Learning AS-to-Organization Mappings with AI/ML
…the Cartographers Guilds made a Map of the Empire whose size was that of the Empire, and which coincided point for point with it.
— J.L. Borges. On Rigor in Science. (1946)
Carlos Selmo, Esteban Carisimo, Fabián E. Bustamante, J. Ignacio Alvarez-Hamelin
Borges is a Python package and CLI tool that maps Autonomous Systems (ASNs) to their parent organizations, revealing the true corporate structure of the Internet. By combining traditional WHOIS data with AI-powered analysis of websites, favicons, and unstructured text, Borges uncovers hidden relationships that existing methods miss.
Our approach achieves 0.947 accuracy in extracting sibling AS relationships and outperforms prior systems by 7% on the Organization Factor metric. This improved mapping reveals approximately 192 million previously uncounted Internet users and enables better understanding of Internet resilience, incident response, and market concentration.
@inproceedings{borges:imc,
author = {Carlos Selmo and Esteban Carisimo and
Fabián E. Bustamante and J. Ignacio Alvarez-Hamelin},
title = {Learning AS-to-Organization Mappings with Borges},
booktitle = {Proc. of ACM IMC},
year = {2025},
month = {10}
}
Internet Society Pulse - An in-depth look at how Borges reveals the hidden corporate structure of the Internet and its implications for policy and resilience.
Excited to share our work on Borges at IMC 2025! We built a system that maps who really controls Internet infrastructure by combining AI/ML with traditional network analysis. Check out how we uncovered 192M hidden Internet users: github.com/NU-AquaLab/borges
— Esteban Carisimo (@estcarisimo) November 2025
The project is named after Jorge Luis Borges, the renowned Argentine writer whose story “On Rigor in Science” describes cartographers creating a map so detailed it becomes as large as the territory itself. Like Borges’s fictional map, our tool aims to create a comprehensive representation of Internet ownership—though thankfully, in a more manageable format.