Research Progress of Physical Adversarial Examples for Anti-Intelligent Detection

This paper focuses on the research of physical domain adversarial examples in the field of machine learning security. Through in-depth analysis of the mainstream generation algorithms of digital domain adversarial examples, it provides a solid theoretical basis for the research and development of physical domain adversarial examples. In this paper, the research status of physical domain adversarial examples in visible light, infrared, laser and other detection fields is comprehensively reviewed, and the current novel physical adversarial methods are summarized, which provides a rich reference for researchers in this field. Looking forward to the future, the article points out that intelligent confrontation will tend to be diversified, and multi-mode confrontation will become the mainstream. At the same time, it emphasizes the importance of innovation in combining application scenarios and various physical confrontation technologies. In addition, the paper also proposes the necessity of improving the strong robustness and concealment of confrontation means in complex environments to ensure the effectiveness of physical domain confrontation. This review not only provides some guidance for the research of confrontation in the field of physics, but also lays a foundation for subsequent scholars to further explore this field.