A Link-Based Method for Hiding Sensitive Patterns on Data Mining

DOI: 10.5176/978-981-08-6308-1_40

Authors: Yu-Chiang Li, Chien-Ting Huang

Abstract:

Shared data has become a trend among businesses to increase mutual benefit for all participants. However, sharing data also increases the risk of unexpected information leaks, so it is important that an organization develops an efficient and effective method restricting information leaks to maintain its competitive edge. The direct release of databases for query is particularly vulnerable to sensitive information leaks. The challenge is to modify the data in order to protect the sensitive individual information while maintaining the data mining capability. This article proposes a link-based model for effective association rule hiding that reduces the unexpected loss of association rules, which is called the side effect or “misses cost.” Experimental results reveal that the proposed link-based model is effective and can generally achieve a significantly lower misses cost than the Algo2b and MaxFIA methods do in test datasets.