Privacy level evaluation of differential privacy for time series based on filtering theory

Abstract

The current differential privacy preserving methods on correlated time series were not designed by protecting against a specific attack model,and the privacy level of them couldn’t be measured.Therefore,an attack model was put forward to solve the above problems.Since the noise series added by these methods was independent and identically distributed,and the time series could be seen as a short-time stationary process,a linear filter was designed based on filtering theory,in order to filter out the noise series.Experimental results show that the proposed attack model is valid,and can work as a unified measurement for these methods.

Keywords

• differential privacy;privacy preserving;correlated time series;attack model