PAVED: Perturbation analysis for verification of energy data

Sensor integrity is arguably the most critical feature to protect in cyber-physical systems. Since power systems are cyber-physical systems with ubiquitous sensors that monitor and protect the grid, data must be trustworthy. Process safety and control decisions ultimately depend on data. The focus of this paper is how to design and apply perturbation based detection for sensor verification, under full AC unobservable false data injection (AU-FDI) attacks, by combining an active probing strategy with cyber-side data based on the cyber-physical situational awareness model CyPSA. A case study on a cyber-physical eight substation model is presented, where we construct an AU-FDI attack and introduce our probing-based detection solution and evaluate it with varying probe signals, values, and locations. Results demonstrate how sensor data in power systems can be systematically authenticated using perturbation-based techniques and how different perturbation types and locations affect the results. The case study then demonstrates the improvements to verification by using both physical and cyber data, as CyPSA provides risk prioritization in the form of authenticity weight measure of the sensors, for enhancing the security of power systems from a cyber-physical point of view.