Face Biometrics in eKYC¶
Definition¶
Face biometrics in eKYC encompasses the complete pipeline of capturing a user's face, verifying they are a live person (liveness detection), and matching their face against their identity document photo — the core technology enabling remote identity verification.
The Face Verification Pipeline¶
graph LR
A[Selfie Capture] --> B[Face Detection<br/>Locate face in image]
B --> C[Quality Check<br/>Blur, lighting, pose]
C --> D[Liveness Detection<br/>Real person vs spoof]
D --> E[Face Embedding<br/>Extract 512-d vector]
E --> F[Face Matching<br/>Compare with ID photo]
F --> G{Match?}
G -->|Score > threshold| H[✅ Verified]
G -->|Score < threshold| I[❌ Rejected]Key Components¶
| Component | Purpose | Key Challenge |
|---|---|---|
| Face detection | Find face in image | Multiple faces, small faces, occlusion |
| Face quality | Ensure image is usable | Low light, blur, extreme angles |
| Face liveness | Prove it's a live person | Deepfakes, printed photos, 3D masks |
| Face recognition | Match selfie to ID photo | Cross-quality (selfie vs printed ID photo), cross-age |
| Face deduplication | Check against existing database (1:N) | Scale (millions of comparisons), speed |
Key Takeaways¶
Summary
- Face verification is a 5-step pipeline: detect → quality → liveness → embed → match
- Liveness detection is the most critical security component — prevents spoofing
- Cross-quality matching (selfie vs printed ID photo) is the key accuracy challenge
- The entire pipeline must run in < 3 seconds for good user experience