SaTC: CORE: Small: Mitigating Threats of Physical-Domain Signal Injections on Security, Reliability, and Safety of Sensing and Control Systems

University Of Louisiana At Lafayette, Lafayette LA

Investigators

Abstract

Light, motion, and sound sensors are used in many applications, such as smart homes, self-driving cars, and industrial control systems. As with many computing technologies, sensors can be vulnerable to attacks. In particular, attackers may try to inject false information into sensors; this could lead to the cyber-physical systems they rely on having an incorrect model of the world, in turn affecting these systems’ reliability and safety. This project’s goal is to detect and mitigate those attacks, based on the fundamental insight that the underlying physics of light, motion, and sound can be used to detect inconsistent signals. Developing methods that use physical properties to detect possible attacks, as well as sensor noise and errors, can lead to safer and more robust cyber-physical systems. The planned work will support a number of post-graduate, graduate, undergraduate, and high school educational activities through integrating projects with coursework and research projects, while developing research tools, testbeds, and datasets that can be used by others for both research and education. This project aims to secure sensor-based systems with untrustworthy analog sensing components through three research thrusts. The first thrust involves determining how to reliably detect and mitigate transducer-stage o ut-of-band acoustic injections on inertial sensors, through embedding modulated patterns at the digitization stage of out-of-band signals and develops robust inertial sensing on multiple sensors. The second thrust explores defenses against light-based attacks on cameras to secure vision systems, modeling threats of light injections that exploit the same modality (i.e., light) as image sensors and utilizing adversarial optical physics to defend against various light injections. The third thrust will develop testbeds, tools, datasets, procedures, and metrics to test, improve, and validate sensing system security and reliability under physical-domain signal injection attacks. Together, the project outcomes will provide direct help to secure kinematics-based sensing and actuation systems, as well as autonomous vehicle, against various adversarial injection attacks. 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.