Quantifying information leakage of randomized protocols

Abstract. The quantification of information leakage provides a quan-titative evaluation of the security of a system. We propose the usage of Markovian processes to model and analyze the information leakage of deterministic and probabilistic systems. We show that this method generalizes the lattice of information approach and is a natural framework for modeling refined attackers capable to observe the internal behavior of the system. We also use our method to obtain an algorithm for the computation of channel capacity from our Markovian models. Finally, we show how to use the method to analyze timed and non-timed attacks on the Onion Routing protocol.