According to one of the widely accepted definitions, cyber-physical systems (CPS) are engineered systems where functionality emerges from the networked interaction of computational and physical processes. Complex CPS abound in modern society and it is not surprising that many of these systems are safety and mission critical that makes them a target for attacks. Even under normal conditions, CPS face complex issues crosscutting many disciplines with significant implications on essential system functions. Adding cyber-attacks in all their insidious variety creates a massive challenge that cannot be neglected due to the potential consequences.

Because of its significance, security and resilience have attracted considerable attention in many CPS application domains. Because of the heterogeneity and complexity, methodologies that improve CPS security are very diverse with different objectives, specifications, and constraints resulting in a broad body of knowledge. Research efforts are starting to use scientific methods and results to shape technology, practice, and policy in protecting systems from attackers, detecting intrusions, and recovering from compromises. Beyond the complex structure and interactions, security and resilience properties emerge from complex interrelationships, they are not explained by understanding the individual elements of the system, and are highly dynamic in response to changing environment and circumstances.

The objective in this NSF Frontier project is to develop the foundations of secure and resilient CPS. The Vanderbilt team, in particular, focused on the following research areas: (1) Robust monitoring, diagnosis, and networked control, (2) Learning and control for resilience, (3) Threat assessment and diagnostics, and (4) System-security co-design. The developed approaches build upon our strengths on model-based design and system and game theory. The overall objective has been to develop integrated solutions that increase our understanding of complex interrelationships in CPS and allow design of defense strategies that improve security and resilience.

The team accomplished the following key results in the project:

1.  Developed and evaluated a game theoretic framework for attack and anomaly detection in CPS. The novelty of the approach lies on incorporating dynamical models of CPS including invariant constraints such as energy and traffic flow as well as data driven component models. The results have been demonstrated using case studies of water distribution systems, process control systems, and transportation networks.

2.  Developed resilient distributed control and coordination algorithms in the presence of malicious agents.  Innovations in this area include the use of trusted nodes for improving resilience of consensus algorithms multi-agent systems and resilient distributed diffusion for multi-task state estimation that can be used for tracking multiple targets.

3.  Developed learning and control methods to improve resilience in CPS that include methods for sensor selection and placement in adversarial setting in the context of transportation networks and water distribution systems, decentralized resilient traffic light control, stochastic message authentication for integrity assurance of resource constraint systems, and integrative methods that combine redundancy, diversity, and hardening for improving security and resilience.

4.  Develop languages and tools for threat and attack modeling in CPS in order to understand the impact of cyber attacks to system performance and operation. The tools allow the representation of novel attack models and the characterization of the actions of that are needed to compromise the operation of the CPS.  The threat and attack models developed were used for vulnerability analysis of transportation networks and power systems.

5.  Developed a suite of methods for system-security co-design using a model-based framework where the software and system implementation task is to derive the code and models such that they satisfy functional, safety, and security as well as the resource constraint requirements imposed by the underlying platform.

6.  Developed simulation and hardware-in-the-loop testbeds and conducted comprehensive evaluation of both system and security properties. The testbeds include a modeling and simulation integration platform that enables evaluation of resilience of transportation networks in the presence of cyber attacks based on attacker-defender games using simulations of sufficient fidelity and a hardware-in-the-loop which connects physics simulators with networked embedded computers for performing analysis of realistic cyber-attack effects in networked CPS.

7.  Broadening participation: The main activity for broadening participation in computing is the organization of summer camps on CPS and CPS security for high school students. The camps use RoboScape, a team-based, hands-on curriculum for teaching cybersecurity, distributed programming, and robotics. We continue offering the camps in Summer 2019 after the completion of the project. More importantly, we are offering a teacher camp in order to involve middle and high school teachers and scale up our outreach and recruitment of women and underrepresented minorities.      

Last Modified: 05/30/2019
Modified by: Xenofon Koutsoukos