HYDRA: A SUBMODULAR OPTIMIZATION FRAMEWORK FOR DYNAMIC NETWORK RESILIENCE IN THE PRESENCE OF ANTAGONISTIC INTERACTIONS AND INTERDEPENDENT FAILURES

Abstract

Networks in Air Force-relevant applications need to be robust against natural failures and resilient against deliberate attacks. Two critical failure modes are (i) antagonistic interactions, in which an interaction between nodes causes their state values to separate, and (ii) interdependent failures, in which the failure of a single node or edge can trigger cascading failures of other nodes and edges. Agile response to such failures requires design of an efficient monitoring system that is integrated into the network. The proposed research is to develop a submodular optimization framework for resilient complex networks. Submodularity is a diminishing returns property that enables the development of computationally efficient algorithms with provable optimality bounds. Since complex networks contain both discrete and continuous design elements, both discrete as well as continuous forms of submodularity will be investigated. The proposal highlights the need to define new notions of submodularity (hybrid submodularity) for optimization of joint continuous-discrete functions in complex networks. The main research thrusts are:: Resilience to Antagonistic Interactions: This thrust will investigate resilience to antagonistic interactions in networks with linear dynamics as well as networks with nonlinear Kuramoto dynamics. Resilience to Interdependent Failures: This thrust will investigate mitigation of interdependent failures, with a focus on controlled islanding, in which a subset of edges is deliberately removed, allowing additional nodes to fail, in order to stop the spread of the cascade. Monitoring for Responsive Resilience: This thrust will investigate and develop monitoring algorithms for selecting a set of information to gather to detect network failures and attacks with minimal false positives/negatives and performance overhead.

Document Details

Document Type
DoD Grant Award
Publication Date
Aug 12, 2021
Source ID
FA95502010074

Entities

People

  • Linda Bushnell

Organizations

  • Air Force Office of Scientific Research
  • United States Air Force
  • University of Washington

Tags

Fields of Study

  • Computer science

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Distributed Systems and Data Platform Development
  • Educational Psychology