Hide Association Rules with Fewer Side Effects

Peng CHENG  Ivan LEE  Jeng-Shyang PAN  Chun-Wei LIN  John F. RODDICK  

Publication
IEICE TRANSACTIONS on Information and Systems   Vol.E98-D   No.10   pp.1788-1798
Publication Date: 2015/10/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.2014EDP7345
Type of Manuscript: PAPER
Category: Artificial Intelligence, Data Mining
Keyword: 
Association rule hiding,  side effects,  border rules,  

Full Text: PDF(461KB)>>
Buy this Article




Summary: 
Association rule mining is a powerful data mining tool, and it can be used to discover unknown patterns from large volumes of data. However, people often have to face the risk of disclosing sensitive information when data is shared with different organizations. The association rule mining techniques may be improperly used to find sensitive patterns which the owner is unwilling to disclose. One of the great challenges in association rule mining is how to protect the confidentiality of sensitive patterns when data is released. Association rule hiding refers to sanitize a database so that certain sensitive association rules cannot be mined out in the released database. In this study, we proposed a new method which hides sensitive rules by removing some items in a database to reduce the support or confidence levels of sensitive rules below specified thresholds. Based on the information of positive border rules and negative border rules contained in transactions, the proposed method chooses suitable candidates for modification aimed at reducing the side effects and the data distortion degree. Comparative experiments on real datasets and synthetic datasets demonstrate that the proposed method can hide sensitive rules with much fewer side effects and database modifications.