Skip to main content

Biometric Functions

Detailed reference for the biometric capabilities exposed by Service & ABIS.

Multimodal support​

Native processing for the following modalities, supporting both images and templates:

  • Face — verification, identification; ICAO-aligned quality validation; PAD/liveness
  • Fingerprint — slap and rolled; NFIQ2 quality scoring
  • Contactless Fingerprint — capture via standard cameras without specialized hardware; same matching pipeline as contact fingerprint
  • Palmprint — verification and identification
  • Iris — verification and identification; deep-learning-based feature extraction
  • DNA — image and template ingestion (where applicable)

Document handling​

Travel and identity documents are first-class records in the platform alongside biometrics:

  • Supported document types — Passport (Ordinary, Diplomatic), National ID, Visa, Residence Permit, Other (see the Document entity for the documentType enum and full attribute schema)
  • Capture pipeline — OCR for printed text, MRZ parsing, NFC chip reading (ICAO Doc 9303 eMRTD; TD1/TD2/TD3); document authenticity checks
  • Persistence — Documents are stored as records linked to a Traveler; the raw MRZ payload is preserved alongside the parsed fields. storedModalities on the Document references the biometric records derived from chip data (e.g., the eMRTD portrait)
  • Lifecycle — documentStatus tracks Valid, Expired, Revoked, Flagged for operational use during border processing

Automatic extraction & template management​

Biometric samples are automatically extracted into standardized templates and stored within the matcher component for processing. The system maintains:

  • Per-modality quality thresholds (configurable)
  • ICAO Doc 9303 portrait quality compliance for face
  • NFIQ2 fingerprint quality scoring
  • IREX-aligned iris quality assessment

Flexible biometric record model​

Every enrollment record (biometric record) comprises three configurable groups:

  • Demographic data
  • Biometric data — templates and references to images
  • Metadata — lifecycle, status, audit context

Verification and Identification modes​

Verification (1:1 / 1:few)​

Confirms whether a presented probe matches a specific known reference (or a small set).

Identification (1:N)​

Searches the full database or logically isolated galleries to find candidate matches.

Targeted identifications​

Limits search scope to specific galleries to reduce latency and optimize system throughput.

Scoring hierarchy​

Granular scoring model with multiple tiers:

  • Partial scores (instance-level) — per individual modality instance (e.g., per fingerprint position)
  • Modality-specific scores — aggregated across all instances of a modality
  • Fused multimodal score — combined score across all modalities for a candidate

Configurable matching threshold​

Parameters can be tuned during deployment:

  • Acceptance / rejection thresholds per modality
  • Multimodal fusion weights
  • Candidate count limiters for search results
  • Per-gallery threshold overrides

Stateless / Semi-stateful / Stateful matching​

ModeProbeReferencesUse case
Fully statelessSupplied in requestSupplied in requestQuick 1:1 verification with no persistence
Semi-statefulSupplied in requestRetrieved from stored recordsIdentification or verification against the gallery
Fully statefulPulled from stored recordsPulled from stored recordsPersistent operations with audit trails

Proprietary template support​

Full compliance with PFT III and international standards for seamless interoperability. Existing templates from migration sources can be ingested directly.

See also​