Research map / Point clouds / Point cloud classification
AdvPC: Transferable Adversarial Perturbations on 3D Point Clouds
Abdullah Hamdi, Sara Rojas, Ali Thabet, Bernard Ghanem
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 AdvPC data adversarial loss instead of 3D-Adv | point cloud attack · ModelNet40 transferability matrix | average transfer success rate: 11.5% -> 24.9% | Improved |
| use AdvPC instead of KNN Attack | point cloud attack · ModelNet40 transferability matrix | average transfer success rate: 8.92% -> 24.9% | Improved |
| use AdvPC instead of 3D-Adv, DGCNN victim to PointNet++ (SSG) | point cloud attack · eps_inf 0.18 | attack success rate: 31.2 -> 64.4 | Improved |
| use AdvPC instead of 3D-Adv, DGCNN victim to PointNet++ (MSG) | point cloud attack · eps_inf 0.18 | attack success rate: 11.2 -> 46.0 | Improved |
| use AdvPC instead of 3D-Adv, PointNet++ (MSG) victim to SSG | point cloud attack · eps_inf 0.18 | attack success rate: 28.4 -> 54.8 | Improved |
| use AdvPC instead of 3D-Adv, DGCNN victim to PointNet | point cloud attack · eps_inf 0.45 | attack success rate: 9.6 -> 32.8 | Improved |
| use AdvPC instead of 3D-Adv against SRS defense | point cloud attack · DGCNN, eps_inf 0.18 | attack success rate: 43.2 -> 80.0 | Improved |
| use AdvPC instead of 3D-Adv against adversarial training | point cloud attack · DGCNN, eps_inf 0.18 | attack success rate: 7.2 -> 39.6 | Improved |
| use AdvPC instead of 3D-Adv against the newly trained AE defense | point cloud attack · DGCNN, eps_inf 0.18 | attack success rate: 9.2 -> 17.2 | Improved |
| use AdvPC instead of 3D-Adv against SOR defense | point cloud attack · DGCNN, eps_inf 0.18 | attack success rate: 18.8 -> 36.8 | Improved |
| use AdvPC instead of 3D-Adv against DUP Net defense | point cloud attack · DGCNN, eps_inf 0.18 | attack success rate: 28 -> 43.6 | Improved |
| use AdvPC instead of 3D-Adv with no defense (victim success) | point cloud attack · DGCNN, eps_inf 0.18 | attack success rate: 100 -> 94.8 | Hurt |
Later papers that compared with this method
5 later accepted papers report AdvPC as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Towards a 3D Transfer-based Black-box Attack via Critical Feature Guidance (ICCV 2025 · better on 2 of 2 shared results)
- Transferable 3D Adversarial Shape Completion using Diffusion Models (ECCV 2024 · better on 2 of 2 shared results)
- Hiding Imperceptible Noise in Curvature-Aware Patches for 3D Point Cloud Attack (ECCV 2024 · better on 1 of 1 shared result)
- Generating Transferable 3D Adversarial Point Cloud via Random Perturbation Factorization (AAAI 2023 · better on 2 of 2 shared results)
- Shape Prior Guided Attack: Sparser Perturbations on 3D Point Clouds (AAAI 2022 · better on 8 of 8 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
- Generating 3D Adversarial Point Clouds (CVPR 2019 · similarity 0.92)
- Generating Transferable 3D Adversarial Point Cloud via Random Perturbation Factorization (AAAI 2023 · similarity 0.92)
- Benchmarking and Analyzing Robust Point Cloud Recognition: Bag of Tricks for Defending Adversarial Examples (ICCV 2023 · similarity 0.91)
- Transferable 3D Adversarial Shape Completion using Diffusion Models (ECCV 2024 · similarity 0.91)
- Robust Adversarial Objects against Deep Learning Models (AAAI 2020 · similarity 0.91)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.89)
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.
