Using DetGen, we have generated the following datasets:
- DetGen-IDS: Moving towards a more general dataset constructed to apply this probing methodology, we constructed DetGen-IDS. This dataset is suitable to quickly probe ML-model behaviour that were trained on the CICIDS-17 dataset. The dataset mirrors properties of the CICIDS-17 data to allow pre-trained models to be probed without retraining. The DetGen-IDS data therefore serves as complementary probing data that provides microstructure labels and a sufficient and controlled diversity of several traffic characteristics that is not found in the CICIDS-17 data. This dataset is further described in the following paper:
Henry Clausen, Robert Flood, David Aspinall, "Controlling Network Traffic Microstructures for Machine-Learning Model Probing", EAI SecureComm conference, September 2021
- Stepping-stone dataset: The detection of interactive stepping-stones is challenging due to various reasons. Like many intrusion attacks, stepping-stones are rare and there exist no public datasets, leading researchers to evaluate their methods on self-provided private data, which makes a direct comparison of the achieved results impossible. We release a large and comprehensive dataset suitable for the training of machine-learning-based methods and in-depth performance evaluation. To our knowledge, this is the first public SSD dataset. This dataset is further described in the following paper:
Henry Clausen, Michael Gibson, David Aspinall, "Evading stepping-stone detection with enough chaff", Conference on Network and System Security, November 2020
Both datasets are currently being uploaded to our data repository, where they will be available to other researchers.