Design and Operation of Secure Multi-Agent Networks
Abstract
Networks and distributed systems will play an increasingly fundamental role on the battlefield and in other operating theaters of interest to the U.S. Army, where success relies critically on integrity and availability of network resources. Despite redundancy in numbers, protected communication, and supervised operation, capable adversaries can disrupt networked operations by observing the behavior of a few units, spoofing exchanged messages, and tampering with coordination protocols. Crucially, it is the networked coordination among the units that introduces these vulnerabilities, which have neither been analyzed nor considered in the design and operation of network systems. This research stems from recent breakthroughs from the PI showing how security of networked multi-agent systems can be ensured at the design stage, by shaping interaction dynamics, and during operation, with reactive monitoring schemes. Blending these methods with classic cyber- and information-based techniques will dramatically increase reliability of networked systems in adversarial environments. To this end, we propose to address three fundamental research questions: What network features facilitate manipulability and increase robustness? We propose to identify structural and dynamic features that allow manipulability from benign controls and robustness against attacks and disturbances, and to characterize manipulability/robustness tradeoffs. How do we design networks with guaranteed manipulability and robustness? We propose to design algorithms to engineer networks that exhibit different desirable behaviors, at the node level, in response to localized stimuli from authorized controls and malicious disturbance. How do we ensure reliable network operation against coordinated adversaries? We propose to design network-centric methods to detect and identify nodes compromised by failure and coordinated attacks, and to reveal fundamental tradeoffs between network structure and security. This project will help the U.S. Army achieve reliable and efficient (semi)autonomous systems in networked, uncertain , dynamically changing, hostile environments. This project will undertake a rigorous approach based on control theory, graph optimization, and statistics to develop novel theories and tools to study and operate network systems. Results will be validated via simulations and experiments using aerial and ground vehicles in realistic environments similar to complex civilian and military platforms of interest to the U.S. Army. Because the results of this project will uncover foundational principles in the design and control of multi- agent networks, the results will have far-reaching implications across science and engineering, including for the social, medical and energy domains, as already demonstrated by the PI s preliminary work.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 11, 2018
- Source ID
- W911NF1710358
Entities
People
- Fabio Pasqualetti
Organizations
- Army Contracting Command
- United States Army
- University of California, Riverside