Quantifying Privacy in Nuclear Warhead Authentication Protocols

International verification of nuclear warheads is a practical problem in which the protection of secret warhead information is of paramount importance. We propose a measure that would enable a weapon owner to evaluate the privacy of a proposed protocol in a technology-neutral fashion. We show the problem is reducible to natural' and corrective’ learning. The natural learning can be computed without assumptions about the inspector, while the corrective learning accounts for the inspector’s prior knowledge. The natural learning provides the warhead owner a useful lower bound on the information leaked by the proposed protocol. Using numerical examples, we demonstrate that the proposed measure correlates better with the accuracy of a maximum a posteriori probability estimate than alternative measures.