Re-Identification risk in anonymized data sets with parent-child information

Abstract

We explore the risk of re-identification in anonymized data sets that preserve genealogical information (i.e. parent-child links). We consider attacks based on the number of children of an individual. We use part of a well known data set in our experiments, which show that a substantial part of the population involved in the set can be identified, even if additional anonymity protection measures based on graph trimming are taken. We have also found that the risk quickly increases with generatiou depth.