Noninterference and the Most Powerful Probabilistic Adversary

Probabilistic noninterference extends the classical possibilistic notion introduced by Goguen and Meseguer in order to capture the information leakage caused by adversaries that set up probabilistic covert channels. In this setting we investigate how to evaluate the observational power of an adversary to the purpose of establishing the maximal security degree of a given system. We introduce three classes of probabilistic adversaries, which represent the different observational power of an adversary, and then we establish properties for each such classes which state the complexity of effectively computing the most powerful adversary.