Research map / Adversarial robustness and security / Black-box and transfer attacks
Towards Effective Adversarial Textured 3D Meshes on Physical Face Recognition
Xiao Yang, Chang Liu, Longlong Xu, Yikai Wang, Yinpeng Dong, Ning Chen, Hang Su, Jun Zhu
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 |
|---|---|---|---|
| optimize adversarial meshes in 3DMM coefficient space rather than mesh space | AT3D · CelebA-HQ, Eye & Nose, ArcFace source, API-1 | attack success rate (%): 75.00 -> 87.00 | Improved |
| optimize adversarial meshes in 3DMM coefficient space rather than mesh space | AT3D · CelebA-HQ, Eye & Nose, ArcFace source, ResNet50 | attack success rate (%): 78.25 -> 93.75 | Improved |
| add chamfer, Laplacian, and edge-length losses to mesh-based optimization (ablation) | AT3D-M · CelebA-HQ, Eye & Nose, ArcFace source, API-1 | attack success rate (%): 75.00 -> 75.00 | No clear effect |
| initialize shape and texture from victim coefficients and use attacker pose (ablation) | AT3D-P · CelebA-HQ, Eye, ArcFace source, API | attack success rate (%): 45.25 -> 84.75 | Improved |
| optimize both shape and texture coefficients instead of shape coefficients only (ablation) | AT3D-P · CelebA-HQ, Eye & Nose, ArcFace source, API-1 | attack success rate (%): 84.25 -> 87.00 | Improved |
| optimize both shape and texture coefficients instead of texture coefficients only (ablation) | AT3D-P · CelebA-HQ, Eye & Nose, ArcFace source, API-1 | attack success rate (%): 77.50 -> 87.00 | Improved |
| add textured 3D meshes to the physical attack | AT3D · Physical attacker-to-victim pairs, API-1 | mean similarity (%): 22.21 -> 82.45 | Improved |
| use AT3D meshes in physical systems with recognition and defenses | AT3D · Physical attacker-to-victim pairs, S-1 | successful physical evaluations: 23/50 | Improved |
Closest papers by meaning
- Physical-World Optical Adversarial Attacks on 3D Face Recognition (CVPR 2023 · similarity 0.90)
- Towards Transferable Targeted 3D Adversarial Attack in the Physical World (CVPR 2024 · similarity 0.88)
- ProjAttacker: A Configurable Physical Adversarial Attack for Face Recognition via Projector (CVPR 2025 · similarity 0.87)
- 3D-Aware Adversarial Makeup Generation for Facial Privacy Protection (TPAMI 2023 · similarity 0.87)
- Non-Adaptive Adversarial Face Generation (NeurIPS 2025 · similarity 0.86)
- Seeing is Not Believing: Adversarial Natural Object Optimization for Hard-Label 3D Scene Attacks (CVPR 2025 · similarity 0.85)
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.
