A New Bayesian-Based Method for Privacy-Preserving Data Mining

Abstract

Recently, data mining developed fast and attracted a lot of attention. When using data mining in real world, privacy protection is an important problem. Over the past ten years, many researchers study this problem and propose a lot of PPDM (privacy preserving data mining) methods. These methods can complete data mining task when protecting privacy. This paper gives a new Bayesian-based PPDM method, which is designed for classification. This method is a data perturbation method and is algorithm-independent, which means the perturbed data can be used by normal classification methods directly. Experiments show that comparing with existing methods, this new method perform better for protecting privacy, when they keeping data utility both.