Approach flawed if ports left in dataset

munhouiani / Deep-Packet

Pytorch implementation of deep packet: a novel approach for encrypted traffic classification using deep learning

MIT License

183 stars 56 forks source link

Approach flawed if ports left in dataset #41

Closed Mr-Pepe closed 1 year ago

Mr-Pepe commented 1 year ago

I just stumbled upon this repository and it looks like only IP information is removed from the dataset. Are you aware of this? I find it hard to believe that your model actually learned anything besides the mapping of network flows to traffic classes.

munhouiani commented 1 year ago

Are you aware of this?

No, I didn't. I was researching the usage of deep learning in packet classification several years ago and found the Deep packet paper. The authors didn't publish their code online, so I implemented it myself in PyTorch. I didn't look into the data, so you might be right. The model could be just mapping the network flows to traffic classes.