Increasing polynomial regression complexity for data anonymization

Abstract

Pervasive computing and the increasing networking needs usually demand from publishing data without revealing sensible information. Among several data protection methods proposed in the literature, those based on linear regression are widely used for numerical data. However, no attempts have been made to study the effect of using more complex polynomial regression methods. In this paper, we present PoROP-k, a family of anonymizing methods able to protect a data set using polynomial regressions. We show that PoROP-k not only reduces the loss of information, but it also obtains a better level of protection compared to previous proposals based on linear regressions.