Why Liveness Matters
A Selfie Isn't Proof of a Person
Identity fraud doesn't need sophisticated equipment any more. A high-quality printed photo, a deepfake video on a phone screen held up to the camera, or a silicon mask — each can defeat face match alone. Liveness is the layer that calls the bluff.
Active Liveness
The user is asked to perform a randomised challenge — blink, smile, turn head left, turn head right. The motion the camera captures is matched against the prompt that was issued. Static spoofs (printed photos, photo-on-screen) cannot respond; replay attacks fall apart because the challenge is unpredictable.
- Randomised challenge sequence
- Higher assurance — ideal for V-CIP, high-value lending
- Slightly longer UX (5–10 seconds)
Passive Liveness
A single selfie frame is analysed for liveness cues — texture, depth, micro-movement, environmental light reflection — without asking the user to do anything. Drop-off goes down significantly. Best paired with risk-based escalation that triggers active liveness on edge cases.
- Single-frame, no user action
- Low-friction UX — ideal for wallet onboarding, prepaid KYC
- Pair with risk-based escalation to active liveness
In-House & Benchmarked
What's Inside the Liveness Model
In-House Model
Trained, owned, and operated by eMudhra. No third-party biometric SDK in the data path. Model improvements roll out under our change-management cycle, not a vendor's.
Diverse Demographics
Training data spans Indian, Southeast Asian, Middle Eastern, and African demographics. Tuned to perform across skin tones, age groups, and facial features — not just majority-population.
Low-Light Optimised
Real-world onboarding happens at midnight on a 2G connection in a poorly-lit room. The model is benchmarked under those conditions, not just studio lighting.
ISO 30107-3 Aligned
Architected against the international standard for Presentation Attack Detection — with structured testing against print, replay, mask, and silicon attacks.
Benchmarked Metrics
FAR / FRR are tracked per release. Customers operating regulated journeys can request the benchmark sheet under NDA before integration.
SDK or API
iOS, Android, and Web SDKs for capture-side enforcement. Or pure server-side API for back-office review of customer-uploaded selfies.
Where It's Used
Typical Deployment Patterns
Mobile Banking Onboarding
Passive liveness on the selfie capture step; escalate to active for high-value accounts.
NBFC Loan Origination
Active liveness as a hard gate before disbursal. Embedded as part of V-CIP for unsecured lending.
Video KYC Pre-Check
Liveness sanity check before the video session starts — saves agent time, reduces drop-off.
