Comprehensive Robustness Analysis of LiDAR-based 3D Object Detection in Autonomous Driving

2026-07-02

AI Summary

Existing LiDAR-only 3D object detection models lack systematic evaluation of their robustness under adversarial attacks, and current benchmarks predominantly rely on mAP, overlooking the multidimensional impact of structural and predictive factors. This work proposes the first multifactor adversarial robustness evaluation framework that integrates structural aspects—such as point cloud density and localization—with predictive dimensions including misclassification, localization error, and distance bias. Leveraging LiDAR-specific adversarial attack methods, we conduct comprehensive experiments on representative architectures, including voxel-based, pillar-based, and anchor-free models. Our findings reveal that high-capacity voxel-based models are more sensitive to coordinate perturbations, anchor-free models exhibit inferior robustness, and state-of-the-art detectors do not consistently outperform earlier approaches under adversarial conditions. These results underscore that accuracy alone is insufficient for safety-critical deployment, necessitating a redefined evaluation benchmark that explicitly incorporates robustness.

Abstract

Recent advancements in LiDAR-only 3D object detection have demonstrated improved detection accuracy over benchmark datasets. However, the adversarial robustness of these models remains untested. Very few adversarial robustness studies exist for LiDAR-only 3D object detection and unfortunately, even they are limited to legacy models. Moreover, there is a systemic gap in the existing evaluation frameworks that rely simply on mAP ignoring other structural and predictive factors. To fill this gap, we propose a holistic framework that evaluates adversarial robustness using two structural factors (point cloud density and point cloud localization) and three predictive factors (misclassification, localization error, distance from ego). Using this framework, we perform an empirical study and critical analysis on recent and legacy state-of-the-art models using adversarial attacks specifically designed for LiDAR-based models. Our key finding is that high-capacity, voxel-based detectors are more susceptible to structured coordinate perturbations than pillar-based detectors. Additionally, non-anchor-based detectors demonstrate poor adversarial robustness, which necessitates rethinking model training techniques. Overall, our results demonstrate that recent models are as vulnerable to adversarial attacks as their predecessors. Therefore, we argue that there is a need to improve the evaluation benchmarks for 3D object detection that not only reward architectural modifications for improving detection accuracy, but also evaluate whether the design choices improve adversarial robustness.

Problem

• adversarial robustness
• LiDAR-based 3D object detection
• evaluation framework
• autonomous driving
• point cloud

Innovation

• adversarial robustness
• LiDAR-based 3D object detection
• holistic evaluation framework
• voxel-based detectors
• non-anchor-based detectors