Research map / Point clouds / Point cloud classification
Good Can Sometimes be Bad: A Unified Attack against 3D Point Cloud Classifier by a Flexible Isotropic Resampling
Linkun Fan, Jiahao Zhang, Juntao Zhang, Lei Zhang, Fazhi He, Daojun Han
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 |
|---|---|---|---|
| use flexible isotropic resampling to hide the attack instead of limiting perturbation magnitude | UAtt3D backdoor · ModelNet40 (PointNet++) | CUD: 103.49 (benign) -> 62.27 | Improved |
| UAtt3D versus PointPBA-I on stealthiness | UAtt3D backdoor · ModelNet40 (PointNet++) | KUV: 4.34 -> 0.33 | Improved |
| UAtt3D versus GeoA3 for adversarial attack | UAtt3D adversarial · ShapeNet16 (PointNet++) | ASR %: 83.86 -> 100 | Improved |
| UAtt3D on DGCNN adversarial attack | UAtt3D adversarial · ModelNet40 (DGCNN) | ASR %: 95.78 vs 99.96 (Eidos) | Hurt |
| apply UAtt3D to defective point clouds (outliers, holes, sparsity) | UAtt3D · Defective ModelNet40 (PointNet++) | ASR %: 96.57 backdoor ASR, 100.00 adversarial ASR with all defects | No clear effect |
| saliency-based defense (dropping 60 highest-saliency points) | PointBA-I vs UAtt3D · ModelNet40 | ASR %: PointBA-I ASR decreased to 15.13%; UAtt3D resists | Hurt |
| PointCVaR defense against UAtt3D | UAtt3D · ModelNet40 | ASR %: backdoor 98.77 -> 98.29; adversarial 95.78 -> 78.42 | Mixed |
| frequency filter defense | UAtt3D backdoor · ModelNet40 | ASR %: 98.77 -> 98.29 (IBAPC to 25.31, PointBA-I to 4.33) | No clear effect |
| replace traditional resampling with the designed FIR (ablation) | UAtt3D · ModelNet40 | ASR / geometry preservation | Improved |
| increase poison rate alpha (ablation) | UAtt3D backdoor · ModelNet10 / ModelNet40 | ASR | Improved |
| vary target label y_t (ablation) | UAtt3D · ModelNet40 | ASR / BAc | No clear effect |
| attacked cloud via UAtt3D used for mesh reconstruction | UAtt3D · ball pivoting meshes | face normal consistency (lower better): 23.15 -> 16.48 | Improved |
Closest papers by meaning
- PointBA: Towards Backdoor Attacks in 3D Point Cloud (ICCV 2021 · similarity 0.90)
- Improving Adversarial Robustness of 3D Point Cloud Classification Models (ECCV 2022 · similarity 0.89)
- Invisible Backdoor Attack against 3D Point Cloud Classifier in Graph Spectral Domain (AAAI 2024 · similarity 0.88)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.87)
- Toward Availability Attacks in 3D Point Clouds (ICML 2024 · similarity 0.86)
- Explicitly Perceiving and Preserving the Local Geometric Structures for 3D Point Cloud Attack (AAAI 2024 · similarity 0.86)
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.
