Belief Propagation during Data Integration in a P2P Network

We examine properties of a peer to peer network comprising several agents that store various types of local data and exchange them through established communication channels. We propose a communication model applicable to a developed platform for data integration between various security agencies and we focus on analysis of consequences of established channels, e. g. an unintended information leakage or a presence of data silos that can be an impediment for cooperation. To detect such situations efficiently, we do not concentrate on exchanged data itself, but on a belief related to known classes of data. In the analyses we use a model, in which communications and belief states are expressed as matrix operations of linear algebra. We show that applying this model we can efficiently reason about the data that can potentially be exchanged between agents not linked directly and about the ranges, which can be reached by the data during communication flows.