Knowledge Hiding in Data Mining by Transaction Adding and Removing

A new approach by transaction adding and removing (TAR) is presented for association rule hiding. Only a few transactions need updating to keep the original features in the mined dataset. Firstly, two definitions of weak associated transaction (WAT) and strong associated transaction (SAT) are defined. Then, the TAR approach and algorithm are stated in detail with two main processes of WAT adding and SAT removing. A kind of WAT modifying approach is described and implemented to avoid the transaction duplication in the database. Furthermore, a modification factor is created to control the updating number of transactions to the database. When the modification factor is set above 0.05, the hiding rate can be reached to 100%, and the side effects of the lost rules and new created rules are very small with rate less than 3%. The robustness to the support attacking is satisfying with suitable hiding rate.