Research map / Point clouds / Point cloud classification
Fooling LiDAR Perception via Adversarial Trajectory Perturbation
Yiming Li, Congcong Wen, Felix Juefei-Xu, Chen Feng
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| attack classification of stage-2 instead of random translation noise | FLAT (Translation) · nuScenes car Easy | AP: additional drops of 5.28, 2.71, and 2.84 in AP versus random perturbation (easy, moderate, hard) | Improved |
| perturb rotation instead of translation | FLAT · nuScenes car Easy | AP: 11.72 -> 2.35 | Improved |
| perturb full trajectory instead of rotation only | FLAT · nuScenes car Easy | AP: 2.35 -> 0.19 | Improved |
| attack stage-2 classification versus stage-2 regression (translation) | FLAT (Translation) · nuScenes car Easy | AP: 26.09 -> 11.72 | Improved |
| attack with polynomial trajectory perturbation instead of per-point translation | FLAT · nuScenes car Easy | AP (stage-2 cls): 11.72 -> 12.51 | No clear effect |
| attack with coordinate-level perturbation instead of trajectory translation | PointRCNN attack · nuScenes car Easy | AP: 11.72 -> 16.42 | Hurt |
| add trajectory smoothness regularization (lambda 0.01) (ablation) | FLAT · nuScenes car Easy | AP: 0.19 -> 0.28 | No clear effect |
| add point cloud l_p regularization lambda=0.01 (ablation) | FLAT · nuScenes car Easy | AP: 0.19 -> 5.66 | Mixed |
| increase point cloud regularization strength from 0.01 to 1 (ablation) | FLAT · nuScenes car Easy | AP: 5.66 -> 23.65 | Hurt |
| transfer attack to black-box PointPillar++ with full trajectory perturbation | FLAT · nuScenes val | mAP: 35.4 -> 14.8 | Improved |
| attack rotation instead of translation in black-box transfer | FLAT · nuScenes val | mAP: 20.6 -> 16.4 | Improved |
Closest papers by meaning
- Physically Realizable Adversarial Examples for LiDAR Object Detection (CVPR 2020 · similarity 0.92)
- Enduring, Efficient and Robust Trajectory Prediction Attack in Autonomous Driving via Optimization-Driven Multi-Frame Perturbation Framework (CVPR 2025 · similarity 0.90)
- A New Adversarial Perspective for LiDAR-based 3D Object Detection (AAAI 2025 · similarity 0.89)
- SlowLiDAR: Increasing the Latency of LiDAR-Based Detection Using Adversarial Examples (CVPR 2023 · similarity 0.89)
- Neural Reconstruction of LiDAR Point Clouds under Jamming Attacks via Full-Waveform Representation and Simultaneous Laser Sensing (NeurIPS 2026 · similarity 0.88)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
