Multiple Methods for Wechat Identification

Wechat is a popular social platform developed in 2011 by Tencent. In this paper wechat traffic is analyzed by a novel hybrid method, which combines statistical method, payload-based method, SVM, CRC and deep learning. Firstly, the statistical method is utilized to extract features from wechat packets header, which can classify different wechat applications and functions, e.g. texts, images and voice, and so on. Secondly, payload-based method is used to identify the traffic, which is corresponding to the above functions and application protocols. Thirdly, SVM is applied to categorize the texts based on their attributes. CRC method is used to classify the images, which effectively protects the user's privacy. Finally, deep learning is presented to extract features of wechat app in order to check the malicious software. Experimental results show that, the proposed method has high accuracy for wechat traffic. It not only identifies wechat app, but also detects the specific functions of app. It even discriminates texts, images, voice and malicious software effectively.