Elements of epistemic crypto logic

Representation of ignorance about large numbers --- agent a does not know agent b's key --- is not feasible in standard Kripke semantics. The paper introduces register models that allow for compact representation of such ignorance. This is used to design a sound an complete language for number guessing games. The probabilities generated by our semantics allow for and motivate Monte Carlo model checking for register models. We show that the approach can be extended to a real life setting, namely the analysis of cryptographic security protocols. We look at a well known security protocol for secret key distribution over an insecure network, and point out how this can be analyzed with our modified version of Kripke semantics.