LiDAttack: Robust Black-box Attack on LiDAR-based Object Detection

TL;DR

LiDAttack is a novel black-box attack method on LiDAR-based object detection that uses genetic algorithms and simulated annealing to create stealthy, effective adversarial perturbations adaptable to real-world scenarios, achieving high success rates.

Contribution

This paper introduces LiDAttack, a robust black-box attack leveraging genetic algorithms and simulated annealing to improve stealth and effectiveness against LiDAR object detectors.

Findings

Achieves up to 90% attack success rate across multiple datasets and models.

Effectively limits perturbation points for stealthy attacks.

Adapts to real-world dynamic scanning variations.

Abstract

Since DNN is vulnerable to carefully crafted adversarial examples, adversarial attack on LiDAR sensors have been extensively studied. We introduce a robust black-box attack dubbed LiDAttack. It utilizes a genetic algorithm with a simulated annealing strategy to strictly limit the location and number of perturbation points, achieving a stealthy and effective attack. And it simulates scanning deviations, allowing it to adapt to dynamic changes in real world scenario variations. Extensive experiments are conducted on 3 datasets (i.e., KITTI, nuScenes, and self-constructed data) with 3 dominant object detection models (i.e., PointRCNN, PointPillar, and PV-RCNN++). The results reveal the efficiency of the LiDAttack when targeting a wide range of object detection models, with an attack success rate (ASR) up to 90%.