Integrated Security and Safety Engineering for Networked Control Systems
Motivation
There is a recent trend to connect controlled process to "open" public networks. This causes a big security issue since unauthorized access ("cyber attacks") is now possible in control systems. Consequently, because some of this controlled process are safety-critical, the study and design of secure and safety control system become of high importance.
The focus of this project is firstly to investigate how cyber attacks manipulate control algorithm and finally also the safety-critical physical process; and secondly to develop new methodologies, from a control perspective, in order to tolerate and mitigate both, the cyber attack and the fault, effects on the system. Specifically, the following main research topics will be investigated in this project:
- Requirement and countermeasure analysis for security and fault-tolerance.
- Detection and isolation of attacks and faults.
- Design of cyber attack- and fault-tolerant control systems.
Local control systems
As a first approach, single networked control loops are considered as the class of systems to be studied. Therefore, the feedback control loop consists on a controller and the process under control. The commands of the controller are applied through the actuators to the process, while the process information needed for the controller is provided by the sensors. Both, the control commands and the measurement of the sensors, are gathered by a diagnosis system in order to determine whether a fault or an attack is present in the control loop (fault/attack detection). And if so is the case, a further step is to determine which fault or attack is present (fault/attack isolation).
Here, we assume that the control loop is networked and somehow connected to an external communication network such as Internet. For the sake of the study, two different kind of networked systems are considered:
- Systems in which control data (i.e., controller commands and sensor measurements) is exchanged by means of a communication network.
- Systems in which the actuators and the sensors are physically wired to the controller (no communication network) but the controller is connected to a network for commissioning and maintenance purposes (see video below).
Distributed control systems
Usually the fault/attack diagnosis and fault/attack tolerant control techniques valid for local feedback control loops are not suitable for large-scale system where the size of the data and the complexity of the computations rapidly increase. In addition, large-scale systems may also involve several control loops with several controllers sharing data among them through several communications networks. To deal with this type of systems, a commonly approach is to design distributed diagnosis systems that work locally but perform globally since they also share information among them.
The study of distributed diagnosis systems is rather new in the research community where a mature machinery equipped with methods and techniques is still lacking. Consequently, the project also aims to contribute in the area of distributed fault/attack diagnosis for large-scale systems.
Researchers: Dr. Yumei Li, Dr. Benjamin Gerard, Dr. Radu State, Dr. Jérôme François, Prof. Dr. Holger Voos, Prof. Dr. Thomas Engel, Prof. Dr. Mohamed Darouach.
