PRIVACY PRESERVING FUZZY ASSOCIATION RULES HIDING IN QUANTITATIVE DATA

Abstract

Data mining is the process of extracting useful patterns or knowledge from large databases. With the advancement in data mining technologies, the protection of confidentiality of sensitive information in databases has become a critical issue to be resolved. For example, association rule analysis is a popular tool for discovering useful associations from large amount of data and some useful hidden information could be easily discovered using this kind of tool. Therefore, there is a need to investigate data mining algorithm from a new point of view that is of personal privacy. Most of the conventional data mining algorithms have been developed to work on transactions with binary values. However, transactions with quantitative values are commonly found in real world applications. Therefore, there is a need to develop some algorithms for hiding useful fuzzy association rules in quantitative data. In this dissertation work, we propose three algorithms that are used for hiding useful fuzzy association rules in quantitative data.' The proposed algorithms integrate the fuzzy set concepts and Apriori mining algorithm to find useful fuzzy association rules in quantitative data and then hide them using privacy preserving technique. For hiding purpose, three different methods have been used. In first method, we decrease the support of the rule to be hidden by decreasing the support value of item in either Left Hand Side (L.H.S.) or Right Hand Side (R.H.S.) of the rule. In second method, we decrease the confidence of the rule to be hidden by increasing the support of L.H.S. of the rule only. In last method, we decrease the confidence of the rule to be hidden by decreasing the support value of item in R.H.S. of the rule only. The implementation of proposed algorithms is done in Java using Gel editor. Experimental results show that the proposed algorithms have low hiding failure and generate minimal side effects and thus, maintain higher data quality of the released database.