Computational efficiency of k-anonymization incorporating clustering

Abstract

Data publicizing pose a threat of disclosing data subjects associating them to their personal sensitive information. k-anonymization is a practical method used to anonymize datasets to be made publicly available. The k-anonymization hides identities of data subjects by ensuring that every record of a publicized dataset has at least k �� 1 (k being a natural number) other records similar to it with respect to a set of attributes called quasi-identifiers. To minimize information loss, a clustering technique is often used to group similar records before k-anonymization is applied. Processing both the clustering and the k-anonymization using current algorithms is computationally expensive. It is within this framework that this research focuses on parallel implementation of the k-anonymization algorithm which incorporates clustering to achieve time effective computations