Course contentsShow
← AI Engineering
Lesson 1533 of 1,88637. Privacy and Data GovernanceFree lesson

Re-identification Risk Assessment

Evaluate and quantify the risk that anonymized or pseudonymized data can be linked back to individuals through auxiliary information.

Re-identification Risk Assessment

What you'll learn: How to measure the likelihood that anonymized or pseudonymized data can be traced back to specific individuals using external information.


Even after you've applied pseudonymization or anonymization techniques (hashing, masking, tokenization), your data isn't automatically safe. Re-identification risk assessment is the systematic process of evaluating whether someone could still figure out who's who by combining your "anonymized" dataset with other publicly available or auxiliary information.

The Core Challenge

Imagine you've masked all names in a medical dataset but kept age, zip code, and diagnosis. Sounds safe? Research shows that 87% of Americans can be uniquely identified with just those three fields using census data. Re-identification risk assessment quantifies this danger before you expose the data.

How Assessment Works

You evaluate risk by:

  1. Identifying quasi-identifiers: Fields that seem harmless alone (birth year, job title, location) but are identifying when combined
  2. Measuring uniqueness: How many records share identical quasi-identifier combinations? Unique records = higher risk
  3. Considering auxiliary data: What external datasets exist? Social media profiles? Public records? Voter registrations?
  4. Calculating k-anonymity: Ensuring every record is indistinguishable from at least k-1 others
  5. Threat modeling: Who might attack? What resources do they have?

Practical Metrics

  • k-anonymity score: Minimum group size sharing same quasi-identifiers
  • l-diversity: Ensuring sensitive attributes vary within groups
  • Uniqueness percentage: Fraction of records that are singletons
  • Linkage probability: Statistical chance of successful re-identification

Unlike your earlier work with PII detection or pseudonymization techniques, this focuses on measuring residual risk after protection.

Key Takeaway: Re-identification risk assessment quantifies whether your anonymization efforts are actually sufficient by modeling how attackers could reverse-engineer identities using auxiliary data sources.