Detection of Double Compression for HEVC Videos With Fake Bitrate

Abstract

Bitrate is an important criterion for digital video quality. Forgers prefer to up-convert bitrate of videos without improving the video quality at all, especially for videos over the Internet. Therefore, proposing an effective algorithm to expose fake bitrate videos becomes an important issue in digital video forensics. In this paper, a novel method based on prediction unit (PU) partition types is proposed to detect high efficient video coding (HEVC) videos with fake bitrate. PU partition type is one unique syntactic unit of HEVC and could be reflected by compression history, while it has not been addressed in the topic of fake bitrate detection. The proposed method adopts support vector machine classifiers with feature vectors formed by histograms of PU partition types in the first P-frame in each group of pictures (GOPs). The performance of the algorithm is compared with state-of-the-art algorithms on videos of various resolutions and bitrates. The experimental results show that the proposed method can identify the fake bitrate videos with high accuracy, and is robust to frame-deletion, copy-paste, and shifted GOP structure attacks.