Toward Rigorous Experimentation and Analysis in Cyber-Physical System Emulation

The number of cyber-attacks to critical infrastructure has increased over the last several years. For power systems, concerns about cyber-attacks are even more timely as system designs move toward reliance on distributed energy resources and the inclusion of more "smart" components. Deep understanding of complex systems and potential threats is critical. For that reason, platforms for rigorous scientific experimentation on high-fidelity models, also called emulations, have become important planning and evaluation tools. However, best practice methodology is still an active area of research, and work remains to be done applying existing rigorous analysis tools in emulation-driven frameworks. This paper discusses work to integrate statistical analysis capabilities into an emulation-driven experimentation platform and to use these capabilities to perform an uncertainty quantification analysis. This work modeled a notional power and fuel distribution network with a notional cyber-attack, and studied how the aggressiveness of the attacker changed the extent to which it impacted the system. The authors expected a linear correlation between these factors but found no such correlation. This finding spurred a series of additional sub-analyses, leading to some key insights and lessons learned which would be generally applicable to other similar experimentation approaches.