A Sensitivity-Adaptive \(\rho \)-Uncertainty Model for Set-Valued Data

Set-valued data brings enormous opportunities to data mining tasks for various purposes. Many anonymous methods for set-valued data have been proposed to effectively protect an individual’s privacy against identify linkable attacks and item linkage attacks. In these methods, sensitive items are protected by a privacy threshold to limit the re-identification probability of sensitive items. However, lots of set-valued data have diverse sensitivity on data items. This leads to the over-protection problem when these existing privacy-preserving methods are applied to process the data items with diverse sensitivity, and it reduces the utility of data. In this paper, we propose a sensitivity-adaptive \({\rho }\)-uncertainty model to prevent over-generalization and over-suppression by using adaptive privacy thresholds. Thresholds, which accurately capture the hidden privacy features of the set-valued dataset, are defined by uneven distribution of different sensitive items. Under the model, we develop a fine-grained privacy preserving technique through Local Generalization and Partial Suppression, which optimizes a balance between privacy protection and data utility. Experiments show that our method effectively improves the utility of anonymous data.