Real-Time Platform Identification of VPN Video Streaming Based on Side-Channel Attack

The video platforms that users watch leak the privacy of their preferences. More and more video streaming is being encrypted to protect users’ privacy. In addition, many users use VPN to enhance their privacy protection further. VPN makes video platform identification challenging because it poses traffic obfuscation and further data encryption. Although the segment-based transmission mechanism and Variable Bit-Rate encoding in HAS make network video traffic show still identifiable patterns, most existing work cannot distinguish different platforms due to the similarity of video streaming. Therefore, we propose a traffic-based side-channel attack method to identify VPN video streaming platforms in real time. The aggregated feature sequence of the unidirectional video streaming is extracted to significantly retain the characteristics of different video platforms. Experiments on 10Gbps backbone background traffic show that the F1-score of the method exceeds 97% and can be processed in real time. In addition, we verify the method’s robustness on datasets with different path features and encryption techniques. A comparison with similar methods shows that our method only requires 1/1260 of the storage and 1/60 of the processing time to identify accurately.