Enabling Temporal Bit Rate Adaptation in Encrypted Video Streams

In practice, video streaming over wireless network through home gateways requires bandwidth adaptation methods to deal with the fluctuation bandwidth the network. A popular method to compensate for temporary insufficient data throughput is dropping of B-frames. However, this method cannot be applied for encrypted commercial content, because frame types cannot be identified due to encryption of frame headers. Dropping I-frames or P-frames is not an option, because the video quality would drop dramatically. We developed a new approach using a sliding window to estimate the types of frames in an encrypted video stream. We validate the approach with probability calculations and extensive testing to demonstrate feasibility. Based on a large set of video sequences, our proposed algorithm drops the right frames in 99.99% of the cases.