Driver’s licenses · State IDs · KYC · Onboarding

How to detect a fake ID or driver’s license.

Most fake IDs presented today aren’t physical forgeries — they’re a photo or scan of a genuine ID with the photo, a date, or a name changed before it was uploaded. The card was never touched; the image was. That edit leaves a trace in the pixels and the file, even when it’s invisible to the eye.

§01 / WHAT GETS FAKED
The ways they are made

What a faked ID image actually looks like.

ID fraud in an upload flow clusters into a small number of image-level techniques. Each leaves a different kind of trace in the pixels or the file.

§02 / HOW TO CHECK
Do this first

Five checks, in the order that catches the most.

Run these on the image itself before treating an ID as proof of identity or age. None of them require the physical card.

1. Run error level analysis on the photo and text

An edited region — a swapped photo, a retouched date — is often saved at a different JPEG quality than the rest of the image, even after the whole file is re-saved once more. Error level analysis (ELA) surfaces that mismatch as a visibly brighter patch.

2. Look for copy-move duplication

A pasted photo, a cloned hologram, or a repeated background texture leaves matching regions elsewhere in the image. Docurensic’s copy-move detector clusters matching patches and draws a box around both the source and the destination.

3. Check the noise floor

A genuine photo of a physical card carries a consistent sensor noise pattern across the whole frame. A patch that’s noticeably smoother than its surroundings — a retouched date, a swapped photo — was edited or re-rendered after the fact.

4. Check internal consistency

Issue date before expiry, a birth date consistent with the stated age, and an ID number in the format the issuing authority actually uses. A field that breaks that pattern calls the rest of the document into question, even in an otherwise convincing forgery.

5. Read the file’s metadata and compression history

An ID photographed on a phone carries EXIF camera data and a single JPEG generation. A screenshot, a re-saved crop, or a missing EXIF block where a live photo is claimed is inconsistent with how the image says it was captured.

§03 / IN PRACTICE
Illustrative example

What a retyped birth date looks like.

A constructed example — not a real customer file — showing the kind of finding Docurensic surfaces.

Exhibit · illustrative

State driver’s license · scan upload

NameJ. Okafor Reyes
ClassC · Standard
Issued03/2023
Date of birth04/12/2003
Structure1 image · JPEG Q92, re-saved once
Review recommendedForensic axis · image altered
Date of birth field re-typed

The DOB digits sit on a different noise floor from the rest of the card and carry their own compression generation. The recovered patch underneath is consistent with a birth year four years later.

Forensics · noise + ELA · strong
Applicant would be under the minimum age

At the recovered birth year, the applicant is under the role's stated age requirement.

Validation · consistency · indicative
§04 / WHY EYEBALLING FAILS

Why looking at it never works.

A fake ID built this way is not a bad forgery. It’s a real government template with one field changed, so everything a reviewer checks by eye — the seal, the layout, the security pattern printed on the card — is genuine, because none of it was touched on the physical original. The edit happened to the photo of it, after the fact, and photos don’t carry the physical security features a real card has.

The evidence that catches it lives in the pixels and the file: compression history, noise consistency, and whether the fields agree with each other. Docurensic runs error level analysis, copy-move detection and a noise-floor check on every image, and reports exactly which region doesn’t match the rest of the frame.

Five layers, one verdict. Every document runs through security, forensics, reasoning, validation, and external lookups — see how the engine works.

§05 / FAQ
Questions

Frequently asked.

Can you tell if an ID photo was edited?

Often, yes. An edited region — a swapped photo, a retouched date — usually carries a different compression history or a smoother noise floor than the rest of the image, which error level analysis and the noise-floor check are built to surface. A pasted or cloned element also shows up in copy-move detection as a matching pair of regions.

What if it's a photocopy or scan, not a phone photo?

The same pixel-level checks still apply — a scanned or photocopied ID has its own consistent noise and compression signature, and an edited region still stands out against it. What changes is the metadata: a scanner leaves a different trace than a phone camera, and Docurensic reports which one the file is actually consistent with.

Do you verify identity against a government database?

No. Docurensic is a forensic analysis of the image you were given — it establishes whether that image was altered, not whether the person or the record exists. It pairs well with a dedicated identity-verification or liveness check, which answer the other half of the question.

How fast is a check?

Seconds for a typical ID image. You upload the photo or scan and get a plain-English verdict with each finding located on the region it appears in, so a reviewer can see the evidence rather than take the score on trust.

Other document types

Bank statements · Pay stubs & payslips · Invoices · Utility bills · W-2s · Lease agreements · Diplomas & certificates · All document fraud detection

Check an ID in the next minute.

Free to start. No install, no sales call.