English

IF-Defense: 3D Adversarial Point Cloud Defense via Implicit Function based Restoration

Computer Vision and Pattern Recognition 2021-03-19 v3

Abstract

Point cloud is an important 3D data representation widely used in many essential applications. Leveraging deep neural networks, recent works have shown great success in processing 3D point clouds. However, those deep neural networks are vulnerable to various 3D adversarial attacks, which can be summarized as two primary types: point perturbation that affects local point distribution, and surface distortion that causes dramatic changes in geometry. In this paper, we simultaneously address both the aforementioned attacks by learning to restore the clean point clouds from the attacked ones. More specifically, we propose an IF-Defense framework to directly optimize the coordinates of input points with geometry-aware and distribution-aware constraints. The former aims to recover the surface of point cloud through implicit function, while the latter encourages evenly-distributed points. Our experimental results show that IF-Defense achieves the state-of-the-art defense performance against existing 3D adversarial attacks on PointNet, PointNet++, DGCNN, PointConv and RS-CNN. For example, compared with previous methods, IF-Defense presents 20.02% improvement in classification accuracy against salient point dropping attack and 16.29% against LG-GAN attack on PointNet. Our code is available at https://github.com/Wuziyi616/IF-Defense.

Keywords

Cite

@article{arxiv.2010.05272,
  title  = {IF-Defense: 3D Adversarial Point Cloud Defense via Implicit Function based Restoration},
  author = {Ziyi Wu and Yueqi Duan and He Wang and Qingnan Fan and Leonidas J. Guibas},
  journal= {arXiv preprint arXiv:2010.05272},
  year   = {2021}
}

Comments

17 pages, 8 figures. Update several experimental results compared with v2, e.g. cross dataset evaluation and baseline results of adversarial training

R2 v1 2026-06-23T19:15:12.928Z