Research map / Point clouds / Point cloud classification
Towards a 3D Transfer-based Black-box Attack via Critical Feature Guidance
Shuchao Pang, Zhenghan Chen, Shen Zhang, Liming Lu, Siyuan Liang, Anan Du, Yongbin Zhou
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 |
|---|---|---|---|
| increase perturbation budget epsilon from 0.18 to 0.45 (ablation) | CFG · ModelNet40 | transferability score: 53.1 -> 63.6 | Improved |
| increase perturbation budget epsilon from 0.18 to 0.45 (ablation) | PF-Attack · ModelNet40 | transferability score: 40.2 -> 46.8 | Improved |
| use CFG instead of PF-Attack | PointNet source · ModelNet40 | transferability score (eps 0.18): 40.2 -> 53.1 | Improved |
| use CFG instead of PF-Attack | PointNet source · ModelNet40 | transferability score (eps 0.45): 46.8 -> 63.6 | Improved |
| use CFG instead of PF-Attack against SRS defense | PointNet source · ModelNet40 | attack success rate: 55.7 -> 79.2 | Improved |
| use CFG instead of PF-Attack against adversarial training defense | PointNet source · ModelNet40 | attack success rate: 10.4 -> 22.6 | Improved |
| use CFG instead of PF-Attack on real-world scans | PointNet source, DGCNN target · ScanObjectNN | attack success rate: 40.3 -> 58.3 | Improved |
| use CFG instead of PF-Attack on state-of-the-art targets | PointNet source, CurveNet target · ModelNet40 | attack success rate: 22.7 -> 29.2 | Improved |
| use CFG instead of PF-Attack on DGCNN target | PointNet source · ModelNet40 | attack success rate: 24.8 -> 39.3 | Improved |
| use CFG instead of AdvPC when PointNet++ MSG is source and DGCNN target | PointNet++ MSG source · ModelNet40 | attack success rate: 39.6 -> 25.3 | Hurt |
| use CFG instead of PF-Attack (generation time) | PointNet source · ModelNet40 | running time (min): 0.10 -> 0.07 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- AdvPC — AdvPC: Transferable Adversarial Perturbations on 3D Point Clouds (ECCV 2020) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Transferable 3D Adversarial Shape Completion using Diffusion Models (ECCV 2024 · similarity 0.93)
- AdvPC: Transferable Adversarial Perturbations on 3D Point Clouds (ECCV 2020 · similarity 0.91)
- Imperceptible Transfer Attack and Defense on 3D Point Cloud Classification (TPAMI 2022 · similarity 0.91)
- Generating Transferable 3D Adversarial Point Cloud via Random Perturbation Factorization (AAAI 2023 · similarity 0.90)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.87)
- Curvature-Invariant Adversarial Attacks for 3D Point Clouds (AAAI 2024 · similarity 0.87)
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.
