Signature-Based Inference-Usability Confinement for Relational Databases under Functional and Join Dependencies

Abstract

Inference control of queries for relational databases confines the information content and thus the usability of data returned to a client, aiming to keep some pieces of information confidential as specified in a policy, in particular for the sake of privacy. In general, there is a tradeoff between the following factors: on the one hand, the expressiveness offered to administrators to declare a schema, a confidentiality policy and assumptions about a client's a priori knowledge; on the other hand, the computational complexity of a provably confidentiality preserving enforcement mechanism. We propose and investigate a new balanced solution for a widely applicable situation: we admit relational schemas with functional and join dependencies, which are also treated as a priori knowledge, and select-project sentences for policies and queries; we design an efficient signature-based enforcement mechanism that we implement for an Oracle/SQL-system. At declaration time, the inference signatures are compiled from an analysis of all possible crucial inferences, and at run time they are employed like in the field of intrusion detection.