Data Anonymisation

Removing names is rarely enough. Age, postcode and job title together can single someone out, and a dataset can often be matched against other public data. Real anonymisation therefore combines techniques such as generalising values (age bands instead of exact ages), removing rare records, adding statistical noise or aggregating figures. Testers then try to re-identify people from the result. If identification stays possible, the data is only pseudonymised and the privacy rules still apply. Truly anonymous data usually falls outside data protection law, which is why organisations want it for research, analytics and sharing. Because linkable data keeps growing, a dataset that was anonymous years ago may no longer be. A claim of anonymity that fails a re-identification test leaves the data fully regulated.

A health insurer wants to share claims data with university researchers. It replaces dates of birth with age bands, drops names and policy numbers, groups small postcodes into regions and removes patients with very rare conditions. A privacy team then tries to match records against public registers and finds no unique matches before release.