Sampled Data Driven Attack Detection and Adaptation for Security in Control Systems

点击翻译按钮获取中文摘要

This project aims to contribute to cyber-attack resilient control methods and architectures, which are critically important to safe operation of power grid, medical, manufacturing, traffic and many other control systems. Malicious attacks on modern time cyber infrastructure have been a daily phenomenon and a constant threat for the nation's security and well-being. There have been numerous cyber-attacks that infiltrated critical industrial control systems, vehicle control units, GPS systems used for navigation and control, air-vehicle autopilots, SCADA systems in the power grid, etc. In all of these cyber-attacks the aim is to deceive the control and monitoring mechanism and potentially lead the system to instability and malfunction, thus causing physical damages that could be of catastrophic nature. As control systems are becoming sophisticated and frequently include components from cyberspace, their security is generating greater challenges. The project focuses on theoretical advancements in resilient control methods as well as experimental safety demonstrations with UAV systems. It also promotes the engagement of undergraduate students with the topic of safe control system analysis and design. This project aims to address cyber-security for control systems from a control theory perspective, in order to develop controller analysis and synthesis methods applicable for secure software and hardware architectures in order to accurately assess and reduce the risks of such attacks. In practice controllers are implemented digitally and physical plants evolve in a continuous-time domain. The sampled-data control implementation generates additional vulnerability to stealthy attacks, which, if ignored, can result in a catastrophic event. Security associated with the attacks of interest cannot be dealt with solely by robust controller design nor by standard security software. Instead this project takes the approach of co-designing the control algorithm along with the secure software/hardware platform on which the controller operates, all in a sampled-data framework. The proposed schemes and theoretical results will be verified experimentally using unmanned aerial vehicles.