Fake ID Detection: How AI Detects Forged and Tampered Documents
Fake ID detection uses AI to identify forged, tampered, or fraudulently captured identity documents before they pass verification. It analyzes the document capture itself, looking for signals such as screen glare, moire patterns, edge distortions, and pixel inconsistencies rather than relying only on the data printed on the ID. This helps detect cases such as a genuine document photographed from a screen or a photocopy presented as an original.
A genuine document does not always mean a legitimate identity. Fraudsters can use a real ID that does not belong to them, or present a genuine document through a screen, photocopy, or altered image. Detecting these attacks requires analyzing how the document was captured, not just the information it contains.
What Is Fake ID Detection?
Fake ID detection is a fraud check that determines whether an identity document submitted during onboarding is authentic and captured directly, or whether it has been forged, altered, or re-photographed. It goes beyond reading the text on an ID by examining the image for signs of tampering or indirect capture.
The distinction matters because modern ID fraud does not always involve a completely fabricated document. A fraudster may photograph a genuine ID from a phone screen, submit a color photocopy, or replace the photo on an otherwise valid document. These documents can contain accurate, readable information. Fake ID detection addresses a different question: is the document genuine and was it captured directly?
Types of ID Forgery and How Detection Works
Strong fake ID detection separates different attack types instead of returning a single generic fraud flag. Verihubs evaluates tampering, screen capture, and photocopy as independent checks, each with its own pass or fail signal. This gives your team more context about why a document was flagged.
| Forgery type | What the fraudster does | What detection looks for |
|---|---|---|
| Screen recapture | Photographs an ID displayed on a phone or monitor | Screen glare, moire patterns, pixel density |
| Photocopy | Submits a color or black-and-white copy as the original | Flatness, print texture, missing depth cues |
| Photo substitution | Replaces the face on an otherwise valid ID | Edge distortions, inconsistent lighting, tampering marks |
| Data or field tampering | Alters text such as name, date of birth, or document number | Font and alignment anomalies, pixel-level edits |
Several of these attacks can start with a genuine document. Screen recapture and photocopy, for example, do not require the fraudster to create a fake ID. They simply present a real document indirectly. Because the underlying data remains valid, data-only checks may still approve the submission. Detecting these attacks requires analyzing the document capture itself.
Why Traditional OCR Misses Forgeries
Optical character recognition is designed to read document data, not determine whether the document itself is authentic. It extracts text and checks whether the expected fields are present and readable. If a screen recapture or photocopy contains clear, readable information, OCR can still extract the fields successfully.
That creates a blind spot for document fraud. OCR cannot reliably distinguish a photograph of a genuine ID from the physical document itself when both contain the same readable information. It does not evaluate screen glare, surface characteristics, or whether part of the document has been digitally altered.
This is what makes fake ID detection different from OCR. Rather than asking only what the document says, it also evaluates whether the document appears genuine and was captured directly. On the identity side of the check, this can be paired with liveness detection, which confirms that the person is live, while fake ID detection evaluates the document capture.
How AI Detects Fake IDs From Document Captures
Fake ID detection analyzes the submitted image for signals that can indicate an indirect or altered capture. When a user submits an ID, the system analyzes the capture across three checks and returns a result in under one second per check.
Capture-integrity analysis. The AI analyzes signals such as screen glare, moire patterns, pixel density inconsistencies, and edge distortions. These characteristics can indicate that a document was photographed from a screen or reproduced as a printed copy. A direct capture of a physical card has different visual characteristics from a photograph of a display, allowing the model to distinguish between the two.
Tamper analysis. The system examines the document for signs of digital or physical alteration, including mismatched fonts, misaligned fields, inconsistent lighting around the photo, and pixel-level edits that may indicate a substituted face or modified data.
Independent scoring. Instead of returning one generic verdict, the system provides a separate pass or fail result for tampering, screen capture, and photocopy detection. This gives your team a more specific reason for the result and makes it easier to approve, reject, or route a submission for manual review.
US Document Coverage and Where Forgery Risks Appear
For US onboarding, fake ID detection needs to cover the documents that legitimate users actually carry as well as those commonly targeted by fraud. Verihubs supports driver’s licenses and state IDs across more than 30 states, US passports, and passport cards, with coverage updated as state document formats change.
Forgery risks are particularly relevant across several industries. Digital banks can encounter synthetic identities using screen-captured IDs during remote onboarding. Online lenders may see forged documents used to apply for credit under fabricated identities. Crypto exchanges need to meet KYC and AML requirements while detecting forged passports and other documents. Insurance and e-wallet platforms also need to replace manual document checks with faster automated detection without adding unnecessary friction for legitimate customers.
Fake ID Detection in the KYC Flow
Fake ID detection does not replace your KYC provider. It adds a forgery check to the existing KYC flow and can run immediately before or after document verification.
A common implementation is to analyze the document when it is submitted, then combine the result with the existing identity decision. A document can contain valid identity information and still fail the screen-recapture check. Pairing document analysis with device intelligence adds another layer of context. A suspicious document submitted from a high-risk device can generate signals from both the document and the environment, giving your risk team a clearer basis for review.
Verihubs ID Forgery Detection is designed to work this way. It returns a clear pass or fail result for each check in under a second, works alongside third-party KYC providers or in-house verification flows, and can be paired with Device Intelligence and NFC passport verification. It holds ISO/IEC 27001:2022, and is backed by Y Combinator with Meta as a business partner.
Frequently Asked Questions
Does fake ID detection replace my existing KYC provider?
No. It runs alongside your existing KYC flow, whether you use a third-party provider or an in-house solution, as an additional forgery check. It can run before or after document verification, with the result combined with your existing identity decision.
Can it tell a photocopy apart from a screen capture?
Yes. Each is evaluated as a separate check with its own pass or fail signal. Tampering, screen capture, and photocopy detection are assessed independently, so your team can identify the type of forgery rather than receiving a single generic flag.
What US documents does fake ID detection support?
Verihubs supports driver’s licenses and state IDs across more than 30 states, US passports, and passport cards. Coverage is updated as document formats change at the state level.
How fast is a fake ID detection check?
The check returns a result in under one second per check in most cases. This allows it to run as part of an inline onboarding flow without adding significant processing time for legitimate users.
Can fake ID detection catch a real ID that was stolen?
Fake ID detection can identify suspicious presentations of a genuine ID, such as a screen capture or photocopy. It cannot determine from the document alone whether a physical ID has been stolen. For a genuine stolen ID presented directly, additional signals such as identity matching, liveness, and device intelligence are needed to assess the person and the surrounding environment.
Does fake ID detection work with deepfakes?
Fake ID detection focuses on the document. Deepfakes target the face, so the complementary defense is liveness detection, which helps confirm that the person is live. Together, these checks address both the document and biometric sides of an onboarding attack.
Fake ID Detection Looks Beyond the Document Data
Fake ID detection adds a layer that standard OCR does not provide. Instead of looking only at the information on a document, it analyzes whether the document appears genuine and was captured directly.
This matters for US digital banks, lenders, crypto exchanges, and wallets that need to identify forged or indirectly captured documents before they become part of an approved identity. Combining document forgery detection with device intelligence and liveness provides additional signals across the document, person, and device.
Want to see how fake ID detection identifies screen captures, photocopies, and tampered documents? Get a 20-minute Verihubs demo of ID forgery detection for US documents.
