A fast perturbation algorithm using tree structure for privacy preserving utility mining

Abstract

As one of the important approaches in privacy preserving data mining, privacy preserving utility mining has been studied to find more meaningful results while database privacy is ensured and to improve algorithm efficiency by integrating fundamental utility pattern mining and privacy preserving data mining methods. However, its previous approaches require a significant amount of time to protect the privacy of data holders because they conduct database scanning operations excessively many times until all important information is hidden. Moreover, as the size of a given database becomes larger and a user-specified minimum utility threshold becomes lower, their performance degradation may be so uncontrollable that they cannot operate normally. To solve this problem, we propose a fast perturbation algorithm based on a tree structure which more quickly performs database perturbation processes for preventing sensitive information from being exposed. We also present extensive experimental results between our proposed method and state-of-the-art algorithms using both real and synthetic datasets. They show the proposed method has not only outstanding privacy preservation performance that is comparable to the previous ones but also 5–10 times faster runtime than that of the existing approaches on average. In addition, the proposed algorithm guarantees better scalability than that of the latest ones with respect to databases with the characteristics of gradually increasing attributes and transactions.