一种面向AMR的音频隐写分析算法

一种面向AMR的音频隐写分析算法 A Novel Audio Steganalysis Scheme for AMR 任延珍 REN Yanzhen 1 first-author 蔡婷婷 CAI Tingting 1 王丽娜 WANG Lina 1 武汉大学计算机学院 武汉大学计算机学院 School of Computer,Wuhan University School of Computer,Wuhan University 针对现有的对固定码本搜索得到的非零脉冲位置进行调制来嵌入秘密信息的AMR（adaptive multirate）隐写算法,提出一种相同脉冲位置概率（SPP）特征的隐写分析方法.由于载体音频和载密音频在固定码本编码过程中,同轨道两个脉冲位置选取策略的不同,因此可以在解码端计算每一轨道第一个和第二个脉冲在整段音频信号中选择同一位置的概率值,以此作为分类特征并用支持向量机SVM进行训练分类.实验结果表明,本文算法在相对嵌入率高于30%时,检测正确率高于85%,对载体音频的检测正确率可以达到90%以上. For those existing AMR（adaptive multi-rate）steganographic algorithms which embed secret messages in the non-zero pulse positions obtained by AMR fixed codebook（FCB）search,a steganalysis method of same position probability（SPP）is proposed.By analyzing the different choosing strategies of two pulse positions in same track during the fixed codebook coding process of cover and stego audio,the probability of the first and the second pulse choosing the same positions for each track in the decoder of whole audio signal is calculated as distinguishing features,and support vector machine（SVM）classifiers are employed to train and classify.Experimental results show that the correct detection rate of our proposed method is above 85% when the embedding bit rate is 30%,and can reach above90%for cover audios. AMR 隐写 隐写分析 固定码本搜索 AMR（adaptive multi-rate） steganography steganalysis fixed codebook search 10.14188/j.1671-8836.2014.06.015 TP309 国家自然科学基金资助项目（61373169） 2014-06-01 2021-04-01 6 60