How to verify a candidate's identity during a live interview

A live video call does not verify a candidate’s identity. Hiring teams need to compare the person on camera with a trusted reference, check that a real person is present, and confirm that the same person remains in the interview. Unusual signals should lead to human review, with candidate consent and local biometric rules addressed before collection.

In KnowBe4’s 2024 account, a North Korean operative was hired after four video interviews and a background check. KnowBe4 wrote:

“Our HR team conducted four video conference based interviews on separate occasions, confirming the individual matched the photo provided on their application.”

The application photo was an AI-enhanced stock image, and the stolen identity belonged to a real person. KnowBe4’s 2024 whitepaper says malware activity on the company laptop began on July 15, 2024, at 9:55 p.m. EST. Matching a face to an application photo had not established who owned the identity.

What are you actually defending against?

Live interview identity checks must distinguish off-camera assistance, a proxy who takes the interview, and synthetic video or audio. Each calls for a different check. A candidate receiving answers off camera may be the invited person; a proxy is not. A face swap can make those cases harder to tell apart.

The US Department of Justice’s 2024 indictment describes conspirators “paying U.S. persons to attend job interviews and work meetings remotely under fake identities.” The indictment alleges at least $88 million generated over approximately six years. In a separate nationwide action, the DOJ reported in 2025 that the FBI searched 21 suspected laptop-farm premises across 14 states and seized approximately 137 laptops.

The risks extend beyond the interview. The FBI’s IC3 warned in October 2026 that North Korean IT workers increasingly use KVM-over-IP devices, including TinyPilot and PiKVM, to control company laptops remotely. Interview verification cannot replace post-hire device security.

A Gartner survey of 3,000 candidates in 2025 found that 6% admitted to posing as someone else in an interview or having someone pose as them. Gartner also predicted that one in four candidate profiles worldwide will be fake by 2028. That forecast concerns profiles, not people attending interviews.

Which free checks can you use during a live interview?

Hiring teams can make live interview identity checks more useful by asking for video, testing unscripted responses, and comparing the person across sessions. The FBI’s 2025 IC3 alert recommends unobscured backgrounds, location questions, a hand waved in front of the face, and images captured for comparison with later meetings. It also suggests in-person identity checks when possible.

Ask a specific question about the candidate’s work, then follow up on the answer. If video appears distorted, a hand wave may reveal a synthetic overlay. A recruiting lead at Make described a suspected deepfake interview with distorted facial edges and off-sync lip movements in a 2025 Fortune report. As of 2026, a hand wave is a useful prompt, not a definitive test: models improve, and a refusal can have an innocent explanation. Apply checks consistently and give candidates a chance to resolve technical problems.

What does automated identity verification add?

Automated identity verification adds repeatable face, liveness, lip-sync, and voice checks, but each measures a different risk. A Head of TA should ask which attacks a vendor has actually tested and what happens when a check is inconclusive.

1. Compare the face with a trusted ID reference. NIST’s 2025 morph-detection guidance reported best-case accuracy of 72% to 90% when a questionable image was compared with a trusted reference across morph generators. Single-image detection fell well below 40% on unfamiliar morphs. Those are morph-detection results, not a general accuracy claim for interview face matching.

2. Check 3D liveness. Three-dimensional liveness checks whether a real person is before the camera rather than a photo, screen, or mask. iBeta’s 2024 explanation of ISO/IEC 30107-3 says its Level 1 tests cap attack artefacts at $30 with 0% allowable penetration; Level 2 caps them at $300 with up to 1% allowable penetration. Those tests address objects presented to a sensor, not synthetic video injected through a virtual camera.

3. Check lip sync and audio consistency. A 2025 systematic review reported one study in which people detected face-swap deepfakes with 91.3% accuracy but lip-sync deepfakes with 52.6% accuracy. The FBI’s 2022 alert likewise described interview video where lip movement did not coordinate with speech.

4. Compare voice with earlier interactions. A 2023 UCL study found that 529 participants identified speech deepfakes correctly 73% of the time. Automated results need scrutiny too: the 2025 ASVspoof 5 evaluation reported degradation under neural codecs and compression.

Berri Proctor matches face and voice to the invited candidate at the start of and throughout a remote interview. Berri Proctor also uses 3D liveness, lip-sync checks, and scene analysis; a person reviews every integrity flag before it affects the candidate. Ask for evidence about the attack types relevant to your hiring process rather than treating any one certification as comprehensive.

How should hiring teams handle false positives?

Hiring teams should review every integrity flag before making a candidate decision because stronger detection settings can also flag genuine people. One 2025 virtual-camera detection study reported the following trade-off for its system:

The 2025 study describes one method, not expected rates for every vendor. Ask for the false-rejection rate at the same setting as the advertised detection rate. Record what evidence a reviewer sees, how a candidate can explain a flag, and who makes the final decision. Berri Proctor requires a person to review each flag before it affects a candidate.

What does the law allow you to collect?

Biometric identity verification requires a location-specific legal review before a hiring team collects face or voice data. The requirements below are a starting point for counsel, not a universal consent rule.

India. The DPDP Act, 2023 does not create a separate sensitive-data category for biometrics. Sections 5 and 6 require notice and free, specific, informed, unambiguous, withdrawable consent for each purpose. Section 7’s employment-related legitimate uses are framed around employees; their application to applicants remains debated. The 2025 DPDP Rules notification set staggered enforcement. Obtain candidate consent unless counsel establishes another basis; security-safeguard failures can attract penalties up to ₹250 crore.

Illinois. The 2024 amendment to the Biometric Information Privacy Act confirms that electronic consent can satisfy the written-release requirement and limits damages for multiple scans of one person to a single violation. Section 15(b) still requires written notice, a stated purpose and retention term, and a release before collection. Statutory damages remain $1,000 for negligent violations or $5,000 for intentional or reckless ones.

EU. Article 5(1)(f) of the EU AI Act has prohibited workplace emotion-recognition systems since February 2, 2025, subject to its narrow medical or safety exception. One-to-one biometric verification is excluded from Annex III’s remote-biometric-identification category, while recruitment and candidate-evaluation systems remain high risk. The European Council’s 2026 timeline gives December 2, 2027, for stand-alone Annex III high-risk rules and August 2, 2028, for high-risk systems embedded in regulated products. Counsel should assess the full interview workflow, not only its identity check.

What process works without buying anything?

A hiring team can start with a consistent, consent-based reference check and a documented response to discrepancies. Collect only the identity material counsel approves, explain its purpose and retention, and protect access to it.

1. Obtain the required notice and consent before collecting a candidate’s ID, selfie, or biometric reference.

2. Compare the live candidate with that trusted reference at the first interview and again in later rounds.

3. Ask an unscripted follow-up about the candidate’s work and use additional camera checks when a discrepancy needs clarification.

4. Document the evidence, allow for technical failures, and send unresolved concerns to a person for review before any hiring decision.

At higher volumes, Berri Proctor can run identity and integrity checks during the interview. Recruiters still make the final hiring decision.

What do hiring teams ask about live identity verification?

Hiring teams most often ask whether video is enough, what liveness proves, and how to respond to a flag. The answers depend on the reference used, the attack being checked, and the law governing the candidate’s data.

Is a video interview enough to verify a candidate’s identity?

A video interview alone cannot verify a candidate’s identity. KnowBe4’s 2024 account describes four video interviews in which the person matched an application photo tied to a stolen identity.

How do you tell if a candidate is using a deepfake in an interview?

The FBI’s 2025 IC3 alert recommends video with an unobscured background, location questions, and a hand waved in front of the face. Lip-sync mismatches or visual distortion merit closer review, but no single signal proves a candidate used a deepfake.

What is liveness detection, and does it stop deepfakes?

Liveness detection checks whether a real person is before the camera rather than a photo, screen, or mask. ISO/IEC 30107-3 presentation-attack testing does not, by itself, establish protection against synthetic video injected through a virtual camera.

Can employers require biometric identity verification of job candidates in India?

The DPDP Act, 2023 requires a lawful basis and appropriate notice for biometric identity verification in India. Because the employment-related exception’s application to pre-hire candidates remains debated, employers should obtain purpose-specific candidate consent unless Indian counsel advises another basis.

Does the EU AI Act ban AI interviews?

The EU AI Act does not ban AI interviews as a category. It prohibits workplace emotion-recognition systems under Article 5(1)(f), while classifying recruitment and candidate-evaluation systems as high risk.

What should a hiring team do when an integrity check flags a candidate?

A hiring team should have a person review the integrity evidence before acting on a flag. The reviewer should consider technical problems, give the candidate a way to respond, and document the final decision.