PGD) and typographic attacks.