Disclosure Risk Assessment
Statistical Disclosure Risk Assessment · Also known as: Microdata Disclosure Risk, Statistical Disclosure Control Risk Estimation, Istatistiksel Açıklama Riski Değerlendirmesi, Re-identification Risk Assessment
Disclosure Risk Assessment is a probabilistic framework introduced by Duncan and Lambert (1989) for quantifying how likely it is that releasing microdata — individual-level records from surveys or administrative files — will allow an outside party to identify a specific respondent or infer sensitive attributes. It is used by statistical agencies, data custodians, and researchers charged with protecting confidentiality before any public release of person-level datasets.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Apply Disclosure Risk Assessment when you must release a microdata file containing personal or sensitive records and need a defensible, quantified measure of re-identification risk prior to publication. Key assumptions are that quasi-identifiers can be enumerated, that population frequency estimates are obtainable, and that the intruder model (what the adversary knows) is explicitly specified. The method is less suited to unstructured data (text, images) or settings where the population universe is unknown. Complement it with utility metrics to balance privacy against analytical usefulness.
Strengths & limitations
- Provides a mathematically grounded, auditable probability estimate of re-identification rather than a qualitative judgment.
- Works directly with the actual data structure, requiring no additional generative model or synthetic substitute.
- Scalable to large administrative microdata files when efficient frequency tabulation is used.
- Output (file-level risk score) is interpretable by non-statisticians and meets regulatory reporting needs.
- Requires credible population frequency estimates; when the sampling frame is unavailable or stratification is complex, estimates can be imprecise.
- The intruder model must be pre-specified; underestimating adversary knowledge leads to systematically low risk estimates.
- Focuses on identity disclosure; attribute disclosure (inferring a sensitive value without identifying the person) requires separate modelling.
- Does not directly prescribe which protection method to apply — it diagnoses risk but does not resolve the privacy-utility trade-off.
Frequently asked
How is Disclosure Risk Assessment different from k-anonymity?
k-anonymity is a deterministic structural rule requiring that every record is indistinguishable from at least k-1 others on quasi-identifiers; it is enforced by transformation. Disclosure Risk Assessment is a probabilistic measurement framework that quantifies re-identification probability given a population model and an explicit intruder assumption. The two are complementary: k-anonymity specifies a protection target; disclosure risk measurement verifies whether that target is met.
What is a 'population unique' and why does it matter?
A population unique is a record whose combination of quasi-identifier values belongs to exactly one individual in the entire target population. Population uniques are the highest-risk records because an adversary who spots that combination in the released file immediately knows which person the record belongs to. Estimating the number of population uniques is therefore the core diagnostic in disclosure risk analysis.
Can the method be applied to continuous variables?
Directly, no: the frequency-counting approach assumes categorical or discretised quasi-identifiers. Continuous variables such as exact income or age in days must first be grouped into intervals before the population cell-size calculation. The choice of interval width is itself a modelling decision that affects both the measured risk and the analytical utility of the released data.
Sources
- Duncan, G. T., & Lambert, D. (1989). The risk of disclosure for microdata. Journal of Business & Economic Statistics, 7(2), 207–217. DOI: 10.1080/07350015.1989.10509729 ↗
How to cite this page
ScholarGate. (2026, June 2). Statistical Disclosure Risk Assessment. ScholarGate. https://scholargate.app/en/privacy/disclosure-risk
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Differential PrivacyPrivacy↔ compare
- k-AnonymityPrivacy↔ compare
- Synthetic Data GenerationPrivacy↔ compare