Research map / Adversarial robustness and security / Attacks on text and graphs
Towards Face Encryption by Generating Adversarial Identity Masks
Xiao Yang, Yinpeng Dong, Tianyu Pang, Hang Su, Jun Zhu, Yuefeng Chen, Hui Xue
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 |
|---|---|---|---|
| add multi-target optimization (MT-DIM over single-target DIM) | DIM · LFW black-box, ArcFace surrogate to MobileFace | Rank-1-T: 16.8 -> 18.8 | Improved |
| replace Center-Opt gain with greedy insertion (TIP-IM) (ablation) | Center-Opt · LFW black-box, ArcFace surrogate to MobileFace | Rank-1-T: 36.8 -> 69.8 | Improved |
| replace Center-Opt gain with greedy insertion (TIP-IM) (ablation) | Center-Opt · LFW black-box, ArcFace surrogate to ResNet50 | Rank-1-T: 28.8 -> 56.0 | Improved |
| replace Center-Opt gain with greedy insertion (TIP-IM) (ablation) | Center-Opt · LFW black-box, ArcFace surrogate to CosFace | Rank-1-T: 3.8 -> 11.4 | Improved |
| TIP-IM over best prior baseline MT-DIM | MT-DIM · LFW black-box, ArcFace surrogate to FaceNet | Rank-1-T: 9.6 -> 32.8 | Improved |
| increase MMD weight gamma from 0.0 to 3.0 (ablation) | TIP-IM (ArcFace) · LFW naturalness | PSNR (dB): 25.26 -> 27.63 | Improved |
| increase MMD weight gamma from 0.0 to 3.0 (ablation) | TIP-IM (ArcFace) · LFW naturalness | SSIM: 0.6520 -> 0.7817 | Improved |
| increase MMD weight gamma from 0.0 to 3.0 (ablation) | TIP-IM (ArcFace) · LFW naturalness | MMD (lower better): 0.7567 -> 0.7518 | Improved |
| increase gamma (stronger naturalness) (ablation) | TIP-IM · LFW targeted protection | Rank-1-T | Hurt |
| increase MMD batch size from 10 to 50 (ablation) | TIP-IM (CosFace) · LFW naturalness | SSIM: 0.8518 -> 0.6759 | Hurt |
| increase number of target identities (ablation) | TIP-IM (MobileFace surrogate) · black-box identity protection | Rank-1-T | Improved |
| use StyleGAN generated faces as targets instead of real identities (ablation) | TIP-IM · MegFace/LFW black-box | Rank-1-T: 49.6 | No clear effect |
Closest papers by meaning
- Adv-CPG: A Customized Portrait Generation Framework with Facial Adversarial Attacks (CVPR 2025 · similarity 0.87)
- FaceRSA: RSA-Aware Facial Identity Cryptography Framework (AAAI 2024 · similarity 0.87)
- Non-Adaptive Adversarial Face Generation (NeurIPS 2025 · similarity 0.87)
- Enhancing Facial Privacy Protection via Weakening Diffusion Purification (CVPR 2025 · similarity 0.87)
- Transferable Facial Privacy Protection against Blind Face Restoration via Domain-Consistent Adversarial Obfuscation (ICML 2024 · similarity 0.86)
- IDProtector: An Adversarial Noise Encoder to Protect Against ID-Preserving Image Generation (CVPR 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.
