Eidos: Efficient, Imperceptible Adversarial 3D Point Clouds

• DOI: 10.48550/arxiv.2405.14210
• Publication Date: 2024-05-23
• Authors: Hanwei Zhang; Cheng Luo; Qisong He; Wei Huang; Renjue Li; Ronan Sicre; Shengxi Huang; Holger Hermanns; Lijun Zhang
• Keywords: FOS: Computer and information sciences; Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV); Computer Science - Computer Vision and Pattern Recognition; FOS: Electrical engineering, electronic engineering, information engineering; Electrical Engineering and Systems Science - Image and Video Processing

Classification of 3D point clouds is a challenging machine learning (ML) task with important real-world applications in spectrum from autonomous driving and robot-assisted surgery to earth observation low orbit. As other ML tasks, classification models are notoriously brittle the presence adversarial attacks. These rooted imperceptible changes inputs effect that seemingly well-trained model ends up misclassifying input. This paper adds understanding attacks by presenting Eidos, framework providing Efficient Imperceptible aDversarial on pOint cloudS. Eidos supports diverse set imperceptibility metrics. It employs an iterative, two-step procedure identify optimal examples, thereby enabling runtime-imperceptibility trade-off. We provide empirical evidence relative several popular cloud established attack methods, showing Eidos' superiority respect efficiency as well imperceptibility.