Re-identification

Polity & Governance glossary

Also called: De-anonymisation · Topic: Fundamental Rights and Constitutional Remedies · NCERT: Beyond NCERT

Meaning

Re-identification, also called de-anonymisation, means tracing data that was meant to be anonymous back to one real person. "Anonymised" data has names and other direct identifiers taken out. But small details like age, town and job can still be put back together to point to a single person. AI models make this easier. During training a model can memorise personal details, and the right questions can pull them back out. Re-identification matters because it shows that anonymisation alone does not protect anyone. That makes it a threat to the right to privacy, which the Supreme Court held to be part of Art. 21 in the K.S. Puttaswamy case (2017).

Example

Say an AI company trains a model on millions of public social media posts. One post on its own seems harmless. But if someone queries the model with "anonymised" details about a person, it can join scattered facts and name them. It can then help build a profile of where they live, what they do and who they know. Such a profile can be used for surveillance, discrimination or fraud.

Don't confuse with

  • Regurgitation: this is when an AI model repeats personal data it memorised during training. Re-identification is the further step of linking data to a specific, named person.
  • Anonymisation: this is the safeguard of removing identifiers. Re-identification is how that safeguard gets defeated.

Read more