Privacy Preservation through Random Non-linear Data Distortion

Consider a scenario in which the data owner has some private/sensitive data and wants a data miner to access it for studying “important” patterns without revealing the sensitive information. Privacy preserving data mining aims to solve this problem by randomly transforming (distorting) the data prior to its release. Previous work only considered the case of linear distortions — additive, multiplicative or a combination of both — for studying the usefulness of the distorted output and the privacy preserved. In this paper, we consider a general class of potentially non-linear transformations of the data. We develop bounds on the expected accuracy of our non-linear distortion and also quantify privacy by using standard definitions. We show how our general transformation can be used in practice for two specific problem instances: a linear model and a popular non-linear model viz. neural network. The paper presents a thorough theoretical analysis of the transformation and possible applications. Experiments conducted on real-life datasets demonstrate the effectiveness of the approach.