Novel Approaches for Privacy Preserving Data Mining in k-Anonymity Model

摘要

In privacy preserving data mining, anonymization based approaches have been used to preserve the privacy of an individual. Existing literature addresses various anonymization based approaches for preserving the sensitive private information of an individual. The k-anonymity model is one of the widely used anonymization based approach. However, the anonymization based approaches suffer from the issue of information loss. To minimize the information loss various state-of-the-art anonymization based clustering approaches viz. Greedy k-member algorithm and Systematic clustering algorithm have been proposed. Among them, the Systematic clustering algorithm gives lesser information loss. In addition, these approaches make use of all attributes during the creation of an anonymized database. Therefore, the risk of disclosure of sensitive private data is higher via publication of all the attributes. In this paper, we propose two approaches for minimizing the disclosure risk and preserving the privacy by using systematic clustering algorithm. First approach creates an unequal combination of quasi-identifier and sensitive attribute. Second approach creates an equal combination of quasi-identifier and sensitive attribute. We also evaluate our approach empirically focusing on the information loss and execution time as vital metrics. We illustrate the effectiveness of the proposed approaches by comparing them with the existing clustering algorithms.