Interactive Trillion Packet Anonymized Network Analysis with the GraphBLAS
Chasen Milner, Michael Houle, Hayden Jananthan, Michael Jones, Jeremy Kepner, Peter W. Michaleas, Inna Voloshchuk, Alex Pentland · 2025
Interactive network analysis is a key element of improving and protecting networks. Modern communication networks deliver enormous volumes of traffic; often exceeding trillions of individual communications events (packets). Finding indications of anomalous activity within these vast streams is mathematically and computationally challenging. Recent innovations with the GraphBLAS high performance math library makes such interactive analysis tractable. The IEEE/MIT/Amazon Anonymized Network Sensing Graph Challenge is an ideal venue for highlighting these innovations. Merging zone matrix concepts frequently found in network firewalls with the Anonymized Network Sensing Challenge analysis a rapid interactive tool is demonstrated that enables anomalous activity on a wide range of scales to be identified in seconds with modest computing resources. These methods are inspiring a new class of network analysis products to more rapidly analyze network traffic.