Cyberattacks usually bring to mind cyber crime, which ultimately seeks money (e.g. ransomware), hactivism, which seeks to disrupt an entity’s IT systems, or espionage, which seeks corporate and government information.  A less frequent association is cyber conflict, which seeks to degrade or destroy physical infrastructure such as power, petrochemical, and manufacturing plants.  What was accomplished by aerial bombing campaigns in WWII can now be done by malicious code introduced through networks to the embedded computers controlling physical processes.  This type of attack has great potential for widespread economic disruption as demonstrated by the December 2015 power disruption in Ukraine that left more than 230,000 residents in the dark.  While the response to cyber crime, hactivism, and espionage is usually reactive (software patching and/or updating virus signatures), cyber warfare defenses need to be proactive since there may be nothing left to defend after the initial attack or it may not be possible to return processes to stability beyond a certain degree of disturbance.  Despite being networked, the simple microcontrollers bridging the cyber and physical worlds generally have few internal protections against malware.  Network exploits developed in the IT realm carry over to process control systems.  We therefore seek to add a last line of defense beyond conventional network protections such as firewalls.

One option, inspired by the movie Minority Report, is prediction and pre-emption.  Rather than have psychics see what will occur in the future, the physical world is mirrored by a predictive virtual world that connects a copy of the (possibly compromised) plant control software to a computer model of the physical plant.  The virtual world can be fast forwarded to see what latent malware in the microcontroller may attempt to do in the near future.  Even a short preview may help to avoid process instability and permit an orderly transfer to a simple, trusted backup controller.  A second option is to use machine learning to classify the possible states of the plant into safe and unsafe states.  When the classifier detects the plant moving into an unsafe region it assumes that the controller has been compromised and transitions to the trusted backup controller.  In both options, it is important that the monitoring, prediction, switchover, and backup components be isolated from any and all microcontroller software, which is accomplished by using programmable system-on-chip platforms that integrate microcontrollers and configurable hardware on the same chip.  The configurable hardware, which cannot be modified over the network, implements the trusted components.

In the cybersecurity realm, workforce education is as important as technical solutions.  Designing protections for cyber-physical systems requires a knowledge of two fields that traditionally have little intersection: network security, which is traditionally associated with computer engineering and computer science programs, and control system engineering, which is normally part of electrical, mechanical, and aerospace engineering programs.  Lab modules created in this project incorporate cyber-physical security concepts into learning modules that highlight real-world technical issues.  This approach helps students understand control system architectures and their vulnerabilities to cyber-attacks via experiential learning, and acquire practical skills through actively participating in the hands-on exercises.  The goal of these lab modules is to show how an adversary could destabilize a computer-controlled process by exploiting vulnerabilities in network, supervisory, and process controller layers.  A mock testbed environment is created with low cost commercial-off-the-shelf hardware suiting undergraduate embedded system, control system design, or cybersecurity courses.  The ultimate goal is to have students appreciate the awesome responsibilities of embedded microcontrollers.

Last Modified: 09/20/2016
Modified by: Cameron D Patterson