Anonymizing JSON without breaking its structure

Questions to answer first

Identify the recipient, the purpose of disclosure and the minimum information needed for that purpose. A closed internal team and a public data release create different risk. Consider whether the recipient possesses another dataset that can link a record back to a person.

Step-by-step workflow

1. Work on a copy. 2. Load the material locally. 3. Select the appropriate detection profile. 4. Review every match and mark missed data manually. 5. Choose masking, a pseudonym or generalization. 6. Apply changes. 7. Run the second scan. 8. Open the output in its normal application. 9. Remove metadata and unnecessary copies.

Before-and-after example

Before: “John Smith from a small town reported the event on 14 February 2026. Contact: john@example.com.” After: “[PERSON_1] from [REGION_1] reported the event in February 2026. Contact: [EMAIL_1].” The event description and rare attributes may still point to one person.

Common mistakes

Common failures include removing only a name, leaving comments or presenter notes, keeping the map beside the output, publishing very small groups, retaining exact dates, forgetting hidden worksheets and trusting a zero-detection result without manual inspection.

How to verify the output

Search manually for surnames, email fragments, identifiers, places and distinctive events. Compare before and after without sharing the original with a recipient who should only receive the output. For tabular data, inspect rare quasi-identifier combinations. For documents, review comments, notes, properties and metadata.

Important limitation

A technical tool cannot issue a legal determination. Pseudonymized information may remain personal data. Anonymity depends on context and on the means that can realistically be used to identify people now or later.

Use the tool

Open the relevant preset, review matches and retain the report together with a record of the decisions taken.

FAQ