Digital Identity Verification: How It Fits Into KYB and Business Verification
On May 16, 2024, the Department of Justice unsealed an indictment against Christina Chapman, an Arizona woman who had spent three years running what prosecutors called a “laptop farm” — a spare room of US-based computers logged in under stolen American identities, so that North Korean IT workers, physically in China and Russia, could clear remote-hire identity checks at 309 US companies. The scheme ran from October 2020 to October 2023 and generated more than $17.1 million. Chapman pleaded guilty in February 2025 and was sentenced on July 24, 2025, to 102 months in prison.
That should have been the end of it. It wasn’t. Between September 21 and 25, 2026, federal agents raided more than 40 laptop farms across Texas in an operation the DOJ called “Lonestar Roundup,” charging 14 people and seizing 187 devices. The Justice Department’s own headline for the action: more than 70 US companies affected. The same week, the FBI issued an advisory telling employers to “implement identity-verification processes during interviewing, onboarding, and throughout the employment of any remote worker.”
That advisory was written for HR teams hiring remote engineers. It describes, almost word for word, the gap sitting underneath every KYB (Know Your Business) check too. A KYB check confirms that a company is real and that a named person is recorded as its director, signatory or beneficial owner. It does nothing to confirm that the human filling in the onboarding form is that person. A filing is not a face.
Confirming the human behind an onboarding claim is one half of the problem. Confirming that the claim itself — “I am the director,” “I hold 30% of this company” — matches a real filing is the other half. Zavia connects directly to government ownership registries in 173 countries to resolve a claimed directorship or stake against the actual record, instead of taking the onboarding form’s word for it.
01Digital Identity Verification: What It Actually Checks
Digital identity verification (IDV) is the set of technical checks that confirm a physical person is who their credentials say they are, without a human reviewer meeting them in person. Three distinct layers do the work, and they catch different things.
Document verification
A passport or national ID is read two ways. Optical character recognition parses the Machine Readable Zone (MRZ) — the fixed-format line at the bottom of the document — against its built-in cryptographic check digit. Where the document carries an embedded chip, NFC reading pulls the chip’s digitally signed data and checks it against the issuing country’s public key, the same mechanism airport e-gates use. A forged physical document can mimic a printed MRZ. It cannot forge a chip signature it was never issued.
Biometric liveness
A selfie compared against a passport photo answers one question: do these two faces match. It does not answer whether a living person produced that face in real time, which is why liveness detection exists as a separate check. NIST’s glossary defines liveness detection as a subset of presentation attack detection (PAD) — measuring involuntary physical responses that a photo, mask, or screen replay cannot reproduce. NIST’s SP 800-63-4, finalized July 31, 2025 in its first full rewrite since 2017, states the requirement plainly in companion volume 800-63A-4: the credential service provider “shall perform a Liveness check when collecting biometric data.”
Database and registry cross-checks
The third layer checks the document against an external source of truth — the issuing government’s database, a sanctions or PEP list, a corporate registry. Here identity verification stops being about the person alone and starts touching KYB directly: a verified human claiming to be “the director of Acme Ltd” is only useful once that claim is checked against Acme’s actual filing.
| Check | What it catches | What it misses | Typical use |
|---|---|---|---|
| Document verification | Forged, altered, or expired physical documents; mismatched MRZ checksums; invalid chip signatures | A genuine document held by someone other than its owner | First-mile onboarding, account opening |
| Biometric liveness | Photos, masks, screen replays, and — per NIST 800-63-4 — AI-generated presentation attacks | A real person coerced or paid to front for someone else | Binding a document to the live person presenting it |
| Database / registry cross-check | Identities that don’t exist on an issuing register; a role claim with no matching filing | Stale or unfiled registry data; jurisdictions with no searchable register | Confirming a claimed company role is real |
Source: NIST SP 800-63A-4, Digital Identity Guidelines — Identity Assurance Level Requirements (final, July 31, 2025); FATF Guidance on Digital Identity (March 2020).
02Where Person-Level IDV Sits Inside a KYB Workflow
KYB and IDV answer two different questions. KYB asks: is this company real, and who does its paperwork say owns and runs it. IDV asks: is the human in front of me right now that person. Running KYB without IDV lets anyone assert a role. Running IDV without KYB verifies a real person asserting a role no filing supports. Neither check, alone, closes the loop.
A directorship or UBO claim only becomes evidence once a verified human is bound to it and that binding is checked against a filing — the sequence FATF’s March 2020 Guidance on Digital Identity treats as a precondition for relying on digital ID in customer due diligence.
The ultimate beneficial owner (UBO) a company discloses on paper and the person who logs into the onboarding portal claiming that identity are not automatically the same fact twice confirmed. They are two separate facts that happen to point at each other, and IDV is the mechanism that checks whether the pointer actually resolves.
03The Regulatory Anchors: FATF, NIST, eIDAS 2.0
Three bodies, writing independently, converge on the same structure: verify the document, verify the live person, verify the claim against a source of truth.
FATF published its Guidance on Digital Identity in March 2020, a technology-neutral framework telling governments and regulated entities when a digital ID system is reliable enough to satisfy the customer due diligence obligations in FATF Recommendation 10. It is risk-based, not pass/fail: a low-risk relationship can rely on a lighter-assurance digital ID; a higher-risk one — a new corporate signatory with no transaction history, for instance — cannot.
NIST sets the technical floor underneath that risk judgment. Its Identity Assurance Levels, defined in SP 800-63A-4, grade how much proofing a given relationship needs.
Source: NIST SP 800-63A-4, Digital Identity Guidelines — Identity Assurance Level Requirements (final, July 31, 2025). Most remote KYB onboarding operates at IAL2; a high-risk UBO claim is the case for stepping up to IAL3.
eIDAS 2.0 — Regulation (EU) 2024/1183, which the European Parliament and Council adopted on April 11, 2024 and which entered into force on May 20, 2024 — takes a different approach again: instead of grading assurance after the fact, it builds a government-issued credential upfront. The European Digital Identity Wallet (EUDI Wallet) requires each EU member state to issue, free of charge, a wallet holding verified identity attributes that a business can check at the point of onboarding, rather than re-deriving assurance from a passport scan each time. Member states are still finalizing the technical rollout; the regulation itself has been in force since May 2024.
04Failure Modes: Synthetic Identity and Deepfake Liveness Bypass
The Federal Reserve Bank of Boston estimated, in a Synthetic Identity Fraud Mitigation Toolkit released in February 2022, that US financial institutions lost roughly $20 billion in 2020 alone to synthetic identity fraud — identities built from a blend of real and fabricated data rather than a single stolen identity. A synthetic identity has no live victim to report it missing, which is precisely why it survives longer inside an onboarding pipeline than a stolen one does.
Deepfakes attack the newer layer: liveness itself. Sumsub’s Identity Fraud Report, published in November 2025, found that AI-driven, multi-step identity fraud rose from 10% of all identity fraud cases in 2024 to 28% in 2025.
Source: Sumsub Identity Fraud Report, published November 2025.
A liveness check works because a live face responds to light and challenge in ways a static image cannot. A laptop running under someone else’s name has no such tell to fail — which is exactly why the Chapman and Lonestar Roundup cases were caught by payroll and banking anomalies, not by any identity check at all. No IDV layer was in the loop to catch.
That last point is the one worth sitting with. Document checks, liveness, and registry cross-checks all defend the moment a human proves who they are. None of them defend a process that never asked the question — which is what both North Korean IT-worker schemes actually exploited: not a broken liveness check, but companies that skipped identity verification for a remote hire entirely.
A verified human claiming a directorship is still just a claim until it is checked against the company’s own record. Resolving a claimed UBO or director role against the government registry that actually filed it is the step that turns a verified person into a verified fact about a company.
05Binding a Verified Identity to a Claimed Company Role
A KYB onboarding flow that wants to close the gap runs identity verification as a gate, not a formality, before it accepts any claimed role. In practice, that means five steps, run in order:
- Capture and verify the document. Read the MRZ and, where present, the NFC chip; reject the session if the chip signature does not validate against the issuing country’s published key.
- Run a liveness challenge, not just a selfie match. Per NIST SP 800-63A-4, the system performs a liveness check at biometric collection, not only a face-match comparison, so a presentation attack is caught at the point of capture.
- Record the assurance level achieved, not just a pass/fail flag. A low-risk account opening and a new corporate signatory claiming 25% ownership do not need the same IAL, and the record should show which one was actually run.
- Cross-check the claimed role against the entity’s own filing. The verified name goes against the company’s register of directors or shareholders, the same filing a standard or enhanced due diligence review would pull independently, before the claim is accepted as fact.
- Re-verify on a trigger event, not on a fixed renewal date. A change of director, a change of registered address, or a new large shareholding are each a reason to run the identity check again — the claim changed, so the proof backing it needs to change too.
06What This Means for a Compliance Stack
Identity verification and KYB are not substitutes, and a stack that treats them as interchangeable has a hole in it, even if every individual check passes. The 309 companies in the Chapman case and the 70-plus in Lonestar Roundup did not fail on weak document checks. Most never ran an identity check on the claimed worker at all — a gap no KYB review of the hiring company itself would catch, because KYB checks the entity, not the hand that signs for it.
The two checks belong in sequence: verify the human, then verify what that human claims about the company, then keep checking both. A one-time check, run once at onboarding and never run again, is a snapshot of a relationship that keeps moving after the snapshot is taken. An identity genuine in 2024 says nothing about who controls the login in 2026.
An onboarding check that runs once and never re-runs leaves exactly the gap a trigger-based re-check closes. Pairing identity re-verification with ownership monitoring that flags a registry change after onboarding closes means a director swap or a new majority shareholder gets caught on the entity side even if nobody logs back in to re-prove who they are.
Frequently Asked Questions
What is digital identity verification (IDV)?
Digital identity verification is the set of remote technical checks — document authentication, biometric liveness, and database or registry cross-checks — used to confirm that a physical person is who their credentials claim, without an in-person reviewer. NIST codifies the requirements in Special Publication 800-63-4 (final, July 31, 2025); FATF sets the risk-based framework for using it in customer due diligence in its March 2020 Guidance on Digital Identity.
How is digital identity verification different from KYB?
KYB (Know Your Business) verifies that a company exists and checks its registered directors, shareholders, and beneficial owners against filed records. IDV verifies that the specific human in an onboarding session is the person they claim to be. KYB checks the entity; IDV checks the human presenting a claim about that entity. Neither replaces the other.
Why does KYB need person-level identity verification at all?
Because a KYB check can only confirm what a filing says — it cannot confirm that the person submitting a director or UBO claim during onboarding is actually that filed person. The Christina Chapman case (DOJ, sentenced July 24, 2025) and the DOJ’s September 2026 “Lonestar Roundup” action, affecting more than 70 US companies, both involved onboarding flows that never ran an identity check on the human behind the claim.
What is biometric liveness detection, and why isn’t a selfie match enough?
A selfie match only confirms two images are similar; it does not confirm a living person produced the image in real time. Liveness detection, defined under NIST’s presentation attack detection (PAD) framework, measures involuntary physical responses — micro-movement, light reflection, depth — that a photo, mask, or deepfake video cannot reliably reproduce. NIST SP 800-63A-4 requires a liveness check at the point of biometric collection, not as an optional add-on.
What is eIDAS 2.0 and how does it change business onboarding in the EU?
eIDAS 2.0 is Regulation (EU) 2024/1183, in force since May 20, 2024, which establishes the European Digital Identity Wallet (EUDI Wallet). Each EU member state must issue a free wallet holding verified identity attributes that a business can check directly at onboarding, shifting part of the proofing burden from each individual company to a government-issued credential.
What is synthetic identity fraud, and how is it different from stolen-identity fraud?
A stolen identity uses one real person’s full details without their knowledge. A synthetic identity blends real and fabricated data — a real Social Security number with a fabricated name and date of birth, for example — so there is no single live victim to report it missing. The Federal Reserve Bank of Boston estimated 2020 US losses from synthetic identity fraud at $20 billion, in a toolkit released February 2022.
Can a deepfake defeat a biometric liveness check?
Deepfakes are specifically why liveness requirements exist and keep being updated: Sumsub’s November 2025 Identity Fraud Report found AI-driven, multi-step identity fraud rose from 10% of cases in 2024 to 28% in 2025. NIST’s SP 800-63-4 revision, finalized July 31, 2025, was written in part to address presentation attacks that earlier liveness implementations did not anticipate.
Does verifying a director’s identity also verify their ownership percentage?
No. Identity verification confirms who the person is. It says nothing about what they own or control — that fact has to come from the company’s own filing, not from the person’s own assertion during onboarding. A verified human stating “I own 30%” still needs that figure checked against the actual beneficial ownership record before it is treated as fact.