Research map / Point clouds / Point cloud classification
Imperceptible Transfer Attack and Defense on 3D Point Cloud Classification
Daizong Liu, Wei Hu
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 |
|---|---|---|---|
| add adversarial transformation model (ablation) | ITA attack, PointNet to PointNet++ · ModelNet40 transfer | transfer success rate (%): 14.60 -> 54.62 | Improved |
| add adversarial transformation model (ablation) | ITA attack, DGCNN to PointNet++ · ModelNet40 transfer | transfer success rate (%): 29.89 -> 76.74 | Improved |
| replace learning-free cloud-wise transform with learned transformation (ablation) | ITA attack, PointNet to PointNet++ · ModelNet40 transfer | transfer success rate (%): 31.28 -> 54.62 | Improved |
| use point-wise instead of cloud-wise learning-free transformation (ablation) | ITA attack, PointNet to PointNet++ · ModelNet40 transfer | transfer success rate (%): 31.28 -> 38.41 | Improved |
| replace learning-free point-wise transform with learned transformation (ablation) | ITA attack, PointNet to PointNet++ · ModelNet40 transfer | transfer success rate (%): 38.41 -> 54.62 | Improved |
| perturb along normal direction with bounded width vs GeoA xyz perturbation | ITA vs GeoA, PointNet · ModelNet40, budget B=0.01 | attack success rate (%): 79% -> 93% | Improved |
| perturb along normal direction with bounded width vs GeoA xyz perturbation | ITA vs GeoA, PointNet · ModelNet40 perturbation size | geometric distortion D_m: 0.1243 -> 0.0927 | Improved |
| tighten bound width to B=0.01 (ablation) | ITA, PointNet · ModelNet40 | attack success rate (%): 93% at B=0.01, 100% at B=0.02 | Mixed |
| ITA instead of GeoA against IF-Defense | PointNet · ModelNet40 | attack success rate (%): 38.72 -> 69.32 | Improved |
| add normal-direction Gaussian noise 2% as defense (ablation) | ITA, PointNet · ModelNet40 | attack success rate (%): 100 -> 95.72 | No clear effect |
| add latent intra- and inter-class constraint to adversarial training (ablation) | Defense, ITA transfer to PointNet · ModelNet40 | attack success rate (%): 49.98 -> 18.13 | Improved |
| add latent constraint versus DeepSym adversarial training | Defense, ITA transfer to PointNet · ModelNet40 | attack success rate (%): 37.38 -> 18.13 | Improved |
Closest papers by meaning
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.92)
- Curvature-Invariant Adversarial Attacks for 3D Point Clouds (AAAI 2024 · similarity 0.90)
- Manifold Constraints for Imperceptible Adversarial Attacks on Point Clouds (AAAI 2024 · similarity 0.90)
- Explicitly Perceiving and Preserving the Local Geometric Structures for 3D Point Cloud Attack (AAAI 2024 · similarity 0.90)
- Imperceptible 3D Point Cloud Attacks on Lattice-based Barycentric Coordinates (AAAI 2025 · similarity 0.90)
- Towards Transferable Targeted 3D Adversarial Attack in the Physical World (CVPR 2024 · similarity 0.88)
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.
