Interactive Verification Suite
Progressive LoA & Human Uniqueness Demo
Test how Web YID proves that a real, unique human is present without central biometric custody. Move your pointer and type to accrue passive evidence, or trigger the step-up ceremonies below.
1. Action Under Test
2. Step-Up Verification Ceremonies
- Earned Assurance
- —
- Required Assurance
- —
- Step Up Target
- None
- Trust Metric
- —
Confidence Meter
LoA — / 4
- LoA 4 Sovereign Enclave Anchor ~90% of offered
- LoA 3 Passive Biometrics 87%
- LoA 2 Hardware Passkey 93%
- LoA 1 Behavioural Humanhood 98.5%
- LoA 0 Passive Ambient 100%
Value precedes the ask. No camera before value accrues — and the face is asked for once, ever.
Zeroization probe
Biometrics destroyed in local RAM. Zero data leaves your device.
- Probe reported
- yes
- Non-zero bytes
- 0
- Retention
- none (no template, no frame)
Cryptographic Anti-Sybil Architecture
How Human Uniqueness Is Proven Without Facial Storage
How do we know a human is unique across 750,000+ accounts without keeping a database of face photos? Web YID uses Oblivious Pseudo-Random Functions (OPRF) to derive deterministic, unlinkable nullifiers.
OPRF Sybil Nullifier
The AdaFace 512-D embedding is passed through an Oblivious PRF protocol evaluated with Subnet 54: N = OPRF(K_subnet, bio_features). The same human face always generates the exact same nullifier,
allowing instant O(1) duplicate detection without ever saving face images.
Human-Time Economics
Sybil cost scales with human presence: C_sybil = Θ(s). Bot farms cannot parallelize 15-frame
temporal liveness and neuromuscular Fitts's law kinematics. Faking 10,000 identities requires 10,000 real human
sessions.
Cross-RP Unlinkability
Per Invariant P4, nullifiers are salted by Relying Party ID. Your identity on dFusion AI cannot be correlated or linked to your identity on other apps. Privacy is a mathematical property, not a policy promise.
Interactive Uniqueness & Sybil Test Console