Y. Setiawan “Adversary Path Analysis of a Physical Protection Design Using a Stochastic Approach”, M.S. Thesis, Nuclear Engineering, Texas A&M University, College Station, TX (2018).
The EASI model is a single path analysis model to calculate the Probability of Interruption (PI) of a Physical Protection System (PPS) in a facility. However, the PI¬ value from the EASI model does not have uncertainty value, which is important to represent the confidence level on the performance of a PPS. A Monte Carlo (stochastic) approach to analyze the effectiveness of a PPS, specifically estimating the PI value and uncertainty in the estimation, is implemented into the EASI model approach by developing a new software. The software was tested by analyzing a hypothetical nuclear facility, NARI by estimating the PI values by considering the characteristics of protection elements in the PPS as well as adversary strategies including collusion with an insider. Sensitivity analysis was performed on the characteristics of the protection elements in the PPS in arriving at the distribution of PI values for the Most Vulnerable Path (MVP) considering the Critical Detection Point (CDP). The implementation of Monte Carlo technique successfully produced PI value distribution from which the mean and standard deviation values were estimated. The relationship between probability of detection of protection elements (PD) and PI is found to be linear, while the relationship between delay time of protection elements (td) and PI is found to be non-linear. The PI value was the lowest for the simulations where the insider malicious acts on the protection elements were included. The rushing strategy among the other adversary strategies analyzed produced the lowest PI value for the NARI facility, due to its unbalanced PPS design. This result points out that the CDP approach of analyzing and balancing the PPS may not be appropriate for every facility. In conclusion, the implementation of the Monte Carlo method is found to be valuable in sampling the PD value and tD value distributions otherwise used as single values for these variables in the EASI model approach.rn