Technology

Face Match for Loan KYC: Thresholds, Liveness, Consent

|Posted by Hitul Mistry / 20 Nov 26

Last reviewed: November 2026

Face match for loan KYC (know your customer) is the automated comparison of a borrower's live selfie with the photographs on identity (ID) documents and with photos taken by field staff, producing a similarity score for each pair. It catches impersonation, swapped photos and one person applying under several names before money leaves the account.

This guide is for chief compliance officers and heads of fraud control at banks, non-banking financial companies (NBFCs) and fintech lenders. It starts with a ready comparison matrix, then covers Reserve Bank of India (RBI) KYC rules, threshold maths, a consent script and a pilot plan, all built by Digiqt's artificial intelligence (AI) development team.

How Do You Set Up Face Match Across Selfie, ID and Field Photos?

You set up face match with six image pairs, each with its own reference image, timing and fallback. Copy this comparison matrix into your KYC operating procedure; score bands are illustrative and must be tuned on your own data.

Image PairReference Image SourceWhenAction if Score Is Low
Live selfie vs Aadhaar photoAadhaar e-KYC response or Aadhaar paperless offline e-KYC fileOnboardingRetake selfie once, then manual review
Live selfie vs PAN (Permanent Account Number) card imageCard image captured in the appOnboardingManual review; PAN photos are small and old
Live selfie vs other officially valid document (OVD)Passport, driving licence or voter ID imageOnboardingAsk for a second OVD
Video frame vs OVD and PAN photosVideo-based customer identification process (V-CIP) sessionV-CIPV-CIP official decides; no auto-pass
Field officer's live photograph vs selfieDigital KYC app photo with watermarkField visitRevisit by a different officer
New selfie vs earlier applicants' facesLender's own image store (1:N search)Every applicationFraud control review

The first five pairs are 1:1 verification, asking "is this the same person?". The last is a 1:N search, asking "have we seen this face before under another name?", and it links to loan application deduplication.

Which Face Match Scores Should Trigger Which Action?

Four face match score bands should trigger four actions; this decision table is illustrative.

Score BandLiveness ResultActionOwner
HighPassAuto-pass this checkSystem
MediumPassManual review within the same dayKYC operations
LowPassReject the pair; ask for a fresh OVD or visitKYC operations
AnyFail or not runStop; treat as a possible spoofFraud control

A high score never overrides a failed liveness check.

What Is Face Match for Loan KYC?

Face match for loan KYC is one control inside customer identification, sitting beside document checks, Aadhaar authentication and Central KYC Registry integration. It answers one question only: do these images show the same living person?

The living part comes from liveness checks, which detect a presentation attack such as a printed photo, a screen replay or a mask held up to the camera. Face matching without liveness only proves that two pictures look alike.

Face match for loan KYC vs video KYC for lenders: video KYC is the full regulated process run by a trained official, while face match is the photo comparison that runs inside it and across every other image in the file. Edited document photos are a separate risk, covered in forged KYC documents in loan applications.

What Do RBI KYC Rules Say About Face Matching and Liveness?

RBI KYC rules mention face matching and liveness in four places in the RBI's NBFC Know Your Customer Directions, 2025 (RBI/DOR/2025-26/361).

  1. The definitions describe V-CIP as customer identification "with facial recognition" through a live, consent-based audio-visual interaction.
  2. Paragraph 24, the digital KYC process, says the application must capture only a live photograph, not a printed or video-graphed one, and the officer must verify that it matches the photo in the document.
  3. Paragraph 27(1)(v) says the V-CIP application shall have "face liveness / spoof detection as well as face matching technology with high degree of accuracy", while ultimate responsibility stays with the NBFC.
  4. Paragraph 27(2)(xiii) says the official must ensure the customer's photograph in the Aadhaar or OVD and the PAN or e-PAN matches the customer on the call.

The explanation under paragraph 27 adds that blinking, smiling or frowning is not mandatory for a liveness check. Paragraph 27(2)(i) says the liveness check shall not exclude persons with special needs.

How Should Lenders Set Face Match Thresholds?

Lenders should set face match thresholds by trading the false match rate (FMR) against the false non-match rate (FNMR). The National Institute of Standards and Technology (NIST) defines FNMR as the share of same-person comparisons that fall below the threshold, and FMR as the share of different-person comparisons that score above it.

Take an illustrative month of 20,000 applications in which 0.4%, or 80, are impostors.

SettingFMRFNMRImpostors Passed (80 × FMR)Genuine Borrowers Rejected (19,920 × FNMR)
Threshold A0.1%2%0.08about 398
Threshold B0.01%5%0.008996

Both settings stop nearly every random impostor, but Threshold B sends 598 more genuine borrowers to manual review. The real gap is lookalikes, family members and morphed photos, so add liveness, document checks and a 1:N search rather than only tightening the threshold.

Borrowers should give clear, specific consent before face match, because a face image is personal data. Section 6 of the Digital Personal Data Protection (DPDP) Act, 2023 requires consent that is "free, specific, informed, unconditional and unambiguous with a clear affirmative action".

Use this consent text as a starting script and adjust it with your DPDP Act compliance team:

To verify your identity for this loan application, [Lender] will take a live
photo and compare it with the photos on the identity documents you share and
on photos taken by our staff during visits. We will also check that the photo
is of a live person. We will use these images only for identity verification
and fraud prevention on this application, and keep them for [period] as our
policy and the law require. Tap "I agree" to continue. You can withdraw
consent; we will then complete KYC through [alternative route].

The Act calls the borrower the Data Principal, and section 6(4) lets her withdraw consent with the same ease with which she gave it.

Store the consent time, version and device with the application, and mask Aadhaar numbers in stored images as described in Aadhaar masking compliance for lenders.

Which Face Match Failures Do Fraud Teams See Most?

Fraud teams see five face match failure patterns most often. Each needs a different control.

PatternSignalControl
Printed photo or screen replayLiveness fails, flat textureStop; capture again with liveness
Morphed ID photoHigh match to applicant and to another person1:N search, document forensics
Lookalike family memberMedium score, same addressField visit and a second OVD
Old or poor ID photo of a genuine borrowerLow score, liveness passesManual review, not rejection
Same face, different names1:N hit with a different PAN or mobileFraud control review

Feed confirmed cases back into field investigation automation so officers know what to look for at the door.

Are impersonation cases slipping past your KYC checks?

Scope a Face Match Build With Digiqt's AI Team

Share anonymised sample images and ask us to test face match on your own KYC photos.

How Does Digiqt's AI Development Team Build Face Match for Loan KYC?

Digiqt's AI development team builds face match for loan KYC as a custom service through model selection, validation on your images, and deployment with continuous improvement. The AI software development process runs through a dedicated development team.

StepTodayWith a Digiqt Build
Compare selfie with IDOfficer eyeballs two photosSimilarity score per pair, logged
Check livenessRelies on the V-CIP officialLiveness model on every selfie, plus the official on V-CIP
Field photosStored, rarely comparedMatched against the onboarding selfie
Repeat facesNot searched1:N search across past applications
ThresholdsNot measuredFMR and FNMR tracked monthly on your own data
AuditScreenshots in the fileScores, images, consent and decisions in one record for KYC audits

Where the service runs matters for V-CIP: paragraph 27(1)(i) asks an NBFC to house the V-CIP technology infrastructure on its own premises, and to keep data ownership with itself on cloud deployments. Digiqt's team designs the hosting around that rule with your technology and compliance heads.

For document-level identity checks, see Digiqt's KYC document verification AI agent.

What Are the Next Steps for a Face Match Pilot?

A face match pilot needs five steps over about twelve weeks. Assign one owner per step.

  1. Weeks 1–2, compliance: approve the consent text, retention period and the six-pair matrix.
  2. Weeks 3–4, data: collect 2,000 consented, anonymised image pairs, including known fraud cases.
  3. Weeks 5–7, model test: measure FMR, FNMR and liveness errors on that set; pick thresholds.
  4. Weeks 8–10, shadow run: score live applications without blocking them and compare with manual decisions.
  5. Weeks 11–12, go-live: switch on the decision table, with weekly review of manual-review volumes.

After go-live, report four numbers to the risk committee every month: manual-review rate, liveness failure rate, confirmed impersonation cases caught, and complaints from genuine borrowers who were asked to retake photos. A rising complaint count usually means the threshold or the capture guidance needs work, not that fraud has gone up.

Ready to compare every selfie, ID and field photo automatically?

Start a Face Match Proof of Concept With Digiqt

Bring your current KYC procedure, or visit Digiqt to plan liveness and face matching for your onboarding flow.

Sources

Read our latest blogs and research

Featured Resources

Technology

Video KYC for Lenders: RBI-Compliant V-CIP That Converts

Video KYC for lenders: RBI's V-CIP standards mapped to build choices, a journey design that lifts completion, and fraud controls against deepfakes.

Read more
Technology

Forged KYC Documents in Loan Applications: Catch Them Early

Forged KYC documents in loan applications: how PAN, Aadhaar and address proofs are faked, a 7-step detection workflow, 8 red flags and metrics to track.

Read more
Technology

Loan Application Deduplication: Stop Wasted Credit Time

Loan application deduplication for lenders: the duplicate patterns that waste credit time, how exact and fuzzy matching work, and how to resolve matches safely.

Read more

About Us

We are a technology services company focused on enabling businesses to scale through AI-driven transformation. At the intersection of innovation, automation, and design, we help our clients rethink how technology can create real business value.

From AI-powered product development to intelligent automation and custom GenAI solutions, we bring deep technical expertise and a problem-solving mindset to every project. Whether you're a startup or an enterprise, we act as your technology partner, building scalable, future-ready solutions tailored to your industry.

Driven by curiosity and built on trust, we believe in turning complexity into clarity and ideas into impact.

Our key clients

Companies we are associated with

Life99
Edelweiss
Aura
Kotak Securities
Coverfox
Phyllo
Quantify Capital
ArtistOnGo
Unimon Energy

Our Offices

Ahmedabad

B-714, K P Epitome, near Dav International School, Makarba, Ahmedabad, Gujarat 380051

+91 99747 29554

Mumbai

C-20, G Block, WeWork, Enam Sambhav, Bandra-Kurla Complex, Mumbai, Maharashtra 400051

+91 99747 29554

Stockholm

Bäverbäcksgränd 10 12462 Bandhagen, Stockholm, Sweden.

+46 72789 9039

Lewes

16192 Coastal Highway, Lewes, Delaware 19958, USA

software developers ahmedabad
ISO 9001:2015 Certified

Call us

Career: +91 90165 81674

Sales: +91 99747 29554

Email us

Career: hr@digiqt.com

Sales: hitul@digiqt.com

© Digiqt 2026, All Rights Reserved