When LLMs Meet ASNs - Rethinking AS-to-Organization Mapping at Internet Scale

Learn how large language models can revolutionize the mapping of Autonomous System Numbers (ASNs) to their operating organizations through this conference talk from NANOG. Discover the limitations of current mapping approaches that rely on WHOIS and PeeringDB data, which often miss cross-regional connections, corporate rebranding, and complex parent-subsidiary relationships that are crucial for Internet measurement, routing security, and operational coordination. Explore ASINT, an innovative large-scale retrieval-augmented pipeline that combines traditional registry data with unstructured web sources including company websites, Wikipedia, and news reports to uncover hidden organizational relationships. Understand how named entity recognition, relevance filtering, and LLM-based inference techniques enable the unification of 111,470 ASNs into 81,233 organization families, identifying thousands of previously unknown connections. Examine the practical impact of this approach, which increases RPKI misconfiguration detection by 27.5%, reduces false-positive hijack alarms by 9.4%, and corrects 5.9% of misclassified IP leasing cases. Access the complete dataset, API, and searchable web interface at https://asint.netsecurelab.org/, which includes a community feedback mechanism for operators to confirm or suggest corrections to mappings. Gain insights into the system design, review key operational case studies, and learn how this continuously updated dataset can improve Internet security and measurement practices.