déidentification

Déidentification is the process of reducing or removing personal identifiers from data to prevent the identification of individuals. It is used to enable data sharing and secondary analysis while aiming to protect privacy. In practice, it encompasses methods that reduce linkability but may still leave data in a form that could be re-identified under certain conditions.

The term is related to, but distinct from, anonymization and pseudonymization. Anonymization is typically considered irreversible, removing all means to identify a person, while pseudonymization replaces direct identifiers with substitutes that require a separate key to re-link to individuals. Déidentification often falls along a spectrum that includes both approaches, depending on the methods and the risk tolerance of a given context.

Techniques employed include direct removal of identifiers (names, addresses, identifiers), generalization or aggregation of data, masking or perturbation of details, and, in some cases, the creation of synthetic data. Formal risk-reduction approaches such as k-anonymity, l-diversity, and t-closeness provide frameworks for measuring residual identifiability, while differential privacy offers mathematical guarantees to limit what can be learned about individuals from the data.

Applications are common in healthcare, public health, statistical reporting, and research, where data need to be usable without exposing individuals. However, de-identification does not eliminate all privacy risk; re-identification can still be possible with auxiliary information or data linkage. Ongoing risk assessment, governance, data-use controls, and appropriate access restrictions are essential components of a robust de-identification strategy.

Regulatory and professional guidance often distinguishes de-identification from other privacy-preserving techniques. For example, HIPAA in the United States outlines methods for de-identification, while GDPR emphasizes pseudonymization and data minimization as protective measures. See also: anonymization, pseudonymization, differential privacy.