Research map / Point clouds / Point cloud classification
Generating Transferable 3D Adversarial Point Cloud via Random Perturbation Factorization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add randomized perturbation factorization to the attack objective | PF-Attack · ModelNet40 | Transferability Score (%): 30.8 -> 40.2; 33.3 -> 46.8 | Helped |
| Add randomized perturbation factorization | PF-Attack, PointNet victim · PointNet Transformer transfer attack | Attack success rate (%): 21.6 -> 51.28; 23.6 -> 59.32 | Helped |
| Replace Chamfer loss with Hausdorff loss (ablation) | PF-Attack, PointNet victim · ModelNet40, PointNet++(MSG) | Attack success rate (%): 43.59 -> 36.17 | Hurt |
| Replace Chamfer loss with Hausdorff loss (ablation) | PF-Attack, PointNet victim · ModelNet40, PointNet++(SSG) | Attack success rate (%): 58.6 -> 52.8 | Hurt |
| Add the perturbation-factorization loss (ablation) | PF-Attack, PointNet victim · ModelNet40, PointNet++(MSG) | Attack success rate (%): 10.92 -> 43.59 | Helped |
| Add the perturbation-factorization loss (ablation) | PF-Attack, PointNet victim · ModelNet40, PointNet++(SSG) | Attack success rate (%): 12.6 -> 58.6 | Helped |
| Use 3 or 4 sub-perturbations instead of 2 (ablation) | PF-Attack, PointNet victim · ModelNet40, PointNet++(MSG), ϵ=0.45 | Attack success rate (%): 64.8 -> 63.4; 64.8 -> 60.9 | Hurt |
| Add perturbation factorization to 3D-Adv, KNN, or GeoA3 (ablation) | Existing point-cloud attacks · ModelNet40 transfer attacks | Attack success rate (%) | Helped |
| Set the random factorization probability to 0 or 1, eliminating effective factorization (ablation) | PF-Attack · ModelNet40 transfer attacks | Attack success rate (%) | Hurt |
| Vary the regularization coefficient β (ablation) | PF-Attack · ModelNet40 transfer attacks | Attack success rate (%) | No clear effect |
Closest papers by meaning
- Generating 3D Adversarial Point Clouds (CVPR 2019 · similarity 0.92)
- Shape-Invariant 3D Adversarial Point Clouds (CVPR 2022 · similarity 0.89)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.89)
- Towards Transferable Targeted 3D Adversarial Attack in the Physical World (CVPR 2024 · similarity 0.87)
- Towards Transferable Adversarial Attacks with Centralized Perturbation (AAAI 2024 · similarity 0.87)
- Boosting Adversarial Transferability through Augmentation in Hypothesis Space (CVPR 2025 · 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.