Developing a Database for Helping Evaluate Safeguards Effectiveness Against Insider Adversaries

Lawrence Livermore National Laboratory (LLNL) has been developing assessment models to systematically evaluate the performance of safeguards and security systems against a potential theft attempt of nuclear material by a nonviolent insider adversary. These models are part of a comprehensive safeguards evaluation package called ASSESS which is being developed jointly by LLNL and Sandia National Laboratories. To help \"standardize\" and provide comparable and consistent model results when used by different analysts at different facilities, a predefined database was needed. The insider database we are developing provides a comprehensive list of strategies and specifies probabilities of detection that depend on various target material and safeguards characteristics, adversary attributes, and the particular methods employed to defeat individual safeguards components. The model and database structure attempt to adequately address all important considerations while simplifying user input and running within a reasonable amount of time on a microcomputer. In this paper, we describe the organizational structure underlying the ASSESS insider database. We focus on how the database 1) allows most situations of interest to be tractably yet reasonably modeled and 2) provides the users with a welldefined and consistent assignment of probabilities of detection for a core set of events from which overall theft scenarios may be comprised and then evaluated.