Research looks to verify power networks at the device level
Securing our power grids from malicious attacks is critical to ensuring a safe power grid, and many researchers are working on solving this problem by focusing on the entire system. However, TCIPG alum and ITI research scientist Katherine Davis is tackling the problem in a new and innovative way in her new $391,014, three-year NSF grant to study verification of cyber-physical critical infrastructures at the device level.
“We want to verify controllers in power systems that could allow unsafe operations, and we want to be able to prevent the programmable logic controllers from taking actions that could affect the safety and reliability of the power grid,” Davis said.
Davis will be working with co-PI (and TCIPG alum) Saman Zonouz of Rutgers University to use power modeling and power engineering to help determine what constitutes a safe decision and how to improve enforcement of safety and security requirements at the programmable logic controller (PLC) level. PLCs are widely used for control automation in industrial control systems, and they are typically connected to an engineering workstation where engineers develop control logic to process and issue commands to the system.
Davis will be focusing on the power systems portion of the project to improve the decisions made in power systems models, while Zonouz, who specializes in cyber-physical systems, will be developing a “look-ahead” graph model that looks at the state the controller is in and the next steps it can take.
“One of the new things we’re bringing to the table is power information in the ‘look-ahead’ graph,” Davis said. “We’ll be looking at the code and commands and our solution is an in-between step that sees what is sent to the controller and makes sure it won’t cause anything harmful to happen.”
Davis and Zonouz will be working at a very low level with the devices and getting into the machine-level code of the PLCs.
Davis anticipates that it will be challenging to determine how to use distributed sensor data to construct power systems models that can be used in low-level code, which will be the focus of her work under this grant.
“We’re looking at how we’re going to incorporate the power system model and power system data into the enforcement of safety and security in the ‘look-ahead’ model and at such a low-level computing platform,” she said. “We’re trying to strike a balance between the amount of information we can consider sufficient to determine whether a power system is safe, and the constraints on available information that are imposed by this minimal computing platform.”
Davis and Zonouz will be building out part of their model and working on real hardware, which will make it possible to scale their solution quickly and easily and to work in real-time. The end goal is to create a module that will sit in the PLC, or somewhere between the operator and the PLC, and intercept and analyze all communications to the PLC to determine whether they are harmful.