Is your feature request related to a problem? Please describe.
Presidio covers Korean identity/tax numbers (KR_RRN, KR_BRN #1822, KR_FRN #1825, KR_PASSPORT #1814, KR_DRIVER_LICENSE #1820, and KR_CRN proposed in #2177) and has bank account coverage for the US (US_BANK_NUMBER), but there is no recognizer for Korean bank account numbers, one of the most common PII types in Korean financial and business text.
Describe the solution you'd like
A KrBankAccountRecognizer (entity KR_BANK_ACCOUNT) with three patterns:
- NongHyup 302-prefixed 4-segment personal accounts (
302-XXXX-XXXX-XX), medium score
- Common hyphenated 3-segment layouts (e.g.
110-234-567890), weak score
- Mixed-separator 9-16 digit runs, very weak score
Korean bank accounts have no unified format and no check digit, so scores stay conservative and rely on context words (계좌, 계좌번호, 예금주, ...). To keep precision, every pattern embeds negative lookaheads excluding Korean mobile/VoIP phone numbers (010/070, a false positive we actually hit in production), resident registration number shapes, and dates. The patterns are ported from a production Korean PII-masking deployment.
Open questions:
- Entity name:
KR_BANK_ACCOUNT vs KR_BANK_NUMBER (US precedent is US_BANK_NUMBER)
- Whether to include bank brand names (우리은행, 국민은행, 카카오뱅크, ...) as context terms. They help recall, but it's a maintenance list.
- 3-2-5 digit strings are both valid BRN and valid account layouts at some banks; currently left to context disambiguation.
I'd like to submit a PR for this.
Is your feature request related to a problem? Please describe.
Presidio covers Korean identity/tax numbers (KR_RRN, KR_BRN #1822, KR_FRN #1825, KR_PASSPORT #1814, KR_DRIVER_LICENSE #1820, and KR_CRN proposed in #2177) and has bank account coverage for the US (
US_BANK_NUMBER), but there is no recognizer for Korean bank account numbers, one of the most common PII types in Korean financial and business text.Describe the solution you'd like
A
KrBankAccountRecognizer(entityKR_BANK_ACCOUNT) with three patterns:302-XXXX-XXXX-XX), medium score110-234-567890), weak scoreKorean bank accounts have no unified format and no check digit, so scores stay conservative and rely on context words (계좌, 계좌번호, 예금주, ...). To keep precision, every pattern embeds negative lookaheads excluding Korean mobile/VoIP phone numbers (010/070, a false positive we actually hit in production), resident registration number shapes, and dates. The patterns are ported from a production Korean PII-masking deployment.
Open questions:
KR_BANK_ACCOUNTvsKR_BANK_NUMBER(US precedent isUS_BANK_NUMBER)I'd like to submit a PR for this.