Privacy & data

k-anonymity

A dataset property where every record is indistinguishable from at least k-1 others on its quasi-identifiers.

k-anonymity (Sweeney, 2002) generalises or suppresses quasi-identifiers (age, postcode, gender) until each combination appears at least k times. Sweeney showed that ZIP code, birth date and sex alone uniquely identify most of the US population, which is why removing names is not anonymisation.

It fails against homogeneity (all k records share the same diagnosis) and background knowledge, which led to l-diversity and t-closeness. It also degrades badly with many attributes, and releasing several k-anonymous views of the same data can be combined to re-identify people.

The term is also used for the range-query pattern in breached-password checks: the client sends only the first five hex characters of a SHA-1 hash and matches the returned suffixes locally, so the service never learns the full hash.

Learn it properly

Also used in

Related terms