Physical Attack

Physical attacks on computer vision systems, particularly those employing deep neural networks (DNNs), are a growing area of research focusing on how to compromise the accuracy and reliability of these systems through real-world manipulations. Current work investigates various attack methods, including adversarial patches, color projections, and backdoor poisoning using physical triggers, often leveraging optimization algorithms and generative models to create stealthy and effective attacks. This research is crucial for understanding and mitigating vulnerabilities in safety-critical applications like autonomous vehicles and facial recognition, where the consequences of system failure can be severe.