Validating Detection Probabilities for the ASSESS Insider Database

Evaluating protection effectiveness against potential threats requires an extensive amount of information about possible adversary actions and safeguards detection probabilities. To aid this process, the Insider module of ASSESS (Analytic System and Software for Evaluating Safeguards and Security) contains an extensive database of possible insider actions involving theft of special nuclear material, and detection probabilities for safeguards components against those actions. To enhance the credibility of evaluation results, we've designed a program for validating and testing this generic database. The Insider database has been carefully structured so that each detection probability corresponds to a well-defined event. In this paper, we describe how we use these detailed event descriptions for structuring and conducting an effective performance testing program. We discuss how to: (1) determine the type of validation/testing procedures that are appropriate for each probability assignment; (2) define a reasonable level of precision for the probability estimates; (3) combine test results and expert opinions; and (4) address the relationship between similar detection situations. These considerations provide the basis for developing a practical safeguards performance validation/ testing program with minimal impact on a facility's operations and personnel. We conclude with a discussion about designing facility-specific testing programs to take advantage of the ongoing ASSESS Insider database validation and testing effort.