Anonymizing Transaction Data to Eliminate Sensitive Inferences

Publishing transaction data containing individuals? activities may risk privacy breaches, so the need for anonymizing such data before their release is increasingly recognized by organizations. Several approaches have been proposed recently to deal with this issue, but they are still inadequate for preserving both data utility and privacy. Some incur unnecessary information loss in order to protect data, while others allow sensitive inferences to be made on anonymized data. In this paper, we propose a novel approach that enhances both data utility and privacy protection in transaction data anonymization. We model potential inferences of individuals? identities and their associated sensitive transaction information as a set of implications, and we design an effective algorithm that is capable of anonymizing data to prevent these sensitive inferences with minimal data utility loss. Experiments using real-world data show that our approach outperforms the state-of-the-art method in terms of preserving both privacy and data utility.