| Autor: |
Akem Aritside, Bütün, Beyza, Gucciardo, Michele, Fiore, Marco |
| Jazyk: |
angličtina |
| Rok vydání: |
2023 |
| DOI: |
10.5281/zenodo.8125040 |
| Popis: |
Recent endeavours have enabled the integrationof trained machine learning models like Random Forests inresource-constrained programmable switches for line rate inference.In this work, we first show how packet-level informationcan be used to classify individual packets in production-levelhardware with very low latency. We then demonstrate how thenewly proposed Flowrest framework improves classificationperformance relative to the packet-level approach by exploitingflow-level statistics to instead classify traffic flows entirely withinthe switch without considerably increasing latency. We conductexperiments using measurement data in a real-world testbedwith an Intel Tofino switch and shed light on how Flowrestachieves an F1-score of 99% in a service classification use case,outperforming its packet-level counterpart by 8%. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
