Achieving the Security and Reliability of Semi-Autonomous Cyber-Physical Systems
Abstract
The objective of this research is to i) gain insights into the design challenges and vulnerabilities of semi-autonomous (SA) systems, like cars, drones, airplanes and ships, ii) develop an end-to-end (e2e) framework, called CADCA, for control anomaly detection and,resolution, and iii) validate the effectiveness of CADCA using real-world implementation and attack simulation. SA systems ? systems, with both manual and autonomous control capabilities?have become prevalent in our daily lives and jobs. While incorporating the con,venience of autonomous control and the flexibility of manual control, SA systems also inherit the vulnerabilities of both control ty,pes. Specifically, an SA system can behave anomalously due to component failures, design/software bugs, or even malicious attacks wh,ile the manual control is also far from perfect due to human errors or malicious intention. However, safety features in modern SA sy,stems are usually implemented with a static control priority assignment that can lead to catastrophic accidents when accompanied wit,h component failures and attacks. Boeing 737 MAX [1] and Germanwings Flight 9525 [2] incidents are two well-known examples of such a, design principle.Considering the grave safety risk caused by both human errors and system failures (i.e., bugs and attacks), we pro,pose to develop a systematic way of preventing abnormal control behavior that may lead to fatal accidents. We will first conduct an,extensive analysis of the vulnerabilities of modern SA cyber-physical systems and identify the possible consequences of their exploi,tation by adversaries. We will then develop and evaluate an e2e framework, CADCA, for detection and resolution of control anomalies,in SA systems. This framework will have two major functionalities. First, it will detect any abnormal behavior in autonomous/manual,control. Second, the framework will resolve any disagreement between manual and autonomous control decisions in a safety-critical si,tuation by taking the operation ?context? into consideration.The PI, Kang Shin, has been investigating robustness and safety issues,in cyber-physical systems for decades and is now ready to tackle the challenging security/safety problem in SA systems. In particula,r, he and his students have recently been working on the identification of potential attack surfaces and security issues, and the de,velopment of defense mechanisms against the attacks of SA cyber-physical systems. They have also made important strides in developin,g a platform for data collection, implementing and evaluating the detection and diagnostics components of the proposed framework. Th,e proposed research will build on these experiences, results and foundations.Approved for Public Release.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 13, 2022
- Source ID
- N000142212622
Entities
People
- Kang Shin
Organizations
- Board of Regents of the University of Michigan
- Office of Naval Research
- United States Navy