An Algorithm for Hiding Sensitive Frequent Itemsets

Association rule mining is an important data-mining technique that finds interesting association among a large set of data items. Since it may disclose patterns and various kinds of sensitive knowledge that are difficult to find otherwise, it may pose a threat to the privacy of discovered confidential information. This study investigates how to shelter certain information and/or confidential knowledge in the data set and how to create a new database for nonconfidential access. The proposed approach uses the data distortion technique. In this connection, sensitive representative rules are mined based an algorithm named GSRR. Then, in immunization phase, an algorithm named EDSR is presented. In this algorithm, the procedure of hiding the sensitive itemsets is carried out through the reduction of sensitive representative rules confidence rate. Regarding this, the changes occur on the right hand side items of the rules. These changes occur on transactions which fully support sensitive representative rules, and among the transactions, a transaction is selected for the change which has the fewest number of items. The goal is having the minimum change on database. Performance comparison of the recommended algorithm and the two benchmark algorithms on the dense database of Chess, illustrated that the proposed algorithm run time has considerably decreased in comparison with the benchmark algorithms. Also, regarding the number of lost rules, the algorithm is more practical than the benchmark algorithms.