Glossary

Glossary of personal data and anonymization.

The words of data protection, and the identifiers a local anonymizer must recognize in a French case file: a short definition, and what AnonymFiles does with each.

Local anonymizer

Software that detects and replaces personal data in a document on the computer itself, without sending it to an online service. That is what AnonymFiles does: what “local” really means.

Personal data

Any information relating to an identified or identifiable natural person, directly or indirectly (GDPR, art. 4.1): a name, but also an address, a number or a combination of details.

Sensitive data

A special category whose processing is prohibited except in specific cases (GDPR, art. 9): health, political opinions, religious beliefs, trade union membership, origin, sexual orientation, genetic and biometric data.

Direct identifier

Data that designates a person on its own: name, social security number, personal e-mail, phone number.

Indirect identifier

A detail that, combined with others, makes it possible to find a person: a rare job, a date, an amount, a small town, a docket number. Re-identification risk.

Pseudonymization

Replacing identifying data with pseudonyms, the mapping being kept separately and protected (GDPR, art. 4.5). The data remains personal. Anonymization vs pseudonymization.

Anonymization

Irreversible processing after which nobody can be identified by any reasonable means; anonymous data falls outside the GDPR (recital 26).

Redaction

Permanently removing a passage of a document before disclosing it. In a PDF, the text must be deleted from the file, not merely covered with a black box. Pseudonymization or redaction.

Mapping table

The file linking each label to its original value. In AnonymFiles, it is encrypted (AES-256-GCM) and never leaves the computer. The GDPR rules.

Label

A stable pseudonym replacing a value, such as [PERSONNE_1] or [ORGANISATION_2]: the same person keeps the same label across the whole file, and a label can be renamed ([CLIENT]).

Putting names back

Replacing labels with the real values in an AI's answer, on your computer, as text, Word or PDF. Pseudonymize, then put the real names back.

Re-identification

Finding a person again from data that was supposed to no longer identify them, by singling out, linkability or inference.

NIR

The French social security number: 13 digits and a 2-digit check key. AnonymFiles recognizes it and verifies the key.

SIREN, SIRET

French company identifiers (9 and 14 digits). For a sole trader, they designate a person: they are then personal data.

EU VAT number

A company's tax identifier in the European Union, derived from the SIREN in France (FR + key + SIREN).

IBAN, BIC

International bank details. AnonymFiles recognizes IBANs from nine European countries, check digits verified, and BICs announced as such.

RPPS, ADELI

French health professional identifiers (11 and 9 digits), masked when announced as such. AnonymFiles for healthcare.

FINESS

French identification number of health and social care facilities (9 characters).

IPP, INS

The permanent patient identifier assigned by a facility, and the French national health identity: both masked when announced as such.

Metadata

Hidden information in a file: author, software, dates, history, comments. It is removed on export. Remove PDF metadata.

OCR

Optical character recognition: reading the text of a scan or an image. In AnonymFiles, it runs on the computer, in French and English. Anonymize a scanned PDF.

MCP (Model Context Protocol)

An open protocol that lets an AI such as Claude read documents through a connector. AnonymFiles' connector only gives it the redacted copies. The MCP connector.