Autonomic Cyber-Physical Systems: Resilience to Cyber Attacks
Abstract
Autonomic Cyber-Physical Systems: Resilience to Cyber AttacksCyber-physical systems (CPS) combine sensing, control, and actuation in a continuous feedback loop to design complex systems that have many applications from industrial plants to navy platforms to home automation.Given the close interaction between the cyber- and physical components of such systems, they are vulnerable to wider range of cyber-attacks that can have a deleterious effect on a nation s infrastructure. The key to making a CPS system secure is building vigilance and adaptivity at various layers of the CPS hierarchy. To this end, the researchers propose an autonomic computing approach for networked CPS to implement resilient behavior in the event of cyber attacks.This effort consists of a multi-level architecture that presents a hierarchical sensing and decision loop. The feedback path integrates data from heterogeneous sensors from the bottom layer into local information from logic units and, further, into global state awareness from assessment units at the top. The feedforward path evaluates actionable information and plans a new global configuration, converted to action by effectors or actuators. The goal of the feedforward path is to mitigate the impact of a cyber attack. The mid-level logic units form a resilient group since an attack can target any one of them. At the same time, sensors or other components that disagree with a consensus are identifed as suspect and subjected to closer evaluation. Eventually, options to circumvent compromised units are evaluated at the higher levels in order to maintain operation albeit in a constrained fashion or in degraded modes. This framework envisage an expanded range of sensors and actuators beyond whatis necessary for low-level control systems.Several novel research and educational contributions will be realized as a result of this proposal. Research contributions include a hierarchical design loop that facilitates scalability. Appropriate task sharing between the four layers will be aligned with the computational capability of the nodes and type of information available at each level. Such an approach ensures that a set of heterogeneous components work in symbiosis to detect and circumvent cyber-attacks. Secondly, the researchers present methods that use multiple touchpoints tophysical processes to detect attacks within lower levels close to the sensor. For example, the planned framework will leverage combinations of trustworthy and novel indirect measurements of physical and cyber processes. These include electro-magnetic or acoustic emissions of physical plant actuators, and side-channels to measure computing and electronic system behaviors. These will provide a spectrum of additional sensing independent of primary data to determine the inception of an attack, and alongside cryptographic protocols with remote attestation and data signing these measurements can be used to establish trusted truth sources necessary for information assurance. Finally, at various network levels detectors will use hints from the lower layers, to implement more scalable and robust consensus algorithmsin distributed systems. The education contributions build on successful internships at UMBC and student projects at USNA. This effort will continue current student involvement via one-year capstone projects, technical and independent research courses, and summer internships. Additionally, a current USNA Cyber Science course will be updated to include hands-on labs built by this group and development of a new Internet of Things course.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 10, 2018
- Source ID
- N000141812450
Entities
People
- Ryan Robucci
Organizations
- Office of Naval Research
- United States Navy
- University of Maryland, Baltimore