Title: WeChat traffic classification using machine learning algorithms and comparative analysis of datasets
Authors: Muhammad Shafiq; Xiangzhan Yu; Asif Ali Laghari
Addresses: School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China ' School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China ' School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China
Abstract: In this research paper, we present the first classification study to classify WeChat application service flow traffic (text messages, picture messages, audio call and video call traffic) classification and secondly to find out the effectiveness of big dataset and small dataset as well as to find out effective machine learning classifiers. We firstly capture WeChat traffic and then extract 44 features then we combine capture traffic to make full instance of dataset. Then we make reduce instances of dataset from the full instance of dataset to show the effectiveness of large dataset and small dataset. Then we execute well known machine learning classifiers. Using statistical test, we use Wilcoxon and Friedman statistical test for the datasets and ML classifiers to find more deeply its effectiveness. Experimental results show that reduce instance dataset show high accuracy result compared to full instance and C4.5 classifier perform effectively as compared to other classifiers.
Keywords: WeChat traffic classification; machine learning; audio and video call; text and picture messages; comparison.
International Journal of Information and Computer Security, 2018 Vol.10 No.2/3, pp.109 - 128
Received: 17 Jan 2017
Accepted: 07 May 2017
Published online: 23 Apr 2018 *