Home/Guides/Detect Deepfake IDs
Identity Fraud
The face matches the document. The problem is that neither one is real.
A deepfake ID pairs a synthetically generated document with a synthetic or stolen photo - built to pass automated KYC and onboarding. The fraud is in the document, not the person presenting it, which means liveness checks don't address it at all. TamperCheck provides the forensic document layer that identity verification stacks are missing.
- No document storage
- Digital, scanned or photographed
- ~1 minute per verdict
Last updated:
The problem
Why deepfake IDs slip past identity verification
Most identity verification stacks split the problem into two checks: liveness (is a real person holding up a document right now?) and document validation (does this document look like a real ID?). Deepfake IDs are built to pass both - the photo can be a real selfie or a synthetic face, and the document is generated to mimic a genuine template closely enough to clear template-matching checks.
This is a distinct problem from face deepfakes used in liveness bypass. A deepfake ID is the document itself being synthetic - the passport, driver's licence, or national ID was never issued by any government or authority. No amount of liveness checking on the applicant's face addresses this, because the fraud is in the document, not the person presenting it.
Synthetic identity fraud often combines a deepfake ID with a thin or fabricated credit file, so the identity has no prior history to flag it as suspicious. By the time inconsistencies surface downstream, onboarding has already approved the account.
What we detect
Forensic signals that expose fraud
MRZ and security-feature inconsistency
Genuine passports and IDs encode specific check-digit and field relationships in the machine-readable zone, plus security features (holograms, microprint, UV elements) that follow strict issuer specifications. Deepfake IDs frequently get these subtly wrong.
Photo-zone boundary artifacts
Composite a photo into a template and the boundary shows edge blending, lighting mismatches, resolution discontinuities. Pixel-level analysis catches the join.
Issuer template mismatch
Every issuing authority works from a small number of fixed layouts. Submitted documents get compared against known genuine templates for that issuer and type - structural drift gets flagged.
Font and micro-print inconsistency
Government-issued IDs use specific, often security-grade typefaces and micro-print that AI generation rarely reproduces exactly. Character shape, spacing, and rendering differences are detectable forensically.
Holographic and UV simulation gaps
Physical security features that exist on genuine IDs (holograms, UV-reactive ink, kinegrams) either don't render at all in a generated image or are simulated in ways that don't match real optical behaviour under standard imaging.
Cross-field plausibility
Date of birth, issue date, expiry date, apparent age in the photo - all checked for mutual consistency. Field formats validated against the claimed issuing country's conventions.
The solution
The document forensics layer that runs alongside liveness and face matching
Runs alongside your existing liveness and face-matching checks. Focused on the document - whether it's genuine, whether it was ever issued by a real authority.
How it works
- Dedicated deepfake and AI-generation detection on top of traditional ID forensics
- MRZ integrity, photo-zone boundary, and security-feature analysis built for passports, national IDs, and driver's licences
- 200+ forensic checks per document, with a clear verdict fast
- Complements liveness and face-matching - TamperCheck checks the document, your existing stack checks the person
- 190+ issuing countries supported, with issuer-specific template matching
- Zero document storage - sensitive identity documents are processed in memory and discarded immediately
Why us
What you won't get elsewhere
Findings, not a score
We name the field that changed and where it sits on the page.
Others hand you a number to justify.
One endpoint, every type
100+ document types, 190+ countries, no per-type setup.
Elsewhere, each new form is a project.
Catches what was never real
A dedicated pass for AI-generated and deepfake documents.
Template matching only catches edits.
Nothing is retained
Analysed in memory, then discarded. Never trained on.
Most services keep documents for QA.
Live in minutes
Published pricing, no minimum. Test it before you talk to us.
Platforms want a contract first.
Sits alongside your stack
Keep Onfido or Jumio for face match. We add the forensics.
Not a rip-and-replace.
FAQ
Common questions
What is a deepfake ID?
A passport, national ID, or driver's licence generated or substantially altered by AI - the document, the photo, or both - to impersonate a genuine identity document without any real government or authority having issued it.
Is a deepfake ID the same as a face deepfake used in liveness bypass?
No. A face deepfake targets the liveness check - making a fake or stolen face appear live on camera. A deepfake ID targets the document validation check - the identity document itself is synthetic. You need both face-liveness checks and document forensics.
Can deepfake IDs pass automated KYC checks?
Especially against stacks that rely on template matching and OCR. Those checks confirm a document looks like an ID and reads like one - they don't forensically verify it was produced by a real issuing authority. That's the gap TamperCheck fills.
Does TamperCheck replace liveness detection?
No - it complements it. Liveness confirms the person is physically present. TamperCheck confirms the document is genuine. A complete identity-verification stack needs both running in parallel.
Or skip the call entirely
We'll match your first top-up in credits, so you can run your own documents through it now.