Privacy Risk Evaluation of Re-Identification of Pseudonyms
Yuma Takeuchi, Shogo Kitajima, Kazuya Fukushima, Masahiro Mambo · 2019
In recent years, personal data collected from applications or web sites have been utilized in various occasions all over the world. Under such circumstances, evaluating the anonymity of personal data quantitatively is quite important in terms of privacy preservation. In this paper, we propose Average Re-identification Rate(ARR), as one of the anonymity-evaluation method, which is based on a similarity of historical data. Our proposing method adopts a multi-pseudonymization which updates pseudonyms on fixed-time intervals, and evaluates re-identification risk in a certain interval determined in advance. We also apply several methods to evaluate data similarity such as Jaccard index and inverse document frequency(idf) to the proposing method and compare their properties on a collected web-browsing history data. Moreover, we consider speed-up methods to calculate similarity.