A Generic Framework for Data Analysis in Privacy-Preserving Data Mining

Abstract

Directions of Privacy-preserving data publishing are toward research and applications. Previous studies focus on static data sets and some experiments are on dynamic data sets too. The problem of continuous privacy-preserving publishing of data streams is not solved by too complex approaches. Privacy is achieved by applying security on dynamic data which is a challenging task. We propose a method that extends the scope of existing works with a different framework of building ensemble classifier on time window based samples on the data streams and applying perturbation using Perlin noise on selective tuples as selective perturbation.