A K-anonymity model with strongly identifiable attributes
Mei Yu, Yu Du, Tianyi Xu, Jian Yun Yu, Yaqing Liu · 2013
In empirical studies of protecting privacy via anonymity, sensitive attributes are typically studied. Through models or algorithms, researchers guarantee some or all of their private information, resulting in a directed method. Sensitive attributes often are deleted until few. This paper analyzes a unique view of quasi-identifiers and shows that the distribution of quasi-identifiers is far from insignificant. In every information release, without exception, we find that there exists a ranking for quasi-identifiers, from low to high, such that almost all published information consist of lower-ranked quasi-identifiers with higher-ranked ones. We present a k-anonymity model with strongly identifiable attributes for deducing such rankings from observed published data. We hold the view that the rankings produced reflect a method of privacy protection.