CAREER: A Cross-Domain Approach to Improving the Attack Resilience of Robotic Systems

Robotic systems increasingly operate in critical sectors such as manufacturing, healthcare, agriculture, and defense, where security and reliability are essential. These systems combine components from the computational and physical domains that interact closely with one another, creating complex interdependencies and vulnerabilities. Traditional security approaches focus on either the computational or physical domain in isolation, leaving significant gaps in protection. This project aims to address these challenges by creating a comprehensive security framework that models and protects the intricate dependencies between computational and physical components. The project's novelties are in its cross-domain approach to vulnerability analysis, real-time mitigation, and forensic investigation. The project's broader significance and importance are in improving the safety and trustworthiness of robotic systems in high-stakes environments, advancing cybersecurity education, and broadening participation in computing through outreach and hands-on learning. This project develops a cross-domain security framework that systematically mitigates vulnerabilities in robotic systems by integrating advanced modeling, real-time detection, and post-attack analysis. The research consists of three technical thrusts. The first thrust focuses on vulnerability analysis by constructing unified models that capture the interactions between computational and physical components, using static analysis, system identification, and stateful fuzzing to detect cross-domain weaknesses. The second thrust designs real-time attack mitigation methods by leveraging these models for predictive state monitoring, enabling early detection of discrepancies between expected and actual system behavior across both domains, and facilitating recovery from corrupted physical states. The third thrust introduces a post-attack investigation technique that applies deterministic replay and causal inference to trace the root causes of attacks through both computational and physical domains. Together, these efforts contribute to a new generation of tools that enhance the resilience of robotic systems. Results from the project will improve the development and deployment of secure robotic applications, influence cybersecurity education, and foster public trust in robotic technologies. 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.