Minable Data Publication Based on Sensitive Association Rule Hiding

Minable data publication can promote data sharing among commercial companies and further facilitate the development of data-driven services. However, these commercial companies are often reluctant to publish their data due to security concerns. The published data may contain some sensitive information that is minable by malicious entities, leading to data privacy leakage. Therefore, it is highly demanded to develop the technologies supporting minable data publication with privacy protection. In this paper, we propose a p rivacy-preserved m inable d ata p ublication scheme (PMDP). PMDP enables selective sensitive association rules hiding while supporting the association rule mining. In PMDP, how to balance the trade-off between data privacy and data utility is the major problem, which can be formulated as a multi-objective optimization problem. To tackle this multi-objective optimization problem, we develop a customized multi-objective evolutionary algorithm (MOEA). In the customized MOEA, the local optimum trapping issue and slow convergence speed issue are hard to be addressed. First, to avoid being trapped into the local optimum, we carefully design a novel mutation method to guarantee the diversity of solutions. Second, to accelerate the convergence speed, we present a preprocessing method before the evolution process of the MOEA. In addition, we introduce the elite learning strategy into the MOEA, so the convergence speed can be further accelerated. At last, experiments are conducted over several datasets to demonstrate the effectiveness of PMDP.