Liveness Detection
Ayrıca şöyle anılır liveness check, biometric liveness, active liveness
Liveness detection answers one narrow question in a biometric identity check: is the face (or voice) in front of the camera a live, physically present human, right now, or something faked? Without it, a face-match check alone can be beaten trivially, by holding up a photo of the person you're pretending to be. Liveness detection is what closes that gap.
There are two broad families. Active liveness asks the user to do something unpredictable, blink, turn their head, smile, and checks whether the video shows that motion happening naturally. Passive liveness (also called presentation attack detection, or PAD) tries to catch a fake without asking the user to do anything, by analyzing texture, depth cues, and other signals a printed photo or screen replay doesn't reproduce correctly. Active is simpler to build and easier for a stranger to understand; passive is harder to build well and is what actually stops a well-prepared attacker.
Who actually built this
The formal vocabulary, what counts as a "presentation attack," how to grade a defense against one, comes from ISO/IEC 30107, an international standard maintained by ISO/IEC JTC 1/SC 37 (the biometrics standards committee). The specific open-source models Solidus's pipeline is built on are InsightFace's ArcFace (for face recognition) and RetinaFace (for face detection), and Minivision AI's MiniFASNetV2 from the Silent-Face-Anti-Spoofing project (for passive anti-spoof scoring), all published, credited research, not Solidus's own cryptography or model architecture.
Solidus today
Solidus's active-challenge liveness gate, a randomized gesture sequence checked against real camera frames, is built and has been exercised end-to-end in testing. The passive anti-spoof layer exists in code but is deliberately claim-gated: it hasn't been through the adversarial review needed before Solidus will say, in public, that it stops spoofing. That is stated here as the honest current state, not softened.
See also
Presentation Attack is the specific thing liveness detection defends against, defined precisely. IDV and KYC are the processes liveness detection sits inside.
Nereden geliyor
Bunu başkası belirtti. Solidus bir uygulamasını yazdı.
Liveness detection is a standard biometrics technique for confirming that a face (or voice) presented to a camera or microphone belongs to a live human physically present, not a printed photo, a screen replay, a mask, or a synthetic deepfake. The taxonomy of what it defends against is formalized in ISO/IEC 30107 (Biometric presentation attack detection). Solidus's own gate combines a randomized action-sequence challenge (blink/turn/smile/nod, ordered by a cryptographically secure random generator so a pre-recorded clip can't guess the sequence) with face detection and matching built on InsightFace's open-source ArcFace and RetinaFace models, plus a passive anti-spoof model built on Minivision AI's open-source MiniFASNetV2 (Silent-Face-Anti-Spoofing project). None of that underlying model research belongs to Solidus, crediting it here is the point, not a footnote.
Bunu nasıl doğrularsınız
Test ağında çalışıyor. Ana ağda değil.
None a stranger can run standalone, this step lives inside a real KYC verification session, and no separate public demo of it exists. Stated honestly rather than invented.