一种针对HEVC预测模式隐写的检测算法

A prediction mode steganalysis detection algorithm for HEVC

针对高效视频编码(HEVC)帧内预测模式的信息隐藏算法,提出了一种隐写检测算法。基于帧内预测模式的信息隐藏算法,通过修改原始视频的帧内预测模式完成隐秘信息的嵌入。实验表明,这种操作会导致部分帧内预测单元(PU)的大小(4×4、8×8、16×16、32×32和64×64)发生改变。对隐写视频进行相同参数的重压缩后发现,不同大小的PU在数量上有向原始视频恢复的倾向,而原始视频重压缩后PU数量变化很小。定义了视频重压缩前后各尺寸PU数量及占有比例的变化率,并以此以作为特征,结合支持向量机(SVM)分类器进行视频信息隐藏检测。实验结果表明,本文算法简单、高效,而且对不同分辨率和不同量化参数的视频均有良好的检测效果。

In this paper, for detecting secret information through modifying HEVC intra-prediction mode, a steganalysis algorithm is proposed. In the implementation process of these steganographic algorithms, a large number of intra-prediction modes are modified. This operation leads to changes in the size of prediction Unit （PU） ,including the sizes of 4×4,8×8,16×16,32×32 and 64×64. We re-compress the original video and the video with secret information using the same configuration parameters. It＇s found that the changes of PU amount for the stego video after re-compression are larger than those for original video without steganography. In the experiment, we firstly calculate the amount of PU different sizes with before and after re-compression. Then the change rate is defined about the amount of various PU and the proportion of various PU. The change rate before and after the re-compression are taken as effective features and sent to support vector machine （SVM） classifier to finish the detection of hiding information. Experimental results demonstrate that the algorithm is simple and efficient. Moreover, it is suitable for videos with different resolutions and different quantization parameters.