Knowledge transfer and information leakage in protocols

Abstract

A protocol defines a structured conversation aimed at exchanging information between two or more parties. Complete confidentiality is virtually impossible so long as useful information needs to be transmitted. A more useful approach is to quantify the amount of information that is leaked. Traditionally, information flow in protocols has been analyzed using notions of entropy. We move to a discrete approach where information is measured in terms of propositional facts. We consider protocols involving agents holding numbered cards who exchange information to discover each others’ private hands. We define a transition system that searches the space of all possible announcement sequences made by such a set of agents and tries to identify a subset of announcements that constitutes an informative yet safe protocol.