Adaptable Cyber Defense for Autonomous Air Operations

Cyber defense is difficult, but presents a particularly thorny problem for legacy systems, including legacy embedded systems, where in many such cases source code may not even be available. In this project, we proposed to investigate and extend a pattern-based approach recently developed by USC-ISI for analyzing and retrofitting binary code to protect against potential attacks. In this project, we will explore how this approach can be extended by adding execution time analysis so that a greater range of countermeasures can be injected to protect a system. In addition, we will also investigate the use of analytic machine learning algorithms that can learn new patterns by observation, so that a broader and more complex set of threats can be addressed over time, evolving the behavior of the system.