Datasets Download

BEFORE USING THE DATASETS:

As discussed in our paper, we opted to use the "Attempted" label for an attack (e.g. DoS GoldenEye - Attempted) whenever a flow that was intended to be part of the attack does not exhibit any malicious behaviour. Researchers are free to decide for themselves whether they want to combine the "Attempted" flows into the actual attack label, as this will depend heavily on the flow preprocessing method that is applied. When in doubt, we recommend to relabel all "Attempted" flows as Benign (the easiest way to do this is to look at all flows where the "Attempted Category" feature is different from -1). Under no circumstance should the "Attempted" flows be treated as a separate label for your Machine Learning model!

The datasets can be downloaded here