SymAttack: Symmetry-aware Imperceptible Adversarial Attacks on 3D Point Clouds

Keke Tang, Zhensu Wang, Weilong Peng, Lujie Huang, Le Wang, Peican Zhu, Wenping Wang, Zhihong Tian

Abstract: Adversarial attacks on point clouds are crucial for assessing and improving the adversarial robustness of 3D deep learning models. Despite leveraging various geometric constraints, current adversarial attack strategies often suffer from inadequate imperceptibility. Given that adversarial perturbations tend to disrupt the inherent symmetry in objects, we recognize this disruption as the primary cause of the lack of imperceptibility in these attacks. In this paper, we introduce a novel framework, symmetry-aware imperceptible adversarial attacks on 3D point clouds (SymAttack), to address this issue. Our approach starts by identifying part- and patch-level symmetry elements, and grouping points based on semantic and Euclidean distances, respectively. During the adversarial attack iterations, we intentionally adjust the perturbation vectors on symmetric points relative to their symmetry plane. By preserving symmetry within the attack process, SymAttack significantly enhances imperceptibility. Extensive experiments validate the effectiveness of SymAttack in generating imperceptible adversarial point clouds, demonstrating its superiority over the state-of-the-art methods. Codes will be made public upon paper acceptance.

Primary Subject Area: [Experience] Multimedia Applications

Secondary Subject Area: [Content] Media Interpretation

Relevance To Conference:

3D point clouds are frequently utilized alongside data from various modalities to improve a system's comprehension and decision-making capabilities. An illustrative case is in autonomous driving systems, where 3D point clouds (sourced from LiDAR) are merged with video imagery (captured by cameras) to enhance the accuracy of environmental awareness and object identification. Additionally, point clouds inherently encompass multiple modalities, including geometry and color, underscoring their significance in multimedia learning.

Our research investigates the adversarial robustness of deep learning models specifically designed for 3D point clouds. We believe that our work will offer a fresh perspective on robust multimodal deep learning.

Submission Number: 2919