Research map / Federated learning and privacy / Privacy attacks and secure learning
Personalized Privacy Protection Mask Against Unauthorized Facial Recognition
Ka-Ho Chow, Sihao Hu, Tiansheng Huang, Ling Liu
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 |
|---|---|---|---|
| select ensemble of FR models by highest focal diversity rather than lowest (two-model) (ablation) | Chameleon P3-Mask generation · FaceScrub unknown FR models | mean PSR (%): 72.56 -> 96.25 | Improved |
| select ensemble of FR models by highest focal diversity rather than lowest (three-model) (ablation) | Chameleon P3-Mask generation · FaceScrub unknown FR models | mean PSR (%): 76.42 -> 98.27 | Improved |
| choose the highest-focal-diversity two-model team over a random team (ablation) | Chameleon · FaceScrub | protection strength: over 20% stronger | Improved |
| unmask protected images with the correct P3-Mask key (ablation) | Chameleon · FaceScrub M. Baccarin | mean FR accuracy (%): 6.82 -> 100.00 | Improved |
| unmask protected images with an incorrect key (ablation) | Chameleon · FaceScrub M. Baccarin | mean FR accuracy (%): 6.82 -> 1.14 | Hurt |
| cross-image per-user mask replaces per-image optimization | Chameleon vs TIP-IM and Fawkes · protection time per face | seconds: 105.12 -> 0.0076 | Improved |
| add perceptibility (SSIM) optimization and ML pipeline awareness versus OPOM | Chameleon · image quality example | SSIM: 0.8839 -> 0.9493 | Improved |
| adaptive adversary applies JPEG compression, Gaussian filter or median filter to washed-out protected images | Chameleon · FaceScrub | PSR | No clear effect |
| OPOM masks under lossy end-to-end ML pipeline operations | OPOM · FaceScrub and LFW | PSR: PSR can drop by 33% | Hurt |
Closest papers by meaning
- Validating Privacy-Preserving Face Recognition under a Minimum Assumption (CVPR 2024 · similarity 0.86)
- Privacy-Preserving Face Recognition Using Trainable Feature Subtraction (CVPR 2024 · similarity 0.85)
- Privacy-Preserving Optics for Enhancing Protection in Face De-Identification (CVPR 2024 · similarity 0.85)
- Disguise without Disruption: Utility-Preserving Face De-identification (AAAI 2024 · similarity 0.85)
- FSFM: A Generalizable Face Security Foundation Model via Self-Supervised Facial Representation Learning (CVPR 2025 · similarity 0.84)
- Asymmetric Invertible Threat: Learning Reversible Privacy Defense for Face Recognition (NeurIPS 2026 · similarity 0.84)
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.
