Anonymization Of Set-valued Data For Privacy Preserving Data Publishing

The collection of digital information by governments, corporations, and individ-uals has created tremendous opportunities for knowledge-and information-based de-cision making. Driven by mutual benefts, there is a demand for the exchange andpublication of data among various parties. Data in its original form, however, typicallycontains sensitive information about individuals, and publishing such data will violateindividual privacy.Privacy-preservingset-valueddatapublishingisanimportantyetchallengingprob-lem. While most existing techniques involve generalization or global suppression ofthedataitems, wepresentapartial(local)suppressionmethodtoanonymizeset-valueddata. This method ensures no strong inference of sensitive information is possible re-gardless of the amount of background knowledge the attacker possesses. The proposedapproach not only minimizes the number of item deletions, but can also preserve theoriginal data distribution or retain mineable useful association rules depending on therequirements of the downstream applications. Preliminary evaluation shows that ourapproach outperforms the peers in preserving the original data distribution by morethan100times, and retaining many more mineable useful association rules with fewspurious rules, while reducing the number of deletions by30%on average.