Integrations of physical processes with computers and process control, or cyber-physical systems (CPS), are everywhere.

However, the widespread application of CPS also makes them prime targets for hackers. Just one simple change in the value of a sensor can wreak havoc. Vulnerability to malicious attacks has created the need for systems that can withstand the corruption of sensors and still provide safe and efficient process control.

That is why a new method is in development for resilient state estimation (RSE) for systems that are under attack.

State estimation refers to the use of external variables, i.e., sensor readings, to determine the internal state of the system using mathematical models called “observers.” This is a critical step in process control. When the internal state of a system can end up determined despite the corruption of sensors, it is called RSE.

“Would you ride an autonomous vehicle or live near a computer-controlled power plant if safety and security were not considered in their design? The importance of resiliency in control systems has been recognized for over a decade,” said Professor Yongsoon Eun from Daegu Gyeongbuk Institute of Science and Technology, who conducted a study on the subject with his colleague from Hyundai Motor Company, Yechan Jeong.

Schneider Bold

All control systems are subject to variations or “disturbances” in the process, which cause errors in state estimation. However, as the disturbance increases, so does the error, leading to a breakdown in system resiliency. Making use of a kind of observer known as “Unknown Input Observer (UIO),” the new RSE method overcomes this limitation and provides a way for state estimation that can withstand malicious attacks as well as external disturbances.

In this method, a UIO is designed for each sensor, the estimates from each UIO are combined, and the error is processed to provide the true value of the internal state of the system. The benefit of using a UIO is its estimation error always converges to zero, regardless of external disturbances to the process.

This is unlike other observers, which can only provide a range for estimation error. Another novelty of this method is that it deploys a ‘partial state UIO,’ a technique newly developed by Eun’s team, by which as much partial information on internal states as possible is extracted from each sensor when full state information extraction is not feasible. This greatly expands the applicability of the new RSE method based on UIO.

“The proposed method gives a system a level of tolerance for faults and attacks and, in cases where it is inevitable, allows graceful degradation of system functionality. This makes it crucial to the design of CPS,” Eun said.

Pin It on Pinterest

Share This