NeTS: Technological Management of Computer Networks

Modern society depends on computer networks for daily life, economic activity, and essential public services. Yet these networks remain difficult for people to manage safely and reliably because many operational tasks are detailed, repetitive, time-sensitive, and easy to perform incorrectly. This project advances a new research vision called Technological Management of Computer Networks: network systems should be designed not only to carry data, but also to make their own operation understandable, governable, and trustworthy. In this vision, people retain responsibility for judgment, policy, and accountability, while artificial intelligence agents carry out low-level operational tasks within approved limits. The central research challenge is to determine how such delegation can be made safe and robust: automated actions must be tied to human authority, checked against policy, controlled within defined boundaries, recorded as evidence, and open to review. By treating management, operations, and compliance as core technical properties of network systems rather than as external paperwork or after-the-fact automation, the project aims to create a new foundation for reliable and accountable digital infrastructure. The work will support national needs for secure networks by reducing preventable failures, improving trust in AI-assisted operations, and training students in secure, controllable, and robust network systems. This project will develop and evaluate the scientific and technical principles needed to make technological management practical. The research will formalize a human and artificial intelligence division of responsibility in which people provide governance and accountability by defining policies, approving procedures, assigning authority, evaluating tradeoffs, and reviewing evidence. Artificial intelligence agents provide fulfillment by interpreting operational intent, inspecting telemetry, selecting approved workflows, preparing configuration changes, and carrying out routine tasks within technical and organizational limits. The project will use both frontier artificial intelligence models, which provide leading capabilities for reasoning and planning, and locally hosted models, which can support privacy, cost control, availability, and operation close to managed systems. The project will investigate how signed workflows, role-based authority, policy-controlled system components, tamper-resistant operational logbooks, and improving artificial intelligence capabilities can be combined so that network operations are not merely automated, but verifiable, auditable, and accountable. The work will be built and tested using an open-source, HTTP-based overlay networking platform developed from foundational open-source technologies, including Linux, Envoy Proxy, and the Open Policy Agent. The results are expected to be useful for managing these technologies individually, as well as for managing them together as a policy-based overlay network. Experiments will measure safety, responsiveness, governance, and compliance evidence compared with more manual forms of network management. Example tasks include detecting latency problems, adjusting network parameters within approved limits, carrying out staged configuration rollouts, reviewing access permissions, rotating keys and certificates, and preparing evidence for audits. The project will produce open-source software, repeatable benchmark experiments, and a capstone remote-access overlay system that demonstrates a new model of artificial intelligence-assisted network operations under human governance. 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. NSF Award ID: 2551260 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Patrick Crowley | Institution: Washington University, SAINT LOUIS, MO | Award Amount: $675,000 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2551260 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2551260.html