A constraint-based framework for computing privacy preserving OLAP aggregations on data cubes

A constraint-based framework for computing privacy preserving OLAP aggregations on data cubes is proposed and experimentally assessed in this paper. Our framework introduces a novel privacy OLAP notion, which, following consolidated paradigms of OLAP research, looks at the privacy of aggregate patterns defined on multidimensional ranges rather than the privacy of individual tuples/data-cells, like similar efforts in privacy preserving database and data-cube research. To this end, we devise a threshold-based method that aims at simultaneously accomplishing the so-called privacy constraint, which inferiorly bounds the inference error, and the so-called accuracy constraint, which superiorly bounds the query error, on OLAP aggregations of the target data cube, following a best-effort approach. Finally, we complete our main theoretical contribution by means of an experimental evaluation and analysis of the effectiveness of our proposed framework on synthetic, benchmark and real-life data cubes.