Synthesizing temporal logic and human performance models for deception mitigation
Abstract
Deception is a key element in attacks on military and civilian cyberphysical systems. With increasing use of automation and autonomy in such systems, such vulnerabilities to deception are growing, with catastrophic outcomes, as evidenced by recent major breaches in cybersecurity across the DOD. There is a pressing need to understand long-term adversarial strategies where one hostile action islargely harmless and subtle, but multiple actions over long periods of time comprise a successful attack. To this end, we will develop domain-agnostic deception models that blend temporal logic specifications with probabilistic models of human attention over a long period of time that explore multiple episodes of interactions between one or more humans and an adversary. The resulting validated deception models will be used to develop mitigation and counter-deception strategies, specifically to determine whether and how tobest detect and possibly defend against such deceptive attacks. In addition, we will explore the development of visualizations and other cognitive assistance tools to help humans detect possible abnormalities in their work flow, so that they can possible intervene sooner than they otherwise would, if at all.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 24, 2023
- Source ID
- N000142312651
Entities
People
- Missy Cummings
Organizations
- George Mason University
- Office of Naval Research
- United States Navy