Linkage attack

A linkage attack (also known as a record linkage or de-anonymization attack) occurs when anonymized data is combined with one or more external datasets to re-identify individuals. In essence, overlap in quasi-identifying details—like ratings, timestamps, demographic traits, or behavioral patterns—can act like a fingerprint that links a supposedly anonymous record back to a real person.

For example, researchers demonstrated that by matching Netflix’s anonymized movie ratings (ratings, dates, movie IDs) with the same user’s public IMDb ratings, they could re-identify specific users from the Netflix dataset. This minor overlap in non-identifying data enabled them to connect the dots despite the lack of explicit identifiers.