Which ID Numbers Does ONYRI Sanitize Detect, by Country?

ONYRI Sanitize checks a document against country-specific patterns for ID numbers, tax numbers, and social security numbers before it masks anything. Some of these checks only look at the shape of the number. Others run the same checksum formula the issuing government or bank uses. This page lists exactly what the engine looks for in each of the 47 countries it covers, and where you still need to check the result yourself.

How does ONYRI Sanitize check an ID number?

Every detector works one of two ways. A format check only confirms the shape is right, like a letter followed by eight digits. A checksum check goes further. It runs the same math the issuing body uses and drops a number that fails it. Take Jane Example's US passport number, C12345678. ONYRI confirms the shape but cannot confirm the number is real, because US passport numbers carry no public checksum. A French SIREN business number is different. ONYRI can verify its checksum on the spot.

France: which numbers does it detect?

ONYRI also looks for French passport and ID card numbers, plus RIB bank slips and tax reference numbers. All four share the same limit as the NIR. The shape is checked, but there is no public checksum to confirm a fake-looking one is genuine. When in doubt, ONYRI masks it anyway, which is the safer direction to err on.

United States, United Kingdom, and Belgium: which numbers does it detect?

ONYRI also reads US passport numbers, and US driver's license numbers when a label like DL or Driver's license sits right next to the value. Without that label, a short mix of letters and digits is too ambiguous to flag safely. UK and Belgian postcodes are matched by format too, though a Belgian postcode only counts once a city name follows it.

Germany, Austria, and Switzerland: which numbers does it detect?

These four checksums make DACH numbers some of the most reliable ONYRI detects. A German postal code needs more care. German capitalizes every common noun, so ONYRI only accepts a code next to a real place name, or next to a clear label such as PLZ, to avoid masking an ordinary amount like 2000 Euro.

Spain, Italy, the Netherlands, and Luxembourg: which numbers does it detect?

ONYRI also reads postal codes next to a city name: Spain's 01000-52999, Italy's 5-digit CAP with an optional province like (MI), the Netherlands' 1234 AB, and Luxembourg's L-1234. National phone numbers are matched too, plus the +34, +39, +31, and +352 country codes.

A Dutch KvK number is 8 bare digits, too ambiguous on its own: ONYRI needs a label like KvK right next to it. A bare BSN is 9 digits; it's masked once it passes the elfproef check, or with a label like BSN nearby. Without either, check the preview and add it under Also mask.

Dates spelled out in words are read too: 3 de marzo de 2026, 1° marzo 2026, and 12 maart 2026. Spanish, Italian, and Dutch first and last names are matched against built-in dictionaries. Luxembourg documents are often written in French or German, so ONYRI reads dates and names in both. It also guesses a document's country from its text (language, IBAN, phone prefix, postal code), so the default profile adapts on its own. A profile set to a specific country keeps it.

Which numbers does it detect in the 36 other countries?

Since October 2026, ONYRI Sanitize also covers 36 more countries, which makes 47 in all. For each one it knows the main ID, tax and company numbers, plus phone numbers, dates and, where the country uses them, postcodes and street addresses.

The same rule applies everywhere. A number in its distinctive shape, with its dashes, dots, slash or letters, is masked even if its check digit is wrong, so an OCR error cannot leave it exposed. A bare run of digits is masked only when its check digit is valid, or when a label such as PESEL, CPF or TFN sits right next to it.

Europe

The Americas and Oceania

The Middle East and North Africa

Asia

Numbers written with Arabic, Persian, Devanagari, Thai or full-width digits are read as ordinary digits, so a number typed in another script is still found. Names are read after a title or a label, and in signatures. In languages without a names model, such as Polish, Czech, Russian, Japanese or Korean, a bare name with no title or label can stay visible: check the preview and select it by hand. Canada is still not covered.

What about bank details, cards, and technical secrets?

• IBAN, in the covered countries that use it (not the United States, for example). ONYRI matches the shape (two letters, two digits, then 11 to 30 letters or digits) but does not run the interbank checksum or check each country's exact length, except in Saudi Arabia, the United Arab Emirates, Egypt and Morocco.
• Credit and debit card numbers, checked with the Luhn algorithm, the same math your card issuer runs.
• BIC and SWIFT codes, the bank identifier that often travels next to an IBAN.
• VAT numbers in several EU formats. The French and Italian keys are verified against their official formulas; the rest are matched by format only.
• Cloud and API keys, such as AWS, Google Cloud, Azure, Stripe, OpenAI, and Anthropic, plus Slack, GitHub, and GitLab tokens, JWTs, and SSH private keys. These sit on the Pro plan and above only, so a Free-plan export of a developer's log would leave them unmasked.

Free plans get 3 custom rules. Pro removes that limit, along with the cap on documents and pages. Either way, a rule you write once can be reused for every document in a profile, which is the practical way to cover a country ONYRI has not built a detector for yet.

Frequently asked questions

Sources & references