Is exactly what it says: finding some way, for at least one person, of identifying that person despite the dataset in which their data appears having been anonymised or, more likely and precisely, pseudonymised. De-anonymisation is another name for the same process.
Details #
This can happen in many ways but clearly what is necessary is that there is something in the reported data that can uniquely identify that person to at least one person. Legally, it’s failed protection of anonymity/confidentiality if the person in question can recognise him/herself. This actually illustrates one rather crude distinction between qualitative case reports and quantitative data. For some case reports the person described might be unhappy not to be able to recognise him/herself and only concerned that s/he should not be identifiable (“re-identified”) by someone with disrespect or frankly malicious intent. In primarily qualitative work there can be a hard challenge not to change details so much that the report becomes in some way misleading while still making sure de-anonymisation is impossible. (Of course, for some work the participant(s) may be willing to be identifiable.) One advantage of purely quantitative work is that things like questionnaire scores are almost guaranteed not to identify someone (but beware that a fellow client or staff member might remember a record score and hence be able to re-identify someone in open data if not from summary tables, statistical tests/explorations or graphics). However, demographic data in an open dataset might put a participant in an smallest identical subset containing just that person and that may make then identifiable (again, probably mostly to fellow participants/clients or staff).
Try also #
Anonymisation
Cell size
Confidentiality
Jigsaw attack
n > 5 rule
Pseudonymisation
Smallest identical subset
Chapters #
The issues run throughout the book but particularly Chapters 6, 7 and 8.
Dates #
First created 26.xi.23.
