A theoretically-sound accuracy/privacy-constrained framework for computing privacy preserving data cubes in OLAP environments

State-of-the-art privacy preserving OLAP approaches lack of strong theoretical bases that provide solid foundations to them. In other words, there is not a theory underlying such approaches, but rather, an algorithmic vision of the problem. A class of methods that clearly confirm to us the trend above is represented by the so-called perturbation-based techniques, which propose to alter the target data cube cell-by-cell to gain privacy preserving query processing. This approach exposes us to clear limits, whose lack of extendibility and scalability are only the tip of an enormous iceberg. With the aim of fulfilling this critical drawback, in this paper we propose and experimentally assess a theoretically-sound accuracy/privacy-constrained framework for computing privacy preserving data cubes in OLAP environments. The benefits deriving from our proposed framework are two-fold. First, we provide and meaningfully exploit solid theoretical foundations to the privacy preserving OLAP problem that pursue the idea of obtaining privacy preserving data cubes via balancing accuracy and privacy of cubes by means of flexible sampling methods. Second, we ensure the efficiency and the scalability of the proposed approach, as confirmed to us by our experimental results, thanks to the idea of leaving the algorithmic vision of the privacy preserving OLAP problem. © 2012 Springer-Verlag.