TLS/SSL Encrypted Traffic Classification with Autoencoder and Convolutional Neural Network
Ying Yang, Cuicui Kang, Gaopeng Gou, Zhen Li, Gang Xiong · 2018
With the increasing demand for privacy protection, the amount of encrypted traffic tremendously raises. Precise traffic analysis and monitoring has become a challenge since the traditional algorithms do not work well any more. To deal with the problem, many researchers extract a number of statistical features and propose some machine learning algorithms on the field of traffic analysis. In this paper, we utilize more distinctive representation of packet length and packet inter-arrival time. Meantime, we propose two deep learning approaches for better feature learning and compare them with the existing state-of-the-art machine learning algorithms. One model is Autoencoder for the purpose of extracting representative features. Another model is Convolutional Neural Network. It learns high dimensional features, improves the accuracy of classification and has been popularly used. The evaluation results show that the Convolutional Neural Network outperformed competing algorithms.