What it detects
pii-airlock combines Microsoft Presidio, a language model for names, and its own checked recognizers. Use these type names with pii-airlock action.
Common types
| Type | Examples |
|---|---|
PERSON | Names |
EMAIL_ADDRESS | jean@acme.fr |
PHONE_NUMBER | +33 6 12 34 56 78, 079 123 45 67 |
CREDIT_CARD | Card numbers |
IBAN_CODE | Bank account numbers |
IP_ADDRESS | 192.168.1.10 |
LOCATION | Cities, addresses |
SECRET | API keys, tokens, private keys, password=… values |
Extra recognizers that work in any language
These check the number's built-in check digit, so a random number does not trigger them.
| Type | What it is |
|---|---|
FR_NIR | French social security number |
CH_AHV | Swiss AHV/AVS number |
FR_SIRET, FR_SIREN | French business IDs |
DE_TAX_ID | German tax ID (needs a nearby keyword such as Steuer-ID) |
DE_TAX_NUMBER | German Steuernummer (needs a keyword) |
IT_VAT_CODE | Italian VAT number |
SECRET | Keys for AWS, GitHub, Slack, OpenAI, Anthropic, Google and Stripe, plus JWTs |
Ignored by default
These fire on almost every date and web link, so they are off unless you ask:
DATE_TIME, URL, NRP, US_DRIVER_LICENSE, US_BANK_NUMBER, US_PASSPORT, IN_VEHICLE_REGISTRATION
An email or name inside a link is still caught by its own recognizer.
To turn one on, set exclude_entities to a list without it, or an action for that type. To turn off all the extra recognizers: builtin_recognizers: false.