Cyber-physical Systems


Security and Resilience Using a Control Systems Perspective

The comprehensive integration of instrumentation, communication, and control into physical systems has led to the study of Cyber-Physical Systems (CPSs), a field that has garnered increased attention. A key concern that is ubiquitous in CPS is a need to ensure security in the face of cyber attacks. We have carried out a survey of systems and control methods that have been proposed for the security of CPS. We classify these methods into three categories based on the type of defense proposed against the cyberattacks: prevention, resilience, and detection & isolation. A unified threat assessment metric is proposed in order to evaluate how CPS security is achieved in each of these three cases. Also surveyed are the risk assessment tools and the effect of network topology on CPS security.

Machine Learning for Optimal Delay-assignment

Resilience to Cyber-attacks using Deep Neural Networks (DNN)

Given the strong presence of communication networks and networked controllers in cloud-based cyber-physical Systems, there is a strong need to co-design communication and control. Of specific interest is the design of delays that often arise when communication occurs over shared resources. In order to ensure an optimal control design in the presence of such delays, not only is it useful to have a delay-aware controller but a method by which optimal assignment of delays can be imposed on the communication network links. A project in our lab concerns the design of a delay-aware feedback controller that judiciously accommodates the most recent state information and a machine learning (ML) based method for determining the optimal delay assignment. This ML method consists of an offline training of a neural network whose inputs are a set of selected delays and outputs are relevant performance-optimizing metrics. The resulting neural network is shown to be capable of learning the optimal delay assignment to the various links in the communication network and therefore yielding optimal performance. The proposed method is validated using a power system case study of an IEEE 68-bus, and shown to result in a notable performance improvement where in 88% of the cases a near-optimal performance can be realized.