JEPA Representation Learning for Intrusion Detection on Real-World Drifting Network Traffic
Supervisors: Prof. Guido Marchetto, Prof. Alessio Sacco, Cristiano Serra, MSc
Does a representation learned from unlabeled network traffic keep an intrusion classifier more accurate than raw flow features on traffic recorded years later?
Built an intrusion detection pipeline in Python with flow preprocessing, a self-supervised JEPA model in PyTorch and four classifiers (logistic regression, decision tree, random forest, XGBoost), using real Internet backbone traffic from 2011, 2016 and 2021 (MAWIFlow benchmark with CICFlowMeter flow features and MAWILab labels).
Tested each year’s classifiers on every year. The representation made anomalous traffic easier to separate with a linear model but did not reduce the drop across years, and the labels changed meaning after 2011.
Audited the benchmark. Showed that its scaling lost most of the resolution in the flow features and replaced it, and found that its published metrics contradict its published samples for two of the three years.
- Python
- PyTorch (CUDA)
- scikit-learn
- XGBoost
- Polars
- NumPy
- Matplotlib