A FCA framework for inference control in data integration systems

Specifying a global access control policy in a data integration system using traditional methods does not necessarily offer a sound and efficient solution to deal with the inference problem. This is because data dependencies (between distributed data sets) are not taken into account when local policies are defined. In this paper, we propose a methodology, together with a set of algorithms, that can help to efficiently detect inferences by considering semantic constraints. The proposed approach is based on formal concept analysis (FCA) as a representation framework. Given a set of local policies, an initial global policy and data dependencies, we propose a methodology that allows the security administrator to derive a set of queries that, combined, could disclose sensitive information. We also say that the set of queries constitutes an inference channel. We use FCA theories to identify the illegal queries known as disclosure transactions. Then, we propose a run-time solution for neutralizing all suspicious queries while ensuring a trade-off between data protection and data availability. By combining Prime Number with Lattice theory, we keep traces of the previously executed queries so that inferences are blocked at run-time. We also discuss a set of experiments that we conducted.