Skip to main content

Identity verification

Assayra identity verification is a workflow-composed set of evidence controls—not a single opaque “verified” flag.

Document intelligence​

The document path can combine:

  • declared MIME and file-signature agreement;
  • active-content and malformed-image rejection;
  • production malware scanning;
  • image geometry, quality and entropy checks;
  • OCR field extraction and confidence;
  • ICAO MRZ parsing and checksum validation;
  • expiry, issuer/country and declared-field comparison;
  • document portrait extraction and evidence-hash binding; and
  • country/document-template checks and uncertainty routing.

The application result exposes versioned signals and evidence receipts. Do not reduce low-confidence or unavailable results to pass in client code.

Human presence and face comparison​

The hosted journey continuously checks that exactly one face is centred, large enough and usable. It automatically captures the centre, requested pose and return-to-centre evidence; the applicant does not manually operate a shutter. Assayra then recomputes capture binding server-side. The assurance route can include:

  • face detection and subject continuity;
  • document-portrait versus live-subject embedding comparison;
  • multi-frame presentation-attack detection;
  • duplicate, replay and virtual-camera/injection indicators;
  • capture integrity and frame relationship checks; and
  • synthetic-media risk fusion derived from PAD and capture-integrity signals.

The client captures; it never declares a biometric pass. Thresholds and required signals are selected by the approved workflow and Country Assurance Pack.

The bundled baseline does not claim an independently qualified, standalone deepfake detector. A workflow may make that claim only after a distinct model route and representative evaluation evidence have been approved for its exact country, device and attack scope.

Every active-liveness result includes an evidence-bound comparison between the live SFace embeddings and the portrait extracted from the submitted identity document. A workflow that does not require active liveness can add the standalone Face likeness control: it automatically captures a three-frame live burst, verifies one continuous subject and performs the same document-to-live comparison without asserting PAD or challenge completion.

Declared details​

Collect declared identity fields before or alongside document evidence. Assayra compares normalized OCR/MRZ values with the declared person and surfaces explainable field-level agreement. Your workflow decides which mismatches reject, route to review or require recollection.

Delivery choices​

  • Hosted: Assayra renders capture, validation and recovery UI.
  • Drop-in: SDK hosts the same journey inside your product.
  • Headless: your UI calls the scoped session operations while protected capture contracts remain mandatory.
  • NFC: native SDK reads passport chip data, SOD and data-group evidence for server-side trust validation.

Result handling​

Treat the final application status as the policy outcome. Preserve individual signals for explanation, investigations and monitoring; do not implement a second hidden decision engine in frontend code.