A game theoretic approach to nuclear safeguards selection and optimization
This work presents a computational tool that calculates optimally efficient safeguarding strategies at and across nuclear fuel cycle facilities for a cost-constrained inspector seeking to detect a state-facilitated diversion or misuse. The tool employs a novel methodology coupling a game theoretic solver with a probabilistic simulation model of a gas centrifuge enrichment plant and an aqueous reprocessing facility. The simulation model features a suite of defender options at both facilities, based on current IAEA practices, and an analogous menu of attacker proliferation pathway options. The simulation model informs the game theoretic solver by calculating the detection probability for a given inspector-proliferator strategy pair and weighting the detection probability by the quantity and quality of material obtained to generate a scenario payoff. Using a modified fictitious play algorithm, the game iteratively calls the simulation model until the equilibrium is reached and outputs the optimal inspection strategy, proliferation strategy, and the equilibrium scenario payoff. Two types of attackers are modeled: a breakout-willing attacker, whose behavior is driven by desire for high value material; and a risk-averse attacker, who desires high-value material but will not pursue a breakout strategy that leads to certain detection. Results are presented demonstrating the sensitivity of defender strategy to budget and attacker characteristics, for an attacker known to be targeting the enrichment or reprocessing facility alone, as well as an attacker who might target either facility. The model results indicate that the optimal defender resource allocation strategy across multiple facilities hardens both facilities equitably, such that both facilities are equally unattractive targets to the attacker.