A Note on Noninterference in the Presence of Colluding Adversaries ( Preliminary Report )

Whether adversaries can glean information from a distributed system in a formal sense hinges on the definition of such a system and what can be observed by those agents. In the presence of colluding adversaries, the standard definition of noninterference by Goguen and Meseguer and its many variants proposed in the literature fail in a very intuitive sense to capture a simple collusion attack. The crucial difference between what is modelled in those definitions and what we argue needs to be modelled is that teams can observe pomsets as Plotkin and Pratt stated. In this note we expose what goes wrong in the known approaches and explain how to fix the problem.