Privacy preserving and data mining in an on-line statistical database of additive type

Abstract

In an on-line statistical database, the query-answering system should prevent answers to statistical queries from leading to disclosure of confidential data. On the other hand, a statistical user is inclined to data mining, that is, to disclose pieces of information that are implicit in the (explicit) answers to his queries. A key task for both is to find data that is derivable from given summary statistics. We show that this task is easy if data is additive and the set of given summary statistics can be modelled by a graph. © Springer-Verlag 2004.