Visible to the public Cyber-Physical Emulation and Optimization of Worst-Case Cyber Attacks on the Power Grid

TitleCyber-Physical Emulation and Optimization of Worst-Case Cyber Attacks on the Power Grid
Publication TypeConference Paper
Year of Publication2019
AuthorsCastillo, Anya, Arguello, Bryan, Cruz, Gerardo, Swiler, Laura
Conference Name2019 Resilience Week (RWS)
Keywordsbulk electric system, compositionality, control engineering computing, CPS Resilience, CrashOverride, CrashOverride malware, CrashOverride malware attack, Cyber Dependencies, cyber experiments, cyber interdiction optimization, cyber physical systems, cyber-physical emulation, Cyber-physical systems, DC optimal power flow, DC power transmission, design of experiments, human factors, interdiction optimization, load control, Load flow, Metrics, notional electric system, optimisation, power control, power control area, power engineering computing, power grid operations, power grids, power system control, power system security, power system simulation, pubcrawl, Remote Terminal Units, Resiliency, SCADA systems, SCADA/ICS controls, Scalability, security of data, substation controlling, Substations, worst-case cyber attacks, worst-case loss

In this paper we report preliminary results from the novel coupling of cyber-physical emulation and interdiction optimization to better understand the impact of a CrashOverride malware attack on a notional electric system. We conduct cyber experiments where CrashOverride issues commands to remote terminal units (RTUs) that are controlling substations within a power control area. We identify worst-case loss of load outcomes with cyber interdiction optimization; the proposed approach is a bilevel formulation that incorporates RTU mappings to controllable loads, transmission lines, and generators in the upper-level (attacker model), and a DC optimal power flow (DCOPF) in the lower-level (defender model). Overall, our preliminary results indicate that the interdiction optimization can guide the design of experiments instead of performing a "full factorial" approach. Likewise, for systems where there are important dependencies between SCADA/ICS controls and power grid operations, the cyber-physical emulations should drive improved parameterization and surrogate models that are applied in scalable optimization techniques.

Citation Keycastillo_cyber-physical_2019