MP3Stego是经典的MP3音频隐写算法之一。通过分析MP3Stego隐写算法对编码器内循环模块的影响,发现哈夫曼码表索引值在隐写前后发生了不同程度的改变。在此基础上,从待检测的MP3音频的解码参数中提取Huffman码表索引值,计算其二阶差分值,将其作为隐写分析的特征,结合SVM支持向量机实现隐写分析。实验结果表明,所提取的特征能够有效地反映MP3Stego算法在不同嵌入速率下的隐写痕迹。
MP3Stego is a typical steganographic algorithm for MP3 audio. By analysing the influence of the MP3Stego made to inner loop of MP3 encoder, it is found that the index values of Huffman table change differently after embedding. In the proposed algorithm, the index values of Huffman table are extracted from the decoder parameters. The second-order difference of the values is calculated as the steganalysis feature and SVM is used to classify the cover and stego MP3 audios. The experimental results show that the proposed algorithm is effective for detecting MP3Stego.