Winner of the NetML Network Traffic Analytics Challenge 2020
Date:
Winner of the NetML Network Traffic Analytics Challenge 2020 (TAU-Team).
NetML Challenge 2020 is a machine learning driven network traffic analytic challenge where participants are required to apply novel machine learning technologies in order to detect malicious flows or distinguish between applications in a fine-grained fashion among network flows.
In the challenge’s dataset, each flow is represented by 121 features extracted from up to 200 packets in both directions as described by Barut etal. and is labeled by 2 or 3 hierarchical multi-labels. The dataset consists of records of over 1M flows and relies on 3 public sources: NetML, CICIDS2017 and non-vpn2016.
The challenge is composed of 7 tracks, 4 of them for malware detection and the rest are for traffic classification. We , the TAU-Team, are the only team to capture one of the top 3 places in all of the tracks and first place in more than one. In this document, we share technical details of our approach for Track-3 and Track-5, where we got first place.
You can find a detailed description of our approach in the following link.