This project proposes a novel cyber-physical paradigm for enhancing the security in networks of advanced robots from external malicious cyber attacks (and internal non-malicious malfunctions as well). Such systems have tremendous potential for improved productivity, but also carry new risks: malicious actors could exploit the connectivity of the devices to carry out attacks with consequences in the physical world. The core idea of this project is to provide a novel layer of security against such threats by designing distributed security specifications based on introspection: agents use physical-sensing capabilities to surveil the behaviors of other agents in the team in addition to the task-specific mission objective. Such specifications will offer an additional layer of protection in emerging applications with networked robots.
Our proposed research revolves around three main thrusts: A) systematically identify attack models specific to the scenarios considered, together with countermeasures based on measurements in the physical world, B) develop abstraction-based high-level planning algorithms that generate flexible plans that satisfy both the task specifications and the additional security requirements, and C) new low-level controllers that can negotiate the high-level plans with real-time objectives (such as obstacle avoidance) while allowing for collaboration between the agents. The proposed research includes an evaluation plan on a robotic testbed which includes camera-equipped ground and aerial vehicles, as well as short-throw projectors for creating augmented reality environments emulating our motivating applications.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
