Presidio PII Detection
======================

**4 probes**

Integration with `Microsoft Presidio <https://microsoft.github.io/presidio/>`_
for detecting Personally Identifiable Information in dataset samples.  Supports
approximately 75 built-in entity types plus custom patterns defined in YAML.
Includes PAN detection with Luhn checksum validation, FinServ-specific
patterns (IBAN, SWIFT/BIC), and domain-specific recognisers loaded from a
user-supplied YAML file.

Install
-------

.. code-block:: bash

   pip install 'metapod-probes[presidio]'

This pulls in ``presidio-analyzer`` and ``presidio-anonymizer`` as transitive
dependencies.

Configuration
-------------

.. code-block:: yaml

   sources:
     - type: presidio
       dataset_path: /data/landing/*.csv
       sample_rows: 500
       score_threshold: 0.7
       custom_recognizers: /etc/metapod/pii_patterns.yaml
       entities:
         - PERSON
         - EMAIL_ADDRESS
         - CREDIT_CARD
         - IBAN_CODE
         - IT_FISCAL_CODE

Probes
------

.. list-table::
   :header-rows: 1
   :widths: 40 60

   * - Probe
     - Description
   * - ``presidio.compliance.pii_scan``
     - Scan dataset samples for PII entities (~75 built-in types)
   * - ``presidio.compliance.pan_detection``
     - Detect payment card numbers with Luhn checksum validation
   * - ``presidio.compliance.finserv_patterns``
     - Detect FinServ patterns (IBAN, SWIFT/BIC, account numbers)
   * - ``presidio.compliance.custom_patterns``
     - Evaluate user-defined recognisers from YAML configuration

Usage
-----

.. code-block:: bash

   metapod run "presidio.compliance.*" --config config.yaml

Output
------

.. code-block:: json

   {
     "probe": "presidio.compliance.pii_scan",
     "source": "presidio",
     "collected_at": "2026-03-31T08:00:00Z",
     "results": [
       {
         "dataset": "customers.csv",
         "column": "cf",
         "entity_type": "IT_FISCAL_CODE",
         "occurrences": 312,
         "score": 0.95,
         "sample_redacted": "RSSMRA85M01H501Z"
       }
     ]
   }