I'm trying to teach AI how to defend against network exploits over ssh and telnet and make it all free open source. I wanted to know if anyone on here has any insights that I might be missing with the initial proposal to: use a test bed for recording a distributed malware simulation to generate some initial training datasets.

For background, I am hoping to reuse all of the v1 webapp process which yielded datasets where neural networks can hit >99.7% accuracy at predicting attacks (https://github.com/jay-johnson/antinex-core/blob/master/docker/notebooks/AntiNex-Using-Pre-Trained-Deep-Neural-Networks-For-Defense.ipynb). I really just want to hear if I am missing something obvious before I start testing and building the dataset and hacker news is where I have learned so much about so many different things I had never heard about.

Here's my proposal for the simulation and test bed: https://github.com/jay-johnson/antinex-datasets/tree/v2/v2/malware

Docs if you're curious: http://antinex.readthedocs.io/en/latest/