Research Article
Hybrid Deep Neural Network - Hidden Markov Model Based Network Traffic Classification
@INPROCEEDINGS{10.1007/978-3-030-06161-6_59, author={Xincheng Tan and Yi Xie}, title={Hybrid Deep Neural Network - Hidden Markov Model Based Network Traffic Classification}, proceedings={Communications and Networking. 13th EAI International Conference, ChinaCom 2018, Chengdu, China, October 23-25, 2018, Proceedings}, proceedings_a={CHINACOM}, year={2019}, month={1}, keywords={Network traffic Content classification Hidden Markov model Deep neural network}, doi={10.1007/978-3-030-06161-6_59} }- Xincheng Tan
Yi Xie
Year: 2019
Hybrid Deep Neural Network - Hidden Markov Model Based Network Traffic Classification
CHINACOM
Springer
DOI: 10.1007/978-3-030-06161-6_59
Abstract
Traffic classification has been well studied in the past two decades, due to its importance for network management and security defense. However, most of existing work in this area only focuses on protocol identification of network traffic instead of content classification. In this paper, we present a new scheme to distinguish the content type for network traffic. The proposed scheme is based on two simple network-layer features that include relative packet arrival time and packet size. We utilize a new model that combines deep neural network and hidden Markov model to describe the network traffic behavior generated by a given content type. For a given model, deep neural network calculates the posterior probabilities of each hidden state based on given traffic feature sequence; while the hidden Markov model profiles the time-varying dynamic process of the traffic features. We derive the parameter learning algorithm for the proposed model and conduct experiments by using real-world network traffic. Our results show that the proposed approach is able to improve the accuracy of conventional GMM-HMM from 77.66% to 96.11%.
