PE-14 De-identification and re-identification risk assessment (data situation audit; direct and indirect identifier removal; disclosure risk assessment)
Removes or alters identifying details and assesses the risk that a person can still be identified in the context of release, giving a de-identified dataset and a re-identification risk statement.
What it does
Removes or alters identifying details and assesses whether a person could still be identified once the data are released or analysed in context. Artefact: a de-identified dataset and a short re-identification risk statement. Read the risk in the context of who will see the output and what other data they hold.
When to use it
Use when material holds personal or sensitive information and the analysis, sharing or model use needs it removed or reduced first.
When it does not fit
Does the material hold personal or sensitive information that the analysis, sharing or model use does not need? If yes, run this before any other technique.
Result format
The result has a de-identified dataset and a short re-identification risk statement. The statement names the identifiers removed or altered, who will see the output and what other data they hold, and gives the risk judged in that context.
Requirements on the material
| Property | Requirement type | Requirement |
|---|---|---|
| Unit of analysis | Always applies | requires=record; requires=document |
| Appraises source credibility | Applies in some cases | no |
| Personal information | Always applies | requires=personal information; requires=sensitive information; requires=unknown |
Not restricted on: Structure, Content type, Variable types, Time structure, How the sample was chosen, Level of detail available, Number of sources, Whether sources can be joined, Amount of material, Known data problems, Producer interest, Source dependence, Origin, Document genre, Language, Security classification, Licence.
| Minimum material | The intended use of the output, the people who will see it, and a list of other data that could be linked to it. |
|---|---|
| Additional sources needed | none |
| Not for (rules) | none |
| Not for (text) | Removing direct identifiers such as names and addresses is read as de-identification, and complex or publicly released datasets still need specialist advice. |
Choosing and ordering
| Kind of technique | Preparation technique, runs in step 5.2 |
|---|---|
| Phase tag | verify and prepare |
| Other phase tags | none |
| Kind of finding | risk |
| Question answered | what |
| Strongest statement it supports | describe |
| Needs as input |
|
| Makes as output |
|
| Must exist first | None |
| Loop target | return to gather |
Library checks by stakes
Each stakes level includes the checks of the level below it. Checks that need a person apply at high stakes only, so the process leaves them out at lower stakes (decision 38). Repeatability: judgement dependent.
| Stakes | Check | Person involved | Counts as a different method |
|---|---|---|---|
| exploratory | input validation | none | no |
| exploratory | spot check | optional | no |
| decision-informing | input validation | none | no |
| decision-informing | spot check | optional | no |
| decision-informing | human review sample | needed; left out below high stakes | no |
| decision-informing | assumption and limitation statement | none | no |
| high-stakes | input validation | none | no |
| high-stakes | spot check | optional | no |
| high-stakes | human review sample | needed | no |
| high-stakes | assumption and limitation statement | none | no |
| high-stakes | independent review | needed | no |
| high-stakes | expert review | needed | no |
Failure modes and limits
| Model suitability | human judgement required |
|---|---|
| Typical model failure | Removes names and addresses and declares the data de-identified, although the remaining details in combination can still identify a person. |
| Error sources | input data; data transformation; language model |
| Certainty limits | Certainty that the data are de-identified drops when other datasets could be linked in (indirectness) and when the context of release is not known. |
Related techniques
- pairs with PE-08 Named entity recognition (NER; entity extraction): named entity recognition finds the names and places to remove
- pairs with PE-04 Data profiling and rule-based validation (data audit; data quality assessment; edit checks): profiling shows which fields could identify a person
Evidence and tag basis
Documented. The Office of the Australian Information Commissioner (OAIC) states that removing direct identifiers alone may not de-identify data under the Privacy Act and that re-identification risk must be assessed in context. The OAIC and CSIRO Data61 De-identification Decision-Making Framework has three activities: a data situation audit, risk analysis and control, and impact management. (OAIC, De-identification and the Privacy Act; OAIC, De-identification Decision-Making Framework)
Tag basis: input profile inference; selection inference; sequencing inference; robustness inference. All tag values are inference unless a source is named.