Direct comparison
Anonymization vs. Pseudonymization
Anonymization is irreversible and exits GDPR scope; pseudonymization is reversible and stays in scope. Compare techniques, legal basis, and use cases.
Side-by-side comparison
| Dimension | Anonymization | Pseudonymization |
|---|---|---|
| What it does | Irreversibly removes or alters identifying information | Replaces identifiers with a code/token; original identity recoverable via a separately-held key |
| Reversibility | Not reversible, by definition | Reversible by anyone with access to the key file |
| GDPR status | Falls outside GDPR scope entirely (Recital 26) once genuinely anonymous | Remains personal data — GDPR still applies in full (Article 4(5), Recital 26) |
| GDPR basis | Not defined as a processing operation with its own article; outcome described in Recital 26 | Defined in Article 4(5); named as a research safeguard in Article 89(1) |
| Re-contacting participants | Not possible — no link back to identity exists | Possible via the key file, subject to access controls |
| Typical techniques | Suppression, generalization, perturbation, k-anonymity, differential privacy, synthetic data | Coded/tokenized identifiers, encryption with a separately-stored key, linkage tables |
| Key file / crosswalk | None exists — destroyed or never created | Exists, stored separately with access controls |
| Comparable US standard | HIPAA Safe Harbor / Expert Determination (45 CFR 164.514(a)-(b)) removes data from PHI scope | No exact HIPAA equivalent — coded data with a retrievable key generally remains PHI |
| When typically used in research | Public repository deposit, long-term archiving, once re-contact is no longer needed | Active data collection and analysis, when re-contact or correction may be needed |
| Risk if done incorrectly | Re-identification via combination with other datasets (quasi-identifier linkage) | Key file exposure or weak access controls defeat the protection |
Common questions
FAQ
Is pseudonymized data still personal data under GDPR?+
Yes. GDPR Recital 26 states that pseudonymised data which could be attributed to a person using separately-held additional information is still considered information on an identifiable natural person, so it remains fully within GDPR's scope.
Can anonymized data be re-identified later?+
If it truly meets the anonymization standard, no — by definition, re-identification must not be possible by any means reasonably likely to be used, accounting for cost, time, and available technology. If a dataset can later be re-identified by reasonably available means, it was not properly anonymized in the first place.
Does deleting the key file turn pseudonymized data into anonymized data?+
It can move it toward anonymization, but only if no other reasonably available means of re-identification remain — for example, no other copy of the key exists anywhere, and the data cannot be re-linked by combining it with other accessible datasets. This should be assessed and documented, not assumed.
Which one should a data management plan specify?+
State explicitly which technique is used at which stage — for example, pseudonymized during active collection and analysis, anonymized before public deposit — rather than using the terms interchangeably. Funders and ethics boards increasingly expect this level of specificity.
