基于量化索引调制QIM的G.723.1语音信息隐写分析方法

发布时间：2018-05-03 03:33

本文选题：信息隐藏 + 隐写分析　；参考：《中国民航大学》2017年硕士论文

【摘要】：随着信息隐藏技术的快速发展,隐写术作为其主要分支在网络通信中发挥着越来越重要的作用,但隐写术如果被不法分子所利用将会造成严重的安全隐患。隐写分析技术通过检测载体中是否含有秘密信息,阻断非法信息的传递,从而对抗隐写术的非法使用,保证通信的安全性。本文以G.723.1语音编码器为载体,对基于量化索引调制(QIM,Quantization Index Modulation)的隐写方法进行检测。首先,在研究矢量量化过程中QIM隐写前后码字索引值发生变化的基础上,分析索引值出现的概率以及相邻索引值之间的相互影响;然后基于索引值分布特性的变化,通过索引的分布概率矩阵和转移概率矩阵对其进行量化,使用主成分分析法进行降维处理,得到维数较低的特征向量,并通过实验验证了降维后的向量仍对QIM隐写灵敏;最后,将提取的特征向量作为支持向量机(SVM,Support Vector Machine)分类器的输入,针对不同类别的语音样本进行实验,通过大量样本训练分类器,用以检测QIM隐写。针对所提方法的准确性和可靠性进行了实验,实验结果表明对分布概率矩阵和转移概率矩这两类特征向量,检测率均高于90%,虚警率均低于8%,说明该方法准确性较高;全局检测率均在90%以上,说明该方法具有较好的可靠性。针对不同嵌入率和不同时长语音样本进行了实验,实验结果表明语音时长在3s以上时,对分布概率矩阵和转移概率矩这两类特征向量的检测率均达到85%,而分布概率矩阵对时长变化更加敏感;随着嵌入率的增长,检测准确率变高。
[Abstract]:With the rapid development of information hiding technology, steganography plays a more and more important role in network communication, but the use of steganography will cause serious security risks if it is used by illegal elements. The illegal use of anti steganography ensures the security of communication. This paper uses G.723.1 speech coder as the carrier to detect the steganography based on quantized index modulation (QIM, Quantization Index Modulation). First, the index value appears on the basis of the change of the index value of the codeword before and after the study of the QIM steganography in the vector quantization process. The probability and the interaction between the adjacent index values, and then quantizing it based on the distribution probability matrix of index and the transfer probability matrix based on the change of the index distribution characteristics, and use the principal component analysis to reduce the dimension of the dimension by using the principal component analysis method, and verify that the vector of the reduced dimension is still QIM hidden through the experiment. In the end, the extracted feature vectors are used as the input of the SVM (Support Vector Machine) classifier. Experiments are carried out for different class of speech samples, and a large number of samples are used to train the classifier to detect the QIM steganography. The accuracy and reliability of the proposed method are tested. The experimental results show that the distribution is generally distributed. The rate matrix and the transfer probability moment are all two kinds of eigenvectors, the detection rate is higher than 90%, the false alarm rate is lower than 8%, which indicates that the method is more accurate and the global detection rate is above 90%. It shows that the method has good reliability. The experiment is carried out for different embedding rates and different simultaneous long speech samples, and the experimental results show that the speech is longer than 3S. The detection rate of two types of eigenvectors, the distribution probability matrix and the transfer probability moment, is up to 85%, while the distribution probability matrix is more sensitive to the change of the length of time, and the detection accuracy becomes higher with the increase of the embedding rate.

【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP309

【参考文献】