Minimal Adversarial Examples for Deep Learning on 3D Point Clouds

TL;DR

This paper investigates the vulnerability of 3D point cloud neural networks to adversarial attacks, proposing a unified method that achieves high success rates with minimal point manipulations, highlighting security concerns in 3D deep learning.

Contribution

It introduces a unified adversarial attack formulation for point cloud networks that balances attack success with minimal perceptible changes, advancing understanding of model vulnerabilities.

Findings

Achieves over 89% attack success rate on synthetic data.

Achieves over 90% attack success rate on real-world data.

Manipulates only about 4% of points to succeed.

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 safety-critical 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.