Minimal Adversarial Examples for Deep Learning on 3D Point Clouds

Jaeyeon Kim1          Binh-Son Hua3          Duc Thanh Nguyen4          Sai-Kit Yeung1

1Hong Kong University of Science and Technology 2VinAI Research
3VinUniversity 4Deakin University

International Conference on Computer Vision (ICCV), 2021


Modelnet40


ScanObjectNN


Abstract

With recent developments of convolutional neural networks, deep learning for 3D point clouds has shown significant progress in various 3D scene understanding tasks, e.g., object recognition, semantic segmentation. In a safetycritical environment, it is however not well understood how such deep learning models are vulnerable to adversarial examples. In this work, we explore adversarial attacks for point cloud-based neural networks. We propose a unified formulation for adversarial point cloud generation that can generalise two different attack strategies. Our method generates adversarial examples by attacking the classification ability of point cloud-based networks while considering the perceptibility of the examples and ensuring the minimal level of point manipulations. Experimental results show that our method achieves the state-of-the-art performance with higher than 89% and 90% of attack success rate on synthetic and real-world data respectively, while manipulating only about 4% of the total points.

Materials

We will open the code soon.

Citation
@article{jy-minimumpointattack-iccv21,
      title = {Minimal Adversarial Examples for Deep Learning on 3D Point Clouds},
      author = {Jaeyeon Kim and Binh-Son Hua and Duc Thanh Nguyen and Sai-Kit Yeung},
      booktitle = {International Conference on Computer Vision (ICCV)},
      year = {2021}
  }

Acknowledgements

This research project is partially supported by an internal grant from HKUST (R9429).