Liveness detection
Detection that checks whether the face on camera is a real person.
Checks whether the face on camera is a real person rather than a photo, screen or mask. It produces evidence and reasoning; it doesn't decide, and the session goes to review.
Model card
- Attack detection
- 97%
- Operation
- Frame by frame
- Output
- Evidence and reasoning
- Decision
- With a human
What does it do?
Frame-by-frame review
Each frame is assessed separately.
Spoof detection
Against presentation attacks such as photos, screens and masks.
Doesn't decide
It doesn't say pass or fail; it returns evidence.
Verified setup
The models in use are verified in a tamper-proof way.
No unnecessary data
Requests containing extra data are rejected.
Encrypted storage
Biometrics are encrypted with a customer-specific key.
How does it work, step by step?
- 1
Stage 1
Frames
The app sends frames.
- 2
Stage 2
Face
The face in the frame is located.
- 3
Stage 3
Check
The face is tested to see whether it is real.
- 4
Stage 4
Evidence
The result and reasoning are returned.
- 5
Stage 5
Review
The decision goes to a human.
Meet Liveness detection.
Let's review with our team how the model would work in your business.