A comprehensive theoretical framework for privacy preserving distributed OLAP

Abstract

This paper complements the privacy preserving distributed OLAP framework proposed by us in a previous work by introducing four major theoretical properties that extend models and algorithms presented in the previous work, where the experimental validation of the framework has also been reported. Particularly, our framework makes use of the CUR matrix decomposition technique as the elementary component for computing privacy preserving two-dimensional OLAP views effectively and efficiently. Here, we investigate theoretical properties of the CUR decomposition method, and identify four theoretical extensions of this method, which, according to our vision, may result in benefits for a wide spectrum of aspects in the context of privacy preserving distributed OLAP, such as privacy preserving knowledge fruition schemes and query optimization. In addition to this, we also provide a widespread experimental analysis of the framework, which fully confirms to us the major practical achievements, in terms of both efficacy and efficiency, due to our framework.