Multidimensional Data Visualization Using Aggregation Method of Differential Privacy Equipartition k-means
Guangqiang Xie · Journal of Chinese Computer Systems · 2013
Privacy preserving data mining developed rapidly in recent years,on the other hand,there is a dearth of research on privacy preserving data visualization,w hich have w ide range of applications.Differential privacy is a new promising privacy-preserving paradigm,in fact,w e are not aw are of any existing multidimensional data visualization method under differential privacy.In this paper,w e study how to preserve priavcy in the process of data visualization.existing DP k-means algorithm is mainly of theoretical interest because it doesn't w ork at large k w hich is necessary in data aggregation.Motivated by this,w e propose e-Differential Privacy Equipartition k-means(DPE k-means),a method w hich w ork better at large k.w e find it eliminate a majority of data overlapping,greatly improve the visualization image quality under a certain privacy level.Our experiments show that at the same e,DPE k-means gets a much higher aggregation quality level than existing DP k-means method.