CAJ | 학술논문

文章提出了一种基于MDCT量化系数统计特征的AAC音频隐写分析方法。将AAC音频进行部分解码得到MDCT量化系数，在MDCT量化系数中提取广义高斯分布模型的参数、量化系数分布直方图的频域统计矩、帧内和帧间MDCT量化系数的Markov转移矩阵的部分数据作为隐写分析的特征，最后采用支持向量机进行分类。通过对不同比特率的AAC音频的实验结果表明，文中提出的AAC音频隐写分析方法对于MDCT量化系数中的直接扩频隐写方式的检测效果较好，对于比特率为128 kb／s的AAC音频，在隐写容量较低的情况下也能达到较高的检测率。
문장제출료일충기우MDCT양화계수통계특정적AAC음빈은사분석방법。장AAC음빈진행부분해마득도MDCT양화계수，재MDCT양화계수중제취엄의고사분포모형적삼수、양화계수분포직방도적빈역통계구、정내화정간MDCT양화계수적Markov전이구진적부분수거작위은사분석적특정，최후채용지지향량궤진행분류。통과대불동비특솔적AAC음빈적실험결과표명，문중제출적AAC음빈은사분석방법대우MDCT양화계수중적직접확빈은사방식적검측효과교호，대우비특솔위128 kb／s적AAC음빈，재은사용량교저적정황하야능체도교고적검측솔。
A steganalysis method of AAC audio based on statistical features of MDCT quantized coeffi‐cients is proposed .Firstly ,AAC audio is partly decoded to get MDCT quantized coefficients ,and then the parameters of generalized Gaussian distribution (GGD) model ,the statistical moments in fre‐quency domain of the distribution histogram of quantized coefficients ,and some data of the Markov transition matrix of the MDCT quantized coefficients in a frame and between frames are extracted from the MDCT quantized coefficients as the features of steganalysis .Finally ,the support vector machine is used as a classifier .The experimental results of AAC audio at different bitrates reveal that ,the pro‐posed steganalysis method of AAC audio has a good detection effect on the hiding method of the direct spread spectrum modulation on the MDCT quantized coefficients ,and for the AAC audio at 128 kb/s bitrate ,the detection accuracy is high even under the condition of low capacity of steganography .