Shadowsocks Traffic Identification Based on Convolutional Neural Network
Nan Zliang, Tiantian Wu, Yuening Zhang, Mingzhong Xiao · 2020
Network traffic has been massively produced since the development of the Internet. While the encryption of traffic has ensured the security and reliability of information, it also brings great challenge to the traffic identification and monitoring. The present study proposes a method of shadowsocks traffic identification based on the one-dimensional Convolutional Neural Network. This method simplifies the feature extraction of traffic identification and the recognition accuracy is over 98%. Because we can not find the published shadowsocks traffic dataset, we gathered four encryption kinds of shadowsocks traffic to study on the influence of different encryption on shadowsocks traffic. Moreover, we include VPN traffic and do contrast experiment based on four deep-learning models to verify the efficiency of one-dimensional convolutional neural network.