Measuring The Effectiveness Of Anonymised Data

Anonymising data has become increasingly important due to the legal constraints imposed by authorities such as the EU's GDPR and for ethical reasons relating to privacy. One large drawback of anonymised data is its reduced quality (utility). Therefore it is crucial to quantify and minimise the utility loss prior to data sharing. We take a closer look at the question of how well this utility loss can be estimated for a specific task, in terms of effectiveness and efficiency of the resulting dataset. Our evaluation shows that the most valuable utility metrics are also the most expensive to measure, and thus often, a suboptimal solution must be chosen.