A Novel Identification Method for Frequently Switched Encrypted Video Streaming in Complex Network Environment

Abstract

Nowadays, to provide stable video streaming and smooth user experience among different scenarios, more and more online video sharing and social media platforms adopt the Dynamic Adaptive Streaming over HTTP (DASH), which can transmit and play videos of appropriate quality levels according to the user's network conditions. For sake of illegal videos spreading on the platforms, the identification of encrypted DASH video streaming is of great importance. However, related works on the identification of encrypted DASH video streaming are seldom concerned with the streaming under poor network conditions that the video quality levels are switched relatively frequently. Therefore, we propose a novel video fingerprint extraction method as well as a novel multi-layer fuzzy matching scheme. So far as we know, we are the first to solve the identification of encrypted DASH video streaming under poor network conditions. And to reduce the overhead we establish a local video fingerprint database using the fingerprint extraction method. Our scheme can identify the encrypted DASH video streaming of frequently switched quality levels using very short-time network traffic sniffing. Experiments show that the time required for traffic sniffing is just 6 seconds and the accuracy exceeds 98%, which is significantly better than state-of-art methods. Besides, our scheme performs well on the identification of streaming in multiple scenarios.