Secure, Distributed, and Continuous Machine Learning in Navy Tactical Networks
Abstract
This project focuses on developing frameworks and algorithms for secure, distributed, and continuous machine learning in tactical networks that are characterized by agent heterogeneity and autonomy, constraints on computational and energy resources, and stringentperformance requirements, particularly on latency. This new paradigm of distributed machine learning, based on on-device computing and device-to-device communication at the edge of the network, has received significant attention lately. However, critical issues of security and resilience have been relatively less explored and the developments have mostly been fragmented. This is primarily dueto the fact that distributed ML has failed to account for significant security and resiliency challenges of tactical networks. For example, distributed ML has not addressed the need to detect unattended-agent capture, followed by agent removal and replacement without human intervention beyond simple release of new agents; nor has it addressed the need to assure both private and authentic inter-agent communication despite traffic-analysis and message-injection attacks. Also, relying on a common IoT infrastructure, distributed ML has inadvertently adopted Internet protocols with large attack surfaces, and has not exploited the geo-proximity of ML agentsto create new secure and robust protocols. That is, it has inherited all Internet vulnerabilities caused by dependencies on DNS, PKI, NTP at the network edge simply because research attention focused exclusively on the new algorithmic paradigms. This project aimsto fill this gap by developing principled approaches for secure and resilient online machine learning in distributed agent networks. Specifically, by employingscalable randomized key-predistribution-based mechanisms in conjunction with resilient distributed algorithms that exploit task specific redundancies, we propose to develop methodologies and an integrated framework for online distributed machine learning with performance guarantees and security assurances against broad classes of adversarial attacks and anomalies (including node capture and data injection attacks).With the Navys strategic focus on Distributed Maritime Operations, distributed algorithmic processing at the edge and embedded on platform becomes increasingly important to integrate geographically distributed naval forces and synchronize operations across all domains. The output of this research would directly inform the development of and enhanced security for the Naval Operational Architecture as the connective tissue between sensors, platforms and decision makers in adversarial and contested environments - enabling faster decisions, leveraging information from across the fleet, integrating manned and unmanned assets and more. This project will leverage the military experience and knowledge of Navy personnel and ROTC Midshipmen in the Pittsburgh region leveraging the Hacking for Defense methodology to explore the problem/research approachand to develop a high-level Concept of Operations that will be used for a technical demonstration of secure and distributed machine learning prototype algorithms and capability.NEPTUNE Project Sponsor: Naval Surface Warfare Ceter - Dahlgren DivisionPOC: Jenn Clift, Chief Technology Officer
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 05, 2021
- Source ID
- N000142112547
Entities
People
- Osman Yagan
Organizations
- Carnegie Mellon University
- Office of Naval Research
- United States Navy