A Survey of Adversarial Attacks on 3D Point Cloud Object Recognition

Currently, artificial intelligence systems have achieved significant success in various domains, withdeep learning technology playing a pivotal role. However, although the deep neural network has stronginference recognition ability, it is still vulnerable to the attack of adversarial examples, showing itsvulnerability. Adversarial samples are specially crafted input data designed to attack and mislead the outputsof deep learning models. With the rapid development of 3D sensors such as LiDAR, the use of deep learningtechnology to address various intelligent tasks in the 3D domain is gaining increasing attention. Ensuring thesecurity and robustness of artificial intelligence systems that process 3D point cloud data, such as deeplearning-based autonomous 3D object detection and recognition for self-driving vehicles, is crucial. In order toanalyze the methods by which 3D adversarial samples attack deep neural networks, and reveal the interferencemechanisms of 3D adversarial samples on deep neural networks, this paper summarizes the research progress onadversarial attack methods for deep neural network models based on 3D point cloud data. The paper firstintroduces the fundamental principles and implementation methods of adversarial attacks, and then itsummarizes and analyzes digital domain adversarial attacks and physical domain adversarial attacks on 3Dpoint clouds. Finally, it discusses the challenges and future research directions in the realm of 3D point cloudadversarial attacks.