Skip to main content

Biometric login

Biometric login authenticates a user on a device, such as a workstation, POS terminal, or handheld, against a small on-device gallery to replace or supplement passwords and PINs. Recognition runs offline on the local hardware, ensuring biometric data never leaves it.

How it works​

The device runs the whole pipeline locally, composing the Enrollment, Verification (1:1), and Identification (1:N) features, with Liveness to block presentation attacks — see those pages for the pipeline detail.

Two shapes, depending on whether an identity is claimed first:

  • 1:1 verification (personal device) — the user claims an identity (username, card, badge) and the device confirms their identity. App unlock, transaction confirmation.
  • 1:few identification (shared device) — a handful of enrolled users share one device, and the device identifies which of them is present. Shift login on a shared rugged handheld or POS terminal.

Components & modalities​

ModalitySDKPassive livenessAuto-capture
FaceSFE Toolkit✓✓
PalmSFE Toolkit✓✓

For the full modality / SDK / platform / accelerator matrix, see Supported Platforms and Supported Hardware Accelerators. For how each pipeline step maps to an SDK capability, see Verification (1:1), Identification (1:N), Liveness, and Enrollment.

Scenario tuning & constraints​

  • Prefer the fast template extractor for the best login experience. Login is 1:1 or 1:few against a small gallery, so the accuracy demand is far lower than large-gallery 1:N. Use the SFE Toolkit fast extraction solver (face_extraction_fast) to achieve the best performance. Step up to balanced or accurate only if match quality proves insufficient. See Face template extraction.
  • Keep the gallery small. Match only against the users who actually use the device — it keeps the 1:few match fast and login latency low.
  • Use the nearby liveness operating point. The device is held or used at close range, so choose the passive-liveness configuration for close-range capture — not the distant kiosk/turnstile point. See Liveness.
  • Auto-capture improves first-attempt success. The auto-capture component triggers acquisition once framing and quality are met, reducing failed logins. See Auto-capture.

Getting started​

  1. Get the SFE Toolkit for your target and set up licensing.
  2. Enroll each user — see Enrollment.
  3. Authenticate with Verification (1:1) (claimed identity) or Identification (1:N) (shared device), with a liveness check.
  4. Optionally add Auto-capture on Android / iOS for guided capture.