Research publishing method that can accurately estimate the likelihood of a specific person to be correctly re-identified, even in a heavily incomplete dataset.

Basically how to de-anonymised people in random anonymised datasets:

https://www.nature.com/articles/s41467-019-10933-3