Privacy-preserving data publishing: A constraint-based clustering approach

Abstract Privacy-preserving data publishing has drawn much research interest re-cently. In this chapter, we address this topic from the viewpoint of constrained clustering, i.e., the problem of finding clusters that satisfy certain user-specified constraints. More specifically, we begin with the problem of clustering under aggregate constraints (without privacy considerations) and explain how traditional algorithms for the unconstrained problem (e.g., the c-means algorithm) break down in the presence of constraints. From there, we develop scalable algorithms that overcome this problem and finally illustrate how our algorithm can also be used for privacy-preserving data publishing.