Customizing Privacy Protection in Data Publishing

Distortion of data prior to publishing is one of the primary approaches to make sensitive data free of any illegal access or malicious use. Privacy customization has not been emphasized and well-studied in related literature. In this paper, data owners' preferences and data attributes' characteristics are taken into consideration. A privacy customization strategy is proposed and accomplished via a group distortion technique based on matrix decomposition. Several privacy and utility measures are studied. The performance of the proposed strategy is evaluated and compared to a conventional full distortion method. Our evaluation demonstrates that the proposed strategy has some attractive properties including an improved utility. In this way, a tradeoff between privacy and utility becomes more feasible.