Privacy-preserving Classification Mining Based on Probability Theory
Xiaohong Su · Jisuanji gongcheng · 2012
In the existed privacy-preserving classification mining methods based on data perturbation,the privacy data is not protected perfectly because the perturbed data and the original data have been related.The classification algorithm and the data perturbation algorithm have high coupling It is not easy to use these methods in practice.To solve these problems,it proposes a privacy-preserving classification mining algorithm based on probability theory.The perturbed data is independent from the original data and they have the same distribution.This proposed method overcomes the shortcomings of others.The perturbed data is no relation with the original data and the classification methods can be used on the perturbed data directly.