Research map / Adversarial robustness and security / Attacks on text and graphs
Isometric 3D Adversarial Examples in the Physical World
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 |
|---|---|---|---|
| constrain adversarial mesh to be epsilon-isometric via Gaussian curvature regularization (epsilon-ISO) vs Mesh Attack | PointNet · ModelNet40 | attack success rate (%): 88.10 -> 98.45 | Helped |
| use epsilon-ISO instead of Mesh Attack under simple defenses SRS | PointNet · SRS defense | attack success rate (%): 87.38 -> 79.85 | Hurt |
| use epsilon-ISO instead of Mesh Attack under IF-Defence | PointNet · IF-Defence | attack success rate (%): 49.06 -> 60.51 | Helped |
| replace EOT with MaxOT (worst-case transformation) | epsilon-ISO · ModelNet40 rotation | attack success rate (%): 76.20 -> 82.50 (PointNet) | Helped |
| use Bayesian optimization for initial transformation instead of random (ablation) | MaxOT · ModelNet40 PointNet | attack success rate (%): 80.59 -> 82.50 | Helped |
| MaxOT instead of EOT in physical world | PointNet · 3D-printed and scanned | attack success rate (%): 72.36 -> 80.12 | Helped |
| increase lambda2 from 0.2 to 1 (ablation) | epsilon-ISO · ModelNet40 PointNet | attack success rate (%): 98.45 -> 94.22 | Hurt |
| increase lambda3 from 0.8 to 16 (ablation) | epsilon-ISO · ModelNet40 PointNet | attack success rate (%): 98.45 -> 63.01 | Hurt |
| replace Gaussian curvature regularizer with mean curvature (ablation) | epsilon-ISO · ModelNet40 | naturalness | Hurt |
| epsilon-ISO instead of GeoA3 for black-box transfer from PointNet to PointNet++ | PointNet · transfer attack | attack success rate (%): 11.5 -> 47.6 | Helped |
Closest papers by meaning
- Curvature-Invariant Adversarial Attacks for 3D Point Clouds (AAAI 2024 · similarity 0.90)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.90)
- Explicitly Perceiving and Preserving the Local Geometric Structures for 3D Point Cloud Attack (AAAI 2024 · similarity 0.88)
- Manifold Constraints for Imperceptible Adversarial Attacks on Point Clouds (AAAI 2024 · similarity 0.87)
- Seeing is Not Believing: Adversarial Natural Object Optimization for Hard-Label 3D Scene Attacks (CVPR 2025 · similarity 0.87)
- Imperceptible 3D Point Cloud Attacks on Lattice-based Barycentric Coordinates (AAAI 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.