An Effective and Efficient Heuristic Privacy Preservation Algorithm for Decremental Anonymization Datasets

Abstract

(α, k)-Anonymity is a well-known anonymization model that is extended from k-Anonymity. It is proposed to address privacy violation issues in published datasets from using identity linkage attacks, attribute linkage attacks, and probability inference linkage attacks. Unfortunately, (α, k)-Anonymity is generally sufficient to preserve the privacy data in datasets that are focused on performing one-time data publishing. Thus, if published datasets, e.g., decremental datasets, are dynamic, i.e., the data of them is always changed by using deletion methods and multiple time data publishing, then the privacy data of users is collected in these published datasets could be violated by using such an appropriate comparison data attacking. To rid this vulnerability of (α, k)-Anonymity, an effective decremental privacy preservation algorithm is available in existence. Although this algorithm can address privacy violation issues in published decremental datasets, it has a vital vulnerability that must be improved, i.e., it is highly complex in terms of transforming the data which is available datasets to satisfy the specific privacy preservation constraints. For this reason, a heuristic privacy preservation algorithm for publishing decremental datasets based on clustering techniques to be proposed in this work. With the proposed algorithm, aside from privacy preservation, the data utility and execution time are also maintained as much as possible. Furthermore, we show the experimental results which indicate that the proposed algorithm is highly effective and efficient.