These are the Anti-Nex datasets for training deep neural networks to defend against network exploits (only web applications added so far). Under the hood these models are trained and tuned using Keras and Tensorflow.
These datasets were created by merging datasets recorded from OWASP ZAP attack simulations and multi-user non-attack simulations with Django.
Each model has an associated guide for preparing datasets and training models:
- Protecting Flask RESTplus with 89% accuracy
- Protecting React and Redux with 87% accuracy
- Protecting Vue with 83% accuracy
- Protecting Django with 70% accuracy
- Protecting Spring with 66% accuracy
This is a 75 MB gif I recorded while capturing the non-attack training data by running the multi-user simulation with the capture tools all active inside a single vm (note: from inside imgur, right click on the gif and select Open in video in new tab to make it larger to read the text):
