Performance Comparison of PSO and Hybrid PSO-GA in Hiding Fuzzy Sensitive Association Rules

Abstract

Explosion of data analysis techniques facilitate organizations to publish microdata about individuals. While the released data sets provide valuable information to researchers, it is possible to infer sensitive information from the published non-sensitive data using association rule mining. An association rule is characterized as sensitive if its confidence is above disclosure threshold. These sensitive rules should be made uninteresting before releasing the dataset publicly. This is done by modifying the data that support the sensitive rules, so that the confidence of these sensitive rules is reduced below disclosure threshold. The main goal of the proposed system is to hide a set of sensitive association rules by perturbing the quantitative data that contains sensitive knowledge using PSO and hybrid PSO-GA with minimum side effects like lost rules, ghost rules. The performance of PSO and Hybrid PSO-GA approach in effectively hiding fuzzy association rule is also compared. Experimental results demonstrate that hybrid approach is efficient in terms of lost rules, number of modifications, hiding failure.