基于DCT域QIM的音频信息伪装算法  

Audio steganography by quantization index modulation in the DCT domain

作　　者：陈铭[1,2,3] 张茹[1,2,3] 刘凡凡[1,2,3] 钮心忻[1,2,3] 杨义先[1,2,3] 

机构地区：[1]北京邮电大学网络与交换技术国家重点实验室信息安全中心,北京100876 [2]北京邮电大学网络与信息攻防技术教育部重点实验室,北京100876 [3]灾备技术国家工程实验室,北京100876

出　　处：《通信学报》2009年第8期105-111,共7页Journal on Communications

基　　金：国家重点基础研究发展计划("973"计划)基金资助项目(2007CB311203);国家自然科学基金资助项目(60821001;90604022);高等学校博士学科点专项科研基金资助项目(20070013007)~~

摘　　要：音频与图像相比具有信息冗余大、随机性强的特点,在音频中实现无误码的信息提取的难度更大。提出一种基于DCT域QIM(quantization index modulation)的音频信息伪装算法,算法特点如下:应用QIM原理,以量化的方式嵌入信息,根据量化区间与信息比特的映射关系提取信息,可实现盲提取;采用改进的QIM方案,针对信息提取的误码,在嵌入端与提取端进行容错处理,保证了隐藏信息的强顽健性;隐藏容量大,可达357.6bit/s。实验表明,算法与传统QIM方法相比具有更好的不可感知性,100%嵌入的载密音频的信噪比在30dB以上,并且对于MP3压缩、重量化、重采样、低通滤波等攻击具有很强的顽健性,同时算法运算量小,易于实现,实用性强。Audio signal has more information redundancy than image signal, but its randomicity is stronger. It is more challenge to extract embedded bit with non-error from audio. A novel audio steganography based on QIM （quantization index modulation） in the DCT （discrete cosine transform） domain was proposed. The merits of the algorithm were as fol- lows： based on the principle of QIM, the message was embedded by quantization method and was extracted by the map- ping between message bit and quantization interval. The extraction was executed blindly; the robustness of the hidden message was ensured well by the improved QIM, which employed the error-tolerance operation during embedding and extracting; the capacity was large as 357.6bit/s. The experimental results show that the proposed method achieves better imperceptivity than normal QIM and the SNR （signal to noise ratio） of the stego audio with the embedding rate of 100% is above 30dB. The embedded message has good robustness against MP3 compression, requantization, resampling and lowpass filtering. Moreover, the algorithm is easy to implement and highly practicable.

关 键 词：信息隐藏 音频信息伪装 QIM DCT 强顽健性 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]