Project Description

The very fabric of modern society is interwoven with computing, from managing critical infrastructures like transportation systems and power grids to predicting climate changes and tracking epidemics. This omnipresence of computing, supporting the very backbone of the most fundamental operations of society and nation, has transformed the threat landscape by interconnecting both cyber and physical domains with an increasing number of vulnerabilities discovered daily.

This project will develop the theoretical foundations as well as the system implementations to automatically and efficiently discover, prioritize (exploit), and mitigate vulnerabilities in cyber-physical systems. The proposed reasoning system scientifically formulates hacking into formal verification problems on a common abstraction that allows for analysis of essential physical properties of cyber-physical systems, such as real-time availability and control integrity. The theory to bridge the semantic gap among different layers for reasoning will lay the foundation for new algorithms and system designs for safety and security protection. Building on the insights from an in-depth understanding of the vulnerabilities, this project also aims to leverage physics principles to detect, prevent, and operate through adversarial attacks. Such physical-world informed defense will bring a paradigm shift into how adversarial attacks are modeled upon and defended against. Lastly, a cyber-physical battlefield testbed will be developed not only to provide the evaluation environment but also to serve as a platform for community and stakeholder engagement.

The proposed research activities tackle the increasingly pressing societal challenge of cyber-physical system security through the lens of automation. To do so, it tackles several long-standing research problems by reasoning via a common abstraction grounded on cyber-physical control loop and driven by physics principles. The interdisciplinary nature of the approach necessities collaborative research connecting communities from computer security, real-time computing, and control theory, as well as artificial intelligence and machine learning. If successful, the proposed project will be the seed for a new scientific community from different backgrounds under the common goal of securing the nation's most dependent critical cyber-physical assets.

The reasoning capability will fundamentally transform the timescale in both offensive maneuvers and defensive actions, introducing a revolutionary shift in battlefield dynamics and giving the US Army a strategic advantage. Beyond military application, the scientific approach to automatically and efficiently protect critical assets also has profound implications for national security and the lives and livelihood of the citizens.

Acknowledgement

This project is supported by ARO under award W911NF-24-1-0155.