Use EMO to protect sensitive knowledge in association rule mining by adding items

Abstract

When data is released or shared among different organizations, some sensitive or confidential information may be subject to be exposed by using data mining tools. Thus, a question arises: how can we protect sensitive knowledge while allowing other parties to extract the knowledge behind the shared data. In this paper, we address the problem of privacy preserving in association rule mining from the perspective of multi-objective optimization. A sensitive rule can be hidden by adding items into the dataset to make the support of the antecedent part of the sensitive rule increase and accordingly the confidence of the sensitive rule decrease. The evolutionary multi-objective optimization (EMO) algorithm is utilized to find suitable transactions (or tuples) to be modified so as the side effects to be minimized. Experiments on real datasets demonstrated the effectiveness of the proposed method.