Multitask Learning for Network Traffic Classification Article Swipe
YOU?
·
· 2020
· Open Access
·
· DOI: https://doi.org/10.1109/icccn49398.2020.9209652
· OA: W2949072481
Traffic classification has various applications in today's Internet, from\nresource allocation, billing and QoS purposes in ISPs to firewall and malware\ndetection in clients. Classical machine learning algorithms and deep learning\nmodels have been widely used to solve the traffic classification task. However,\ntraining such models requires a large amount of labeled data. Labeling data is\noften the most difficult and time-consuming process in building a classifier.\nTo solve this challenge, we reformulate the traffic classification into a\nmulti-task learning framework where bandwidth requirement and duration of a\nflow are predicted along with the traffic class. The motivation of this\napproach is twofold: First, bandwidth requirement and duration are useful in\nmany applications, including routing, resource allocation, and QoS\nprovisioning. Second, these two values can be obtained from each flow easily\nwithout the need for human labeling or capturing flows in a controlled and\nisolated environment. We show that with a large amount of easily obtainable\ndata samples for bandwidth and duration prediction tasks, and only a few data\nsamples for the traffic classification task, one can achieve high accuracy. We\nconduct two experiment with ISCX and QUIC public datasets and show the efficacy\nof our approach.\n