pii-core

Multi-language PII detection with regex and checksum validation. Pure Python, zero runtime dependencies.

The foundation library that pii-veil (reversible anonymization) and pii-presidio (Microsoft Presidio plugin) build on.

Install

pip install pii-core

What's in v0.1.0

Polish identifiers with regex + checksum validation:

• PESEL (national ID), NIP (tax ID), REGON (business registry — 9 and 14 digit) — weighted-sum checksums.
• Polish IBAN (PL prefix, mod-97 validated).
• Polish ID card (dowód osobisty), passport — regex only (official checksums not yet implemented).
• Polish mobile phone (+48, optional separators).
• Opt-in: KRS court register number (10 digits, no checksum), postal code (XX-XXX). Excluded from DEFAULT_DETECTORS because their raw patterns collide with ordinary text — pair them with a context-word filter.

Cross-language detectors:

• Email addresses (practical subset, not strict RFC 5322).
• Credit-card numbers (Luhn-validated; bare, dashed, and spaced shapes for Visa / MC / Amex / Discover / Diners).

Multi-country IBAN validator (is_valid_iban) covering ~80 countries via the published SWIFT registry. The PlIbanDetector regex is Polish-only, but the checksum function is general.

Quick usage

from pii_core import DEFAULT_DETECTORS

text = "Mój PESEL: 44051401358, kontakt: jan@example.pl."

for detector in DEFAULT_DETECTORS:
    for match in detector.detect(text):
        print(f"{match.detector}: {match.value!r} at {match.start}-{match.end}")

from pii_core import is_valid_pesel, is_valid_iban, is_valid_luhn

is_valid_pesel("44051401358")            # True
is_valid_iban("DE89370400440532013000")  # True
is_valid_luhn("4111111111111111")        # True

Opt-in detectors (KRS, postal code) live in pii_core.pl:

from pii_core import DEFAULT_DETECTORS
from pii_core.pl import PlKrsDetector, PlPostalCodeDetector

# Add them only when you have context-word filtering elsewhere in your pipeline.
my_detectors = [*DEFAULT_DETECTORS, PlKrsDetector(), PlPostalCodeDetector()]

API stability

PIIType value strings, detector .name strings, and the order of DEFAULT_DETECTORS are SemVer-stable: downstream consumers persist them in serialized mappings or use them as overlap-resolution priority keys. Internal changes (regex tweaks, helper renames) can churn within minor versions.

Sibling packages

• pii-veil — reversible anonymization with persisted mapping and CLI, built on pii-core.
• pii-presidio — Microsoft Presidio plugin wrapping pii-core recognizers with optional reversible anonymization.

License

Apache-2.0. See LICENSE.