Research On Resolution Identification Of Encrypted Video Streaming Based On HTTP/2 Features

Video resolution,as a direct reflection of video image quality,is an important factor that affects the users’ experience.If the fine-grained identification of resolution can be accurately realized by ISP(Internet Service Provider),the changes of the video image quality can be accurately described,and then the video Qo E(Quality of Experience)evaluation and the monitoring of the overall network condition can be carried out,which plays an important role in improving the quality of network service.However,encryption technology is widely used in video traffic at present,which makes the traditional DPI-based methods no longer feasible.Although there are some analysis methods for encrypted video traffic,most of them are based on the encrypted video traffic using HTTP/1.1 protocol.However,HTTP/2 protocol is not a simple upgrade of HTTP/1.1.It uses a series of new methods to improve transmission efficiency,making the existing methods based on HTTP/1.1 characteristics cannot be applied to the encrypted video traffic using HTTP/2.To solve the above problems,this thesis proposes a new fine-grained identification method for video resolution after an in-depth analysis of transmission characteristics of video traffic under HTTP/2,which mainly includes the following research contents:(1)An anti-interference method of video fingerprint restoration is proposed.Considering the segmented transmission mechanism of DASH,this method takes the length of segments as the video fingerprint.In view of the characteristics of multiplexing in HTTP/2,the concept of the composite fingerprint is introduced on the basis of the basic fingerprint.By analyzing the interference that may be introduced by various network protocols during the encrypted transmission of video segments,two anti-interference correction models are trained for TLS protocol and HTTP/2 protocol respectively.On this basis,the accurate restoration of video fingerprints is realized.The experimental results show that the method of fingerprint restoration proposed in this thesis has a high accuracy rate,with the error of more than 98% of fingerprints being within 0.25%.(2)A resolution identification method based on video fingerprints is proposed.This method uses the restored fingerprint as the recognition unit.After the restored fingerprint to be identified is matched in the plaintext fingerprint database,the hidden Markov model is used to process the matching results and obtain the final resolution identification result.For video data of different playback times,this thesis designs two processing methods for fingerprint units respectively: For the video traffic with short playback time,the fingerprint processing method containing all composite fingerprints is adopted.While for video traffic with long playback time,the fingerprint processing method of dividing pseudo mixed data is proposed to simplify the space of the fingerprint database.The experimental results show that the resolution identification method proposed can perform fine-grained resolution identification with an accuracy high than 98%.In order to highlight the advantages of the proposed method,two existing resolution identification methods are reproduced and compared with the proposed method on the same dataset.The experimental results confirm the effectiveness of the method proposed in this thesis.(3)Based on the above methods,a prototype system of resolution identification of encrypted video traffic under HTTP/2 is designed and implemented.The system mainly includes three modules: plaintext fingerprint database construction module,fingerprint anti-interference restoration module and resolution identification module.The system provides a display interface for users.It also supports the display of output results of each module,the evaluation of identification results and the preservation of key information.