stile

Identity Fraud Protection

Confirm the person behind a high-risk action, return signed evidence, and let customer policy decide what happens next.

01 · Request

POST /v1/verification_sessions
{
  "type": "identity",
  "workflow_id": "wf_identity_step_up"
}

02 · Hosted challenge

Verify your identity
  • Document or mDL
  • Face and liveness
  • Configured context

Identity check complete

03 · Signed result

verification_session.verified
signature check
verified server-side
evidence
method results
decision
customer policy decides
Stile returns signed identity evidence. The customer's policy decides what happens next.

Proof

Evidence you can inspect before policy acts

Inspect credential paths, full results, signed delivery, and retention before evaluating fit.

Credential choice

Government ID and mDL paths share a compact session event.

Configured checks

Selected checks run; full available results require authenticated session expansion.

Compact signed webhook

Carries status, current method, audit/workflow context, timestamps, and an available document or NFC summary.

Authenticated detail retrieval

Full available results stay outside the webhook; retention remains method-specific and contract-scoped.

Why identity fraud is outrunning fraud-scoring stacks

Risk systems use device, network, payment, and account history to flag actions; they do not establish who presents a credential.

The Entrust 2026 Identity Fraud Report reports rising digital forgery and deepfake pressure: context, not Stile performance.

Signals find the moment

Synthetic identities can look ordinary; customers can look unusual; genuine documents can be misused. Scores decide when evidence warrants friction.

Evidence tests the person present

A high-risk action triggers an identity challenge; the user presents a credential; configured checks run; signed evidence returns; customer policy makes the final decision. Stile does not replace monitoring or authorization policy.

How it works

Challenge only when risk warrants it

A risk signal can request stronger evidence at the moment of action. Stile returns the signed result; the customer's policy makes the decision.

  1. Trigger

    A risk or workflow system requests an identity challenge for a high-risk action.

  2. Verify

    The user presents a Government ID or mobile driver’s license. Stile runs the configured document, face, liveness, and network checks.

  3. Decide

    Stile returns signed identity evidence. Customer policy makes the final decision to approve, escalate, delay, or decline.

Next action: inspect the returned evidence, then Read integration docs or Request technical review.

Interactive simulation

Workflow configuration

Credential path

Configured checks

  • Government ID captureRequired
  • Data extractionNormalized
  • Selfie-to-ID comparisonThreshold 0.80
  • Required
  • IP geographyRegion only
  • VPN/proxy analysisSignal

Customer routing

On check failure: Return signed failed evidence and flag the action for additional review in customer policy.

User challenge

Stile
Simulated demo

Confirm your identity

This check was requested before account recovery can continue.

Choose method

Powered by Stile · Signal-only return

Session report

Identity check complete

Synthetic session · Government ID

Credential
Government ID barcode cross-referencePassed
Data extraction
Name, date of birth, expiry normalizedComplete
Selfie-to-ID
Score 0.94 · above threshold 0.800.94
Liveness
Score 0.98 · above threshold 0.80Passed
IP geography
California, US · synthetic regionContext
VPN/proxy
No proxy signal in fixtureClear

Simulated compact event body

Method rows show raw simulated check evidence. Compact webhook booleans are gated by the overall session outcome; confidence is the overall document score, not face similarity.

{
  "id": "evt_demo_verified",
  "object": "event",
  "type": "verification_session.verified",
  "data": {
    "object": {
      "id": "vks_demo_identity",
      "object": "verification_session",
      "status": "verified",
      "type": "identity",
      "client_reference_id": null,
      "current_method": "document_capture",
      "age_tier": null,
      "jurisdiction": null,
      "jurisdiction_audit_declared": null,
      "jurisdiction_audit_ip_derived": null,
      "jurisdiction_audit_resolved": null,
      "jurisdiction_audit_source": null,
      "jurisdiction_audit_mismatch": null,
      "verification_path": null,
      "workflow_version_id": null,
      "livemode": false,
      "created": 1767225600,
      "completed_at": 1767225604,
      "verification_result": {
        "method": "document_capture",
        "confidence": 0.99,
        "age_verified": false,
        "age_estimate": null,
        "identity_verified": true,
        "face_match_passed": true,
        "barcode_cross_ref_match": true,
        "liveness_score": 0.98
      }
    }
  },
  "created": 1767225604
}

Signed result handoff

Production webhook signatures arrive in HTTP headers and must be verified server-side.

Data minimization

Raw captures
Deleted after completion by default
Signed outcome and audit metadata
Retained under the customer’s configured and contractual scope

Customer policy makes the final decision from the returned evidence. A failed configured check can route the action to additional review.

Verification methods

Evidence paths configured for the action

Government ID and mDL are alternative rails. The signed event stays compact; retrieve available full results through authenticated session expansion. No check decides fraud or authorization.

Method 01

Government ID

Establishes
Captures and extracts a supported credential, then checks visible and machine-readable elements; it does not prove the presenter alone.
Signed output
Compact session plus available document summary.

Method 02

Mobile driver's license

Establishes
Requests selected wallet attributes and validates issuer-signed presentation, subject to issuer, device, jurisdiction, and protocol.
Signed output
Compact session status and current method; wallet detail excluded.

Method 03

Selfie-to-ID comparison

Establishes
Compares a live selfie with the credential portrait against a threshold; the score is evidence, not a fraud verdict.
Signed output
Optional face_match_passed in document summary; threshold and score excluded.

Method 04

Liveness

Establishes
Evaluates camera presentation with the configured method; scope varies by provider, model, device, and deployment.
Signed output
Optional liveness_score in document summary; other method evidence excluded.

Method 05

IP geography and VPN/proxy

Establishes
Returns coarse IP geography and VPN/proxy indicators for routing; they cannot identify a person or establish fraud.
Signed output
Jurisdiction audit fields when available. The per-request VPN heuristic is never persisted or included.
Government ID
Establishes
Captures and extracts a supported credential, then checks visible and machine-readable elements; it does not prove the presenter alone.
Signed output
Compact session plus available document summary.
Mobile driver's license
Establishes
Requests selected wallet attributes and validates issuer-signed presentation, subject to issuer, device, jurisdiction, and protocol.
Signed output
Compact session status and current method; wallet detail excluded.
Selfie-to-ID comparison
Establishes
Compares a live selfie with the credential portrait against a threshold; the score is evidence, not a fraud verdict.
Signed output
Optional face_match_passed in document summary; threshold and score excluded.
Liveness
Establishes
Evaluates camera presentation with the configured method; scope varies by provider, model, device, and deployment.
Signed output
Optional liveness_score in document summary; other method evidence excluded.
IP geography and VPN/proxy
Establishes
Returns coarse IP geography and VPN/proxy indicators for routing; they cannot identify a person or establish fraud.
Signed output
Jurisdiction audit fields when available. The per-request VPN heuristic is never persisted or included.

Integration

Define the challenge and verify the webhook

Create a session, verify its signed webhook, then retrieve deeper results server-side for customer policy.

The retention boundary, stated once

Verification processes sensitive identity and biometric material. The lifecycle separates those inputs from evidence used for delivery, routing, and audit.

Retention is method-specific. Confirm the path, finalization event, configured collision data, retention period, contract, and regional law.

  1. 01 · Capture

    Inputs used for the check

    ID images, mDL data, and biometric captures are processed only when the configured evidence path requires them.

  2. 02 · Complete

    Raw captures leave the path

    When a verification is finalized, document images, raw OCR, first and last name, date of birth, and document number are purged from the capture record.

  3. 03 · Evidence

    The signed outcome remains

    Method results and audit metadata support the downstream decision. Biometric collision data follows its configured retention period and is deleted when that period expires.

Retention is method-specific, not a universal promise. The customer's policy still determines the action taken from the signed evidence.

The surrounding stack

Where an identity layer sits in a fraud stack

Upstream systems request proof; Stile returns evidence; policy acts.

Where an identity layer sits in a fraud stack
CategoryProduct roleOutput
Behavior and device intelligenceInterprets device, network, and behavior.Risk signals.
Decisioning and rules enginesApplies customer scores and policies.Customer action.
Orchestration layersRoutes provider evidence through configured workflows.Orchestrated result.
Payment-fraud scoringEvaluates payment exposure using account, device, and network context.A payment risk score.
Transaction-time identity verificationTests the presenter against identity evidence.Signed identity evidence.

Use cases

Where identity fraud risk concentrates

Target proof where the wrong person creates risk.

  1. Account recovery

    Step up recovery when the expected account holder may be absent.

    Trigger: elevated recovery risk
  2. Marketplace onboarding

    Confirm seller representatives before privileged access; this does not replace KYB.

    Trigger: representative verification
  3. High-value transaction

    Challenge unusual transfers selected by customer policy.

    Trigger: customer risk policy
  4. Returning-user reverification

    Request fresh evidence after dormancy, profile/device change, or elevated permission.

    Trigger: identity freshness

FAQ

Identity fraud protection questions

Implementation, evidence, scope, and retention for technical evaluators.

What does the signed result contain?

The signed webhook contains session status, current method, jurisdiction audit, workflow, timestamps, and an optional document or NFC summary. It excludes collected data, metadata, URLs, wallet detail, and full results. Verify the Stile-Signature header over the raw body server-side. Retrieve and expand the session through an authenticated server request for deeper evidence.

Does Stile replace a fraud-scoring platform?

No. A risk or workflow layer requests proof. Stile runs the configured challenge and returns signed evidence; customer policy decides the action.

What happens when a configured check fails?

The signed event reports failed status and current method, plus a narrow document or NFC summary when available. Retrieve method detail server-side through the authenticated session/results API. Customer policy may retry, review, delay, or decline. Use check failed or additional review required; it does not label a person as fraudulent.

What does the liveness test claim cover?

For the configured AWS Rekognition Face Liveness path, iBeta identifies the listed v1.0 backend, device, and configuration. Liveness tested by an NVLAP-accredited lab to ISO/IEC 30107-3 PAD Level 1 and Level 2, with zero successful spoofs. This covers presentation attacks at the camera in those configurations, not injection, face-match accuracy, bias, or application security. Review the iBeta presentation-attack-detection confirmation letters.

What data remains after a session?

Retention separates method inputs from outcomes, audit metadata, and configured collision data. Confirm path, configuration, contract, and law.

Add identity evidence to the moments your fraud stack flags

Keep risk systems. Add a targeted challenge, inspect signed evidence, and let customer policy decide. Deeper detail stays in canonical sources.

Bring one high-risk action, evidence, and routing.

Technical evaluation

Review one high-risk action with the technical team

Map the trigger, evidence path, signed result, and customer-owned next action.