A Novel Sanitization Approach for Privacy Preserving Utility Itemset Mining

Abstract

Data mining plays a vital role in today’s information world wherein it has been widely applied in various business organizations. The current trend in business collaboration demands the need to share data or mined results to gain mutual benefit. However it has also raised a potential threat of revealing sensitive information when releasing data. Data sanitization is the process to conceal the sensitive itemsets present in the source database with appropriate modifications and release the modified database. The problem of finding an optimal solution for the sanitization process which minimizes the non-sensitive patterns lost is NP-hard. Recent researches in data sanitization approaches hide the sensitive itemsets by reducing the support of the itemsets which considers only the presence or absence of itemsets. However in real world scenario the transactions contain the purchased quantities of the items with their unit price. Hence it is essential to consider the utility of itemsets in the source database. In order to address this utility mining model was introduced to find high utility itemsets. In this paper, we focus primarily on protecting privacy in utility mining. Here we consider the utility of the itemsets and propose a novel approach for sanitization such that minimal changes are made to the database with minimum number of non-sensitive itemsets removed from the database.