Quantitative analysis of leakage for multi-threaded programs

Abstract

We present a quantitative analysis of information flow for a multi-threaded language based on a probabilistic scheduler. The analysis consists of two steps. First, multi-threaded programs are translated into single-thread looping programs with a probabilistic operator. Then an information theoretical semantics of loops with probabilistic operators is used to derive the leakage. Using this analysis classical examples of multi-threaded programs are revisited: it is shown how the analysis is able to deal with, among other, probabilistic leakage, internally observable timing leakage and leakage originated by observing intermediate states of computation