Intelligence & Infrastructure

IRIS

Sovereign biometric watchlist. Zero cloud.

Real-time facial recognition against formal watchlists — running entirely on-device. No cloud API. No foreign biometric processing. Anti-spoofing enforced via blink detection (EAR threshold) and pixel-level spoof analysis. Every match produces an immutable audit entry with identity, confidence score, and timestamp. Multi-camera support via local daemon. Consent-first enrollment. PIPEDA / Law 25 compliant by design.

99.7%+

Confidence

<200ms

Match Latency

Cloud Hops

§ SPECIFICATION

Input

Live camera feed or multi-camera daemon (WebSocket)
Enrolled watchlist — AUTHORIZED / FLAGGED entries with face descriptors
Optional: blink threshold, spoof sensitivity, confidence cutoff

Constraints Verified

All compute on-device — no cloud API, no foreign biometric processing
Blink detection EAR threshold 0.22 — liveness required before match
Spoof pixel threshold 0.9 × 10 frames — static image attacks blocked
Watchlist encrypted at rest (AES via Web Crypto API)
Consent gate enforced before enrollment (PIPEDA / Law 25)

Output

AUTHORIZED / FLAGGED / UNKNOWN verdict per face
Euclidean distance score + confidence percentage
Immutable audit log entry (identity, timestamp, screenshot)
Real-time alert on FLAGGED match

§ SAMPLE PROOF ARTIFACT

ARTIFACT // IRIS-MATCH-01FAILURE DETECTED

// SAMPLE PROOF — IRIS ENGINE

Watchlist match — FLAGGED entity identified at access point

TargetLive camera feed, access control point

Conditionface_descriptor euclidean_distance = 0.24 < threshold (0.40), status = FLAGGED

VerdictSAT

SummaryIdentity confirmed. Distance 0.24, confidence 99.66%. Liveness verified. Immutable audit entry + screenshot created.

StatusReview-ready

→ View full artifact repository

Run IRIS on your system.

Formal engagement starts with a technical intake. We scope, configure, and deliver a proof artifact within the agreed SLA.

Request Briefing →