A Framework for Cyber-Physical Model Creation and Evaluation

Abstract

In power systems, a cyber-physical model can play a significant role in contingency ranking to assist operators with preventive plans for cyber-related contingencies by identifying the most significant ones. Diverse cyber-physical models based on attack trees and graphs, fault trees, Markov state-space etc. have been proposed and are being developed by researches depending on specific objective. However, prior to the deployment of the models in real world, it is essential to evaluate the performance based on their computational bottlenecks, scalability and accuracy. This paper thus introduces a software-based model comparison framework that allows researchers to improve their models and also evaluate new models against existing ones. Additionaly, we present the algorithms of two cyber-physical modeling engines targeted for contingency and critical assets ranking; based on Attack Graph Analysis (AGA) and Markov Decision Process (MDP) and compare their performance. The models are evaluated for three different use cases: IEEE-24, CyPSA 8-substation, and IEEE-300 systems on cyber-physical model parameters such as MDP size, computation time of generation, number of attack paths, etc. This framework will not only allow us to design and validate models but also provide a platform to researches worldwide to test new models. Further an application is developed for visualization with one-line diagram and ranking of contingencies and critical assets.