Information Theoretic Metric Of Verification System Performance

Year
2020
Author(s)
Jason Reinhardt - Sandia National Laboratories, Livermore
Michael Hamel - Sandia National Laboratories, Livermore
Ben Bonin - Sandia National Laboratories, Livermore
Eva Uribe - Sandia National Laboratories, Livermore
Abstract

There remain several challenges to verifying potential future nuclear arms control agreements. These include, among others, protecting sensitive information during verification activities, confirming the authenticity of an accountable item, and confirming the disposition of materials. Technology-specific approaches offer potential solutions to these challenges but understanding and comparing the performance of disparate technologies in the context of a verification regime requires metrics and assessment methods that are technology and regime agnostic. This work attempts to create an objective metric that can quantify confidence and be used as a comparison tool for different information barriers. For a hypothetical arms-control scenario, we used varying information barriers that only allow certain information taken from a gamma-ray spectrum to be “seen” by the monitor. A Bayesian belief network was created that included the monitor’s prior and updates their belief as more information became available through measurement. We use the Kullback-Leibler Divergence as a measure of the information gained by the inspector about a protected fact, given the observation of the verification measurement. The result of this analysis is an objective measure of confidence for the monitoring party, and for information protection for the host. This paper will describe the theory and framework used for this analysis.