Transaction Database Modification Approaches for Achieving Privacy Preserving in High Utility Itemset Mining

Privacy preserving data mining has emerged as an important issue in the data mining area because there is an array of algorithms proposed to find hidden valuable information from databases and privacy information can be included in information founded by these algorithms. High utility itemset mining is one of approaches in association rule mining, which finds itemsets with utilities no less than a given minimum utility threshold. Various researches have been dedicated to developing privacy preserving utility mining algorithms for hiding privacy information from high utility itemset mining algorithms. In this paper, we survey state-of-the-art algorithms for privacy preserving utility mining. Three algorithms are introduced in this paper. In order to hide sensitive information in a given transaction database, they modify several transactions by reducing internal utilities so that sensitive itemsets have utilities less than a threshold. Since modifications of transactions cause some side effects in databases, minimizing side effects is an important factor. Therefore, we evaluate their performances in terms of side effects as well as runtimes.