quasiidentifiers

Quasi-identifiers are attributes that do not by themselves reveal a person, but can become identifying when combined with other information. They are distinct from direct identifiers such as names or social security numbers and from sensitive attributes that individuals wish to keep private. The risk they pose arises when external data sources—public records, voter lists, social media profiles, or commercial databases—can be linked to these attributes to re-identify individuals in an ostensibly anonymized dataset.

Common examples include date of birth, gender, and ZIP code. While any one of these values may not identify a person, certain combinations are highly identifying depending on the population and data context. The concept gained prominence in privacy research through work showing that simple demographic attributes often identify people when linked with public data.

In data publishing and analytics, quasi-identifiers drive re-identification risk assessments and guide anonymization strategies. Techniques such as generalization or suppression are applied to quasi-identifiers to achieve k-anonymity, ensuring that each combination of quasi-identifiers appears in at least k records. Extensions such as l-diversity and t-closeness aim to reduce the risk of attribute disclosure within those groups. More recently, differential privacy provides formal guarantees by adding controlled noise to query results, reducing the usefulness of external data for re-identification.

Limitations remain: risk is context-dependent and evolves with new external data sources. Organizations must continuously assess quasi-identifer risk when releasing data. See also k-anonymity, differential privacy.