Five dimensions of customer intelligence extracted from a single photograph.

An API that processes the KYC selfie captured during onboarding and extracts intelligence from it across five dimensions: face verification, deduplication, background context, image metadata, and capture circumstances. No additional integration required.

JupiterNiyoPunjab & Sind BankJana Small Finance BankCSB BankLenDenClubmoneyview
SnapmintIndiaMARTBajaj FinanceKisshtOneCardSmartCoinOTO

Why it exists

Selfie verification linked to identity, location and device signals

The industry treats the KYC selfie as a compliance check: does the face match the document, and is the person live. That check is necessary. It is also the minimum. A selfie carries information about the environment it was taken in, the device it was taken on, how it was physically captured, and whether the same face has appeared before under a different identity. Image Intelligence reads all of it.

What It Does

One selfie in. Five layers of intelligence out.

Face matched on a verification kiosk with confidence score
  • Document match97%
  • 3D livenessVerified
  • Spoof detectionClear
  • Capture sourceLive camera

Face and Identity Match

The compliance baseline. Every institution requires it. Image Intelligence delivers it as the first of five layers, not the only one.

  • Face matched against the identity document with confidence scoring
  • 3D-depth liveness verification
  • Spoof and presentation attack detection
Face vault showing duplicate match history
  • Vault match9 identities
  • Linked cluster3 accounts
  • Match historyAvailable
  • Prior namesSurfaced

Deduplication (Face Vault)

A single-session liveness check confirms the face is real in this moment. It cannot confirm whether the same face has appeared before, under a different name, in a different application. Face Vault can.

  • Every incoming selfie matched against the full historical database
  • Same face, different name, different application surfaced with match history
  • Cluster identification: groups of linked identities using the same face
  • 1,400+ duplicate identity clusters identified across 19,986 KYC selfies in production
Background intelligence panel with risk score
  • EnvironmentResidential
  • Setting typePrivate
  • Location patternUnique
  • Kiosk indicatorsNone

Background Intelligence

The background of a selfie is a signal most onboarding flows discard entirely. A selfie taken in a residential setting carries a different risk and affluence profile from one taken in a lending kiosk, a cyber cafe, or a commercial establishment.

  • Background classified across multiple environment categories
  • Residential, commercial, institutional, outdoor, and kiosk settings distinguished
  • Environment type scored as a risk and affluence indicator
  • Pattern detection: multiple selfies from the same commercial location flagged
Image metadata and EXIF summary on a monitor
  • Device makeExtracted
  • EXIF dataComplete
  • ResolutionNative
  • Device consistencyMatch

Image Metadata

Every image file carries technical information about the device that created it and the conditions under which it was captured. This cluster extracts and analyses that metadata as an independent signal layer.

  • Device make, model, and camera specifications extracted
  • EXIF data analysis for consistency and completeness
  • Image resolution, compression, and format attributes
  • Device-to-application consistency: was this image created by the device submitting the application
Phone camera capturing a live selfie
  • Camera distanceArm’s length
  • Self-captureYes
  • Handling postureOne-handed
  • Camera usedFront

Capture Context

The physical circumstances of how the selfie was taken carry a signal that no other layer reads. A self-taken photograph at arm's length on a personal device looks measurably different from one taken by another person at a different distance and angle.

  • Camera distance, angle, and orientation analysis
  • Self-capture versus externally-operated indicators
  • Device-handling posture during the photograph
  • Front-camera versus rear-camera identification

What Comes Out

Per-image intelligence profile. Real-time. Three consumption tiers.

Where It Applies

Three deployments, each reading the KYC image for a different decision.

  • Fraud prevention

    Fraud prevention

    Identity reuse, spoofing, externally-operated captures

  • Onboarding

    Onboarding

    KYC image verification, Face Vault screening, environment assessment

  • Credit underwriting

    Credit underwriting

    Scene and device metadata as affluence and lifestyle indicators

Integration

A single REST call on the selfie you already capture. No additional integration, no change to the onboarding flow, no replacement of systems already in production.

JupiterNiyoPunjab & Sind BankJana Small Finance BankCSB BankLenDenClubmoneyview
SnapmintIndiaMARTBajaj FinanceKisshtOneCardSmartCoinOTO

Why it exists

The industry treats the KYC selfie as a compliance check: does the face match the document, and is the person live. That check is necessary. It is also the minimum. A selfie carries information about the environment it was taken in, the device it was taken on, how it was physically captured, and whether the same face has appeared before under a different identity. Image Intelligence reads all of it.

Selfie verification linked to identity, location and device signals

What It Does

One selfie in. Five layers of intelligence out.

Face and Identity Match

The compliance baseline. Every institution requires it. Image Intelligence delivers it as the first of five layers, not the only one.

  • Face matched against the identity document with confidence scoring
  • 3D-depth liveness verification
  • Spoof and presentation attack detection

Deduplication (Face Vault)

A single-session liveness check confirms the face is real in this moment. It cannot confirm whether the same face has appeared before, under a different name, in a different application. Face Vault can.

  • Every incoming selfie matched against the full historical database
  • Same face, different name, different application surfaced with match history
  • Cluster identification: groups of linked identities using the same face
  • 1,400+ duplicate identity clusters identified across 19,986 KYC selfies in production

Background Intelligence

The background of a selfie is a signal most onboarding flows discard entirely. A selfie taken in a residential setting carries a different risk and affluence profile from one taken in a lending kiosk, a cyber cafe, or a commercial establishment.

  • Background classified across multiple environment categories
  • Residential, commercial, institutional, outdoor, and kiosk settings distinguished
  • Environment type scored as a risk and affluence indicator
  • Pattern detection: multiple selfies from the same commercial location flagged

Image Metadata

Every image file carries technical information about the device that created it and the conditions under which it was captured. This cluster extracts and analyses that metadata as an independent signal layer.

  • Device make, model, and camera specifications extracted
  • EXIF data analysis for consistency and completeness
  • Image resolution, compression, and format attributes
  • Device-to-application consistency: was this image created by the device submitting the application

Capture Context

The physical circumstances of how the selfie was taken carry a signal that no other layer reads. A self-taken photograph at arm's length on a personal device looks measurably different from one taken by another person at a different distance and angle.

  • Camera distance, angle, and orientation analysis
  • Self-capture versus externally-operated indicators
  • Device-handling posture during the photograph
  • Front-camera versus rear-camera identification

What Comes Out

Per-image intelligence profile. Real-time. Three consumption tiers.

A device emitting individual image signals as structured data

Raw Signals

The full set of image-level attributes as structured data. For institutions that integrate directly into their own decisioning logic.

  • Face match confidence and liveness score
  • Face Vault match status and cluster ID
  • Scene classification and environment type
  • Image metadata and device extraction
  • Capture context indicators
A scorecard reading the image signals into a risk score

Pre-Built Scores

Decision-ready output, each score accompanied by contributing factors.

  • Image risk score (composite across all five layers)
  • Identity integrity classification
  • Affluence and lifestyle indicators from scene and device metadata
image signals feeding a configurable risk model

Custom Risk Models

Image scoring calibrated to the institution's own onboarding patterns. The weight of a kiosk-environment detection differs between a bank with no agent channel and an NBFC with 60% agent-assisted origination. Sign3 tunes accordingly.

Where It Applies

Three deployments, each reading the KYC image for a different decision.

Fraud prevention

Fraud prevention

Identity reuse, spoofing, externally-operated captures

Onboarding

Onboarding

KYC image verification, Face Vault screening, environment assessment

Credit underwriting

Credit underwriting

Scene and device metadata as affluence and lifestyle indicators

Integration

A single REST call on the selfie you already capture. No additional integration, no change to the onboarding flow, no replacement of systems already in production.

SDK connecting a mobile app and a code editor

Delivery

REST API. No additional integration required.

Input

KYC selfie image.

Response time

Real-time.

Timeline

1–2 weeks.

Compliance

DPDP Act compliant. GDPR compliant.

Run your last week of traffic through Sign3.

We'll score it, surface the fraud patterns your current stack missed, and walk you through what we found. Your data, our scoring. No commitment, no integration, no decision required until you've seen the result.