A Few Shots Traffic Classification with mini-FlowPic Augmentation

What is it about?

We show that for image-based traffic classification, e.g., using FlowPic, we can use Contrastive Representation Learning with augmentation that are based on manipulating the data before transforming it into an image. As a result we show few-shot solutions that have excellent accuracy.

Featured Image

Why is it important?

It is hard to obtain labeled datasets for networking data, but easy to get large unlabeled dataset. Our solution can facilitate using unlabeled datasets with a small work of labeling just a few samples per class.