A Clustering Approach for Achieving Data Privacy.
Alina Câmpan, Traian Marius Truţă, John H. McB. Miller, Raluca Sinca · 2007
Abstract — New privacy regulations together with everincreasing data availability and computational power have created a huge interest in data privacy research. One major research direction is built around k-anonymity property and its extensions, which are required for the released data. In this paper we present such an extension to k-anonymity, called p-sensitive k-anonymity, which solves some of the weaknesses that the k-anonymity model has been shown to have. We also introduce a new algorithm for enforcing p-sensitive k-anonymity on microdata sets based on a greedy clustering approach. To limit the amount of information loss the proposed algorithm uses cell-level generalization for categorical attributes and hierarchy-free generalization for numerical attributes. Our belief is that the above mentioned algorithm can be adjusted and used to enforce other similar privacy models as well, with better results than the algorithms originally proposed along with these models. Our experiments show that the proposed algorithm efficiently generates the masked microdata with p-sensitive k-anonymity property. I.