The Aleph - Decoding DNS PTR Records with Large Language Models

Learn how to leverage Large Language Models to automatically decode DNS PTR records for accurate network infrastructure geolocation in this 21-minute conference talk from NANOG. Discover The Aleph, a novel approach that addresses the persistent challenge of geolocating network infrastructure by using LLMs to classify PTR naming patterns, generate extraction regexes, and map geographic hints to real-world locations at Internet scale. Explore how this method processes billions of PTR records from over 20,000 ASes, achieving 94% confidence in IPv4 and 90% in IPv6 geolocation accuracy through validation with ground truth data from network operators and active measurements of 200+ million IPs. Understand the practical applications for latency tuning, SLA verification, outage forensics, and regulatory compliance, while examining the integration into IPinfo's production geolocation pipeline and opportunities for community adoption through open APIs and datasets. Gain insights into how machine learning can solve complex networking problems and enhance traceroute interpretation for improved network operations and research.