Most remote identity checks still start the same way. Photograph the passport's data page, run OCR, compare the printed photo with a selfie. It is familiar and works on any phone with a camera, and it has a ceiling that fraud teams keep hitting.
Nearly every passport issued today also carries an NFC chip. That chip changes what is possible, because it holds data the issuing state has signed.
What OCR actually checks
OCR turns the image of the data page into text: name, nationality, date of birth, document number, expiry date. Most systems also read the machine-readable zone (MRZ), the two lines of characters at the bottom of the page, and validate its check digits.
Check digits catch typos and misreads. They do not catch forgery. The algorithm is public, defined in ICAO Doc 9303, so anyone who edits a document number or a date of birth can recompute matching digits in seconds.
Beyond the text, image-based systems look for signs of tampering: inconsistent fonts, misaligned fields, a pasted portrait, compression artefacts, screen moiré. These are useful signals. They are also judgements about pixels, made on an image supplied by the person being verified.
Where OCR-only verification breaks
- Edited images. A real data page with a changed name or date, or a swapped portrait, can be rendered cleanly enough to pass template and font checks.
- Generated documents. Generative tools now produce data pages with correct layouts, plausible security patterns and valid MRZ check digits.
- Injected captures. If the image never came from a camera, there is no physical document to inspect. Every tampering check is inspecting a file.
- Thresholds. Image forensics returns a score. Set it strict and genuine customers with worn passports or poor lighting fail. Set it loose and forgeries pass.
None of this makes OCR useless. It means OCR answers the question "what does this document say?" and not "did a state issue this, unchanged?"
What the chip proves
An ePassport chip holds the holder's data, the same fields as the MRZ, and a high-resolution facial image. These sit in separate data groups. A hash of each group is listed in a security object, which the issuing state signs with a Document Signer certificate. That certificate is in turn signed by the state's Country Signing Certificate Authority (CSCA), the root of that country's passport trust.
Verifying this chain is called passive authentication. When it succeeds, you know three things:
- the data was issued by that state, because only the state holds its signing keys;
- the data has not been altered since, because any change breaks the hash;
- the facial image is the one the state enrolled, not one supplied by the applicant.
That is a different class of evidence from OCR. It is not a probability and there is no threshold to tune. The signature verifies or it does not.
OCR tells you what a document says. The chip tells you who said it.
The limits worth knowing
Chip verification is strong, and it has edges a compliance reader should understand before relying on it.
- Genuine data is not the same as the original chip. Passive authentication proves the data is genuine. Proving the physical chip is not a copy relies on separate mechanisms, Active Authentication or Chip Authentication, which many but not all passports support. Matching the chip's signed face against a live selfie closes most of that gap, because a copied chip still carries someone else's face.
- Trust depends on the roots. A verifier is only as good as its CSCA store. A sound system refuses a chip whose issuer it cannot verify, rather than passing it with a warning.
- Not every document has a readable chip. Many national ID cards have none, and some chips are locked by the issuing government. The Israeli ID card is one example: its chip cannot be read by third parties, so it can only be verified from the photo.
- The user needs an NFC-capable phone and has to hold the passport against it for a few seconds. With clear guidance, this is a short step.
Where OCR still belongs
OCR is still part of a chip-based flow. To open the chip, the reader needs an access key derived from the document number, date of birth and expiry date, the same fields printed in the MRZ. Reading the MRZ with OCR is the fastest way to get them.
OCR also remains the route for documents that have no chip. The difference is its role. In a chip-first flow, OCR collects the key. The chip supplies the evidence.
A rule of thumb for fraud and compliance teams
- If the document has a chip and the user has an NFC phone, verify the chip. Treat OCR-only on a chipped passport as a weaker level of assurance.
- Use OCR and image checks where there is no chip, and set expectations for that level of assurance accordingly.
- Match the live selfie against the face signed in the chip, not the printed photo.
- Ask vendors whether chip verification runs against a full CSCA store, what happens with an unknown issuer, and whether verification depends on a live network call.
How JERIX approaches it
JERIX reads the NFC chip on passports and verifies the signature chain to the issuing state's CSCA root, fully offline: no network call, no CRL, no OCSP. The trust store holds 588 certificates from 112 countries, and a chip whose issuer is not in it is refused.
The face is taken from the chip's signed data on the server and matched against the live selfie. On iOS, the selfie capture is also signed on the device with a single-use nonce. Where a document has no readable chip, such as the Israeli ID card, verification is photo-based.
After the check, no images, biometrics or chip data are retained. The platform receives a verified result, not the document.
If you are comparing chip verification across vendors, we can show you a live passport read in a short demo.