An Efficient Rule-Hiding Method for Privacy Preserving in Transactional Databases

One of the obstacles in using data mining techniques such as association rules is the risk of leakage of sensitive data after the data is released to the public. Therefore, a trade-off between the data privacy and data mining is of a great importance and must be managed carefully. In this study an efficient algorithm is introduced for preserving the privacy of association rules according to distortion-based method, in which the sensitive association rules are hidden through deletion and reinsertion of items in the database. In this algorithm, in order to reduce the side effects on non-sensitive rules, the item correlation between sensitive and non-sensitive rules is calculated and the item with the minimum influence in non-sensitive rules is selected as the victim item. To reduce the distortion degree on data and preservation of data quality, transactions with highest number of sensitive items are selected for modification. The results show that the proposed algorithm has a better performance in the non-dense real database having less side effects and less data loss compared to its performance in dense real database. Further the results are far better in synthetic databases in compared to real databases