DataTerminology.comNEW
Process of transforming data to prevent identification of individuals while preserving analytical utility.
DESCRIPTION
Data anonymization removes or transforms identifiers and quasi-identifiers to reduce re-identification risk. Techniques include generalization, suppression, aggregation, perturbation, and synthetic data generation. It supports regulatory compliance (e.g., GDPR, HIPAA) and responsible data sharing. True anonymization should be irreversible; otherwise, it may be pseudonymization. Risk assessments and utility evaluations guide appropriate techniques.
ALTERNATIVE DEFINITIONS
Wikipedia
Data anonymization is the process of protecting private or sensitive information by erasing or encrypting identifiers that connect an individual to stored data.
Focuses on removing links between identity and data.
U.S. HHS
De-identification is the process used to prevent a person’s identity from being connected with information.
Regulatory framing in the context of HIPAA compliance.
ENISA
Anonymisation is a process by which personal data is irreversibly altered in such a way that a data subject can no longer be identified directly or indirectly.
Stresses irreversibility as a key criterion of anonymisation.
RELATED TERMS
K-anonymity HIPAA GDPR Data Masking Tokenization Re-identification Privacy Risk Data Minimization
COMPETING TERMS
Pseudonymization Differential Privacy
RELATIONS
- Broader: Anonymization
KNOWLEDGE AREAS
Data Privacy & Compliance Data Security AI Governance Data Management
TARGET AUDIENCE
Data Governance Manager Data Steward Chief Data Officer Data Engineer Data Scientist
EXTERNAL SOURCES
- Wikipedia - Data anonymization
- GDPR.eu - EU GDPR perspective on anonymization
- NIST - NIST glossary entry for de-identification
- OpenAIRE - Practical guide on data anonymization
TOP QUESTIONS
How does Data Anonymization differ from pseudonymization?
Anonymization is intended to be irreversible; pseudonymization replaces identifiers with tokens but can be reversed with a key.
Category: Comparison
What are common anonymization techniques?
Generalization, suppression, aggregation, noise addition, micro-aggregation, and synthetic data are common approaches.
Category: HowTo
What risks remain after anonymization?
Re-identification risks persist via linkage attacks or auxiliary data. Risk assessments and controls are essential.
Category: Risk
When should I use differential privacy instead of anonymization?
Use differential privacy for provable privacy guarantees in query outputs or synthetic data, especially for high-dimensional datasets.
Category: Guidance
How do regulations view anonymized data?
Under GDPR, truly anonymized data falls outside personal data scope; insufficient anonymization may still be regulated.
Category: Compliance
What is k-anonymity?
A dataset satisfies k-anonymity if each record is indistinguishable from at least k−1 others on quasi-identifiers.
Category: Understanding
How to measure utility after anonymization?
Compare distributions, model performance, and key metrics before and after; use task-specific utility measures.
Category: Evaluation
Which data fields are highest risk?
Direct identifiers (names, SSNs) and quasi-identifiers (DOB, ZIP, gender) pose re-identification risks when combined.
Category: Risk
Can anonymization be applied to unstructured data?
Yes, but it is harder. Use NLP-based redaction and entity recognition, with careful validation.
Category: HowTo