Network traffic identification of several open source secure proxy protocols

Secure proxy technique has a wide range of applications especially in the area of VPN to protect the private communication from being surveilled. However, the adversaries may also take advantage of the VPN services for other purposes. Thus, knowing the characteristics of the corresponding security protocols will benefit the administration to narrow down the targets. Also, given the various VPN services on the market, we also want to know whether there exists some security flaws from the cryptographic protocol's point of view, so that can be used to improve analyzing the protocol. In this paper, we first target several widely used open source secure proxy protocols and provide their weaknesses through the static analysis. Then we apply the deep learning technique to train all the related network traffic, and our results showed that all the secure proxy traffics can be successfully distinguished with high accuracy.