YouTube Video Stream Correlation And Transmission Mode Identification On The Encrypted Traffic From Mobile Devices

In recent years,the demand for video services has grown rapidly.YouTube,one of the most popular Internet video services providers,has more than 100 million users.Users watch more than a million hours of video on YouTube every day,half of which comes from mobile devices.Higher requirements of video service have been asked for by users and the user experience is increasingly being valued.Service industry has gradually considered it as the core competitiveness and the primary means of profit.Video service,as the operator's basic business,is the hottest service around the world.Great video watching experience,is the the key to win the game for the operator.Accordingly,to quantify the user experience and then optimize the network planning,improve the quality of service has become the economic profit point to the operators and the equipment providers.Since YouTube applies HTTPS to protect user privacy and security,many methods based on deep packet inspection methods to obtain QoE measurement parameters have been gradually ineffective.To get this information from HTTPS traffic has become a hotspot in the field of traffic identification.The main contents about this article are as follows:1.YouTube video stream identification:YouTube video stream identification is the foundation for stream association and transmission mode recognition.This article details the server information that YouTutbe video server exposes in DNS,TLS handshaking.Based on the above work,the YouTube video stream identification scheme is proposed and implemented in this paper.Experiments have been given to proves that this scheme is practical and effective.2.YouTube video stream association:In order to correlate adaptive encrypted video streams from the same video,this article designed and implemented the YouTube mobile video streaming associated system based on TLS Resumption.The current HTTP stream association method does not apply to HTTPS flows because the application layer information can not be obtained after encryption,such as HTTP header information.The regulation of time inverval factor has been also elininated by adaptive video streaming,which increased the difficulty of association.Experiments show that the systems proposed by this artical can be accurate,fine-grained association of YouTube video traffic.3.YouTube encrypted video streaming mode identification:By studying the adaptive transmission mode strategy,using man-in-the-middle attack,the paper proposes two kinds of recognition algorithms.The YouTube encrypted video streaming transimtion mode can be identifled by the number of concurrent connections based on the naive Bayesian polynomial model.In the experiment.Another solution is based on the A-I-P-FP identification method.It is only necessary to extract the four characteristics of a little bit of packets from the first two streams of the video.This can quickly and accurately identify the video transmission mode from different mobile devices and browsers to YouTube with an accuracy over 90%.