stile

Identity verification (IDV)

Identity verification software

Identity verification software confirms a person is who they claim to be with three checks: the ID, a live person, and a face match. Government testing shows where most systems fail. Stile is built to close those gaps.

  • iBeta-certified liveness technology
  • Digital IDs
  • AI document checks
A person at a pickup counter holds a phone showing an identity verified result.
Independent research

Commercial systems tested by DHS, names hidden. Stile was not tested.

How the typical system performed on each check

  • Selfie match to document · Strongest99.99%of imposters rejected99% of genuine people passed · 7.8 million+ imposter attempts
  • Document validation · Most genuine failures87%of fake IDs caught79% of genuine people passed · 1,900+ IDs from all 50 states
  • Presentation attack detection · Weakest53%of the strongest spoof caught96% of genuine people passed · 1,500+ attack presentations

Two different checks. 99.99% is face matching: telling one person from another. 53% is presentation attack detection: telling a live person from a photo, screen or mask. A system can be excellent at one and weak at the other.

Together, about 75% of genuine people got through. Success multiplies: 79% × 96% × 99%. In DHS's 2024 test, only 18 of 3,024 system combinations met both usability and security minimums.

A woman holds a silicone mask of an older man in front of her face while taking a selfie; the attempt is marked as blocked.

Presentation attack detection · Stile's biggest advantage

The weakest check for most systems is Stile's strongest

The typical tested system stopped about half of the strongest spoof attacks. Stile's face scanning uses liveness technology certified to iBeta PAD Level 1 and Level 2, the lab tests for presentation attacks.

  • Typical tested system, strongest attack (DHS)53% stopped
  • Liveness technology Stile uses, iBeta Level 1 and 2 lab attacks0 of 1,650 got through

Different tests: an iBeta lab certification of the liveness technology Stile uses, set beside DHS's rally of other systems. Side by side, not head to head.

  • Level 1

    Everyday attacks

    Printed photos, phone and tablet screens, paper masks. Pass mark: no attack gets through.

  • Level 2

    Skilled attackers

    3D-printed, resin and latex masks, days of preparation. Pass mark: at most 1%.

  • Beyond the lab

    Virtual camera checks

    Injected video bypasses the camera and sits outside lab testing, so Stile checks the capture channel too.

Document validation

Fake IDs are getting harder to see

The typical tested system caught 87% of fake IDs and caused more failures for genuine people than any other step. Stile's document model, trained with NVIDIA AI tools on thousands of IDs, is built for the fakes template checks miss.

  • Typical tested system (DHS)87% caught
  • Stile document model96% caught

Stile figure from internal testing on 127 fake IDs, 08/2026. Different test from DHS: side by side, not head to head.

Close-up of two ID cards: a genuine card with a crisp hologram marked with a check, and a counterfeit with a flat hologram marked with a cross.
  • Phones miss security features

    Ultraviolet and infrared marks that ID scanners read don't show in a phone photo.

  • AI-made fakes

    Edited photos, flat or misplaced holograms, and generated IDs that look right on screen.

  • Barcode against print

    A fake barcode can scan fine. Stile checks it matches the printed details.

  • Retake, not reject

    A blurry or glared photo is retaken in the same session, so genuine people aren't turned away.

A hand holds a phone showing a request to share age over 21 and a photo from a digital ID.

Digital IDs

Or skip the ID photo entirely

With a mobile driver's license, the person approves sharing in their phone wallet and the wallet sends data signed by the issuing state. Stile checks the signature, so there is no card photo to fake or misread.

  • Signed by the state

    Removes the weakest document step

  • No glare or wear

    Fewer genuine people fail

  • Only what you ask

    "Over 21" without the address

Issued state by state and works with compatible wallets; photo ID stays as the fallback, and a face check links the phone holder to the license.

Selfie match to document

99.99% on average, but not equal for everyone

Face matching is the most mature check. The average hides who gets turned away.

  • Skin tone

    In a GSA study of 3,991 people, one of five systems rejected darker skin tones well above its own average.

  • Age

    Faces change between the ID photo and today's selfie, and research finds weaker matching for the youngest people.

  • The ID photo

    Holograms and worn cards hide the face. The worst tested system couldn't read the ID photo 37.47% of the time.

Stile retakes poor photos in the same session. Stile has not yet been evaluated for demographic differences.

Why all three

Each check stops a different attack

An attacker only has to beat the weakest check, so 99.99% face matching doesn't help if liveness lets a mask through. And every step turns some genuine people away, so each one has to be strong.

79% × 96% × 99% ≈ 75%Genuine people passing all three, typical tested system

DocumentLivenessFace match
Fake ID with the attacker's photoStops itMissesMisses
Photo or mask of the real holderMissesStops itMisses
Someone else's real IDMissesMissesStops it

Stile, check by check

What Stile does, and what it does not claim

CheckStileWhat we do not claim
Document validationDigital ID first, verified by signature. Photo ID fallback with an AI document model and barcode cross-check, for all 50 US states and dozens of other jurisdictions.mDL availability depends on state, wallet and device.
Presentation attack detectionLiveness technology certified to iBeta PAD Level 1 and Level 2, plus virtual camera checks.Certification covers the tested liveness component and setup. No deepfake detection claim.
Selfie match to documentSelfie compared with the ID photo; poor photos retaken in the same session.Not yet evaluated for demographic differences.
All threeOne flow and one signed result, tied to the request that triggered it.Stile was not part of DHS testing.

Before you buy

Questions to ask any provider, including us

  • Document

    What share of fake IDs do you accept, and of genuine IDs do you reject? Do you accept digital IDs?

  • Liveness

    Which iBeta level, for which version and device? What about virtual cameras?

  • Face match

    How do results differ by age and skin tone?

  • All three

    What share of real users finish on their own phones?

What your team gets

A clear answer, not a copy of someone's ID

Your team sees whether the person was verified and which checks passed, right where the request is waiting. You decide what happens next.

  • You get

    Verified, needs review or not verified. Which checks passed. When it happened, for your records.

  • You don't have to keep

    Photos of the ID, face scans or ID numbers. Stile deletes ID images and face scans by default once there is a decision.

Payment-detail change · Vendor portal

Identity verified

Document
Passed
Face match
Passed
Liveness
Passed

Ready for your team's approval.

FAQ

Common questions.

Still have a question? Book a demo

Software that checks a person is real and is who they claim to be: it validates a government ID, confirms a live person is present, and matches their selfie to the ID photo. Some products also accept digital IDs from a phone wallet.

In DHS testing of commercial systems, selfie match to document was the most accurate: the typical system rejected 99.99% of imposters. Presentation attack detection was the weakest at stopping attacks, catching 53% of the strongest spoof. Document validation caught 87% of fake IDs. These figures describe the systems tested, not Stile.

Face matching checks that the selfie and the ID photo are the same person. Liveness (presentation attack detection) checks that the selfie comes from a live person, not a photo, screen or mask. A system can match faces well and still be fooled by a mask. More on biometric verification.

Stile's face scanning uses liveness technology certified by iBeta to ISO/IEC 30107-3 presentation attack detection Level 1 and Level 2. The certification covers the tested component and setup. What the levels test.

A mobile driver's license or state ID stored in a phone wallet. The wallet sends data signed by the issuing state, so there is no card photo to fake. Availability depends on the state and the wallet. How digital ID verification works.

No. DHS ran the rally with each system packaged to run inside its own test setup, and part of Stile’s checks runs on cloud services that can’t be packaged that way. The DHS results on this page describe other commercial systems. Ask any provider, including us, which independent tests their checks have passed.

Send a link or add one button where the request happens, then act on the result. See the integration guide.

DHS S&T, A Quantitative Framework for Evaluating Remote Identity Validation Systems (2025); the DHS RIVR-1 results briefing, September 30, 2026; Maryland Test Facility RIVR results; and GSA and CITeR, performance and equity of remote identity verification (IJCB 2024).

Get started

Start verifying for free

1,000 free checks every month. Document, face match and liveness in one flow. No credit card required.