Research map / Federated learning and privacy / Privacy attacks and secure learning
IDFace: Face Template Protection for Efficient and Secure Identification
Sunpill Kim, Seunghun Paik, Chanwoo Hwang, Dongsoo Kim, Junbum Shin, Jae Hong Seo
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 |
|---|---|---|---|
| lower query truncation beta from 341 to 63 (alpha=341) (ablation) | IDFace CKKS · 1M identities | identify time (s): 0.753 -> 0.126 | Improved |
| lower query truncation beta from 341 to 63 (ablation) | IDFace · CFP-FP | accuracy: 99.24% -> 98.20% | Hurt |
| lower query truncation beta from 341 to 63 (ablation) | IDFace · IJB-C TPIR@1e-2 | TPIR: 95.95% -> 92.87% | Hurt |
| lower query truncation beta from 341 to 63 (ablation) | IDFace · AgeDB | accuracy: 97.97% -> 96.48% | Hurt |
| lower query truncation beta from 341 to 63 (ablation) | IDFace · LFW | accuracy: 99.78% -> 99.73% | No clear effect |
| use top-341 ternary transform instead of full 512 components (ablation) | IDFace · CFP-FP | accuracy: 98.96% -> 99.24% | Improved |
| use top-341 ternary transform instead of full 512 components (ablation) | IDFace · IJB-C TAR@1e-3 | TAR: 97.93% -> 98.27% | Improved |
| apply almost-isometric ternary transform plus HE (IDFace beta=63) vs plain matching | IDFace CKKS · 1M identities | identify time (s): 0.063 -> 0.126 | Hurt |
| use IDFace (341) CKKS instead of HERS (512) | IDFace · 1M identities | identify time (s): 48.9 -> 0.753 | Improved |
| use IDFace (63) CKKS instead of MFBR-ID | IDFace · 1M identities | identify time (s): 0.545 -> 0.126 | Improved |
| use IDFace (63) CKKS instead of MFBR-ID | IDFace · 1M identities | storage (GB): 132.1GB -> 4.125GB | Improved |
| reduce HERS feature dimension from 512 to 128 | HERS · 1M identities | identify time (s): 48.9 -> 12.3 | Improved |
Closest papers by meaning
- CryptoFace: End-to-End Encrypted Face Recognition (CVPR 2025 · similarity 0.88)
- SlerpFace: Face Template Protection via Spherical Linear Interpolation (AAAI 2025 · similarity 0.86)
- FaceRSA: RSA-Aware Facial Identity Cryptography Framework (AAAI 2024 · similarity 0.86)
- How Private is Private? A Comparative Study for Face De-Identification (NeurIPS 2026 · similarity 0.85)
- Privacy-Preserving Face Recognition Using Trainable Feature Subtraction (CVPR 2024 · similarity 0.84)
- Gaussian Differentially Private Human Faces Under a Face Radial Curve Representation (ICLR 2025 · similarity 0.83)
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.
