On the Flow of Data, Information, and Time.

We study information flow in a model for data-parallel computing. We show how an extant notion of virtual time can help guarantee information-flow properties. For this purpose, we introduce functions that express dependencies between inputs and outputs at each node in a dataflow graph. Each node may operate over a distinct set of virtual times--so, from a security perspective, it may have its own classification scheme. A coherence criterion ensures that those local dependencies yield global properties.