【作者】 张湛；

【导师】 王执铨； 戴跃伟；

【作者基本信息】 南京理工大学 ， 控制科学与工程， 2010， 博士

【摘要】 隐写术是信息隐藏研究领域的一个重要分支,主要运用于隐秘通信。隐写将需传输的隐秘信息以尽可能不引起监控者怀疑的方式,嵌入到“无害”的载体信息中,并利用公共传输通道发送给接收方,接收方则从接收到的含密载体中,正确提取出其中隐藏的隐秘信息。相反,隐写分析则作为公共通信的监控者,对各种看似正常的通信行为进行检测和分析,从而发现隐秘通信的存在,甚至进一步提取含密载体中的隐秘信息。隐写和隐写分析技术相互对抗共同发展,均为当前信息安全研究中极为活跃和重要的组成部分,是涉及信息安全、多媒体信号处理以及模式识别等多个学科的交叉课题。隐写的安全性评估以及高安全、低感知失真和高容量的嵌入算法设计是隐写术研究的核心和难点问题。本文对于隐写的高阶统计安全性评价、隐写分析算法设计、高阶统计安全隐写算法设计、隐写编码技术的应用等几个关键的理论和技术问题进行了深入研究,主要研究成果如下：(1)针对隐写高阶统计安全性的评价问题,提出利用数字图像高阶Markov模型,刻画载体图像空域像素间相关性的方法,比较了几种常用图像扫描方式对该模型建立效果的影响,根据该模型建立了原始图像和载密图像的高阶统计测度,并进一步分析了这一测度与传统ε-secure安全性指标的关系。(2)通过分析载体图像空域和DCT系数域的加性和乘性四种扩频隐写算法,对嵌入前后载体像素高阶Markov模型统计分布的影响,利用模型经验矩阵提取载体像素统计分布的特征,运用支持向量机分类器,提出一种针对扩频类隐写的计算规模可调的隐写分析算法。(3)利用载体空域像素Markov模型统计测度公式,对LSB匹配嵌入算法进行优化,提出一种二阶统计安全的LSB匹配优化隐写算法；进而针对大规模隐写的情况,分析LSB匹配隐写对载体空域像素Markov模型统计分布的影响,利用动态补偿的方法提出一种二阶统计安全的快速LSB匹配隐写算法。(4)将图像高阶Markov模型扩展到载体像素预测误差域,基于Markov统计测度提出在载体像素预测误差域二阶统计安全的优化量化隐写算法；并针对大量隐写的情况,提出二阶统计安全的快速量化补偿隐写算法。(5)将载体局部像素复杂度与隐写编码相结合,针对载体各部分的不同特性在一簇隐写编码中选用合适的编码,提出一种基于循环隐写码的自适应图像隐写算法；进而将图像高阶Markov模型和动态补偿的思想,与基于隐写编码的自适应隐写的思路相结合,提出一种兼顾感知失真和统计安全性的自适应隐写算法。最后,论文分析了本文研究中还存在的问题,并指出了进一步研究的方向。更多还原

【Abstract】 Steganography is an important branch of information hiding research, and its main purpose is covert communication. Steganography is to embed secret information which needs to be transmitted into’innocent’cover-object without causing suspicion of monitors, and send the stego-object to the receiver through public transmission channels, and the receiver correctly extract the secret information from the stego-object he received. On the contrary, steganalysis is to analyze vary kinds of’innocent’communication behavior as a communication monitor, and detect the existence of covert communication, even further to extract the secret information being transmitted. Steganography and steganalysis confront each other and co-develop. They are the interdisciplinary subjects of information security, multimedia signal processing, pattern recognition and so on as the very active and important component in the research of information security field. It is the core and difficulty problem to evaluate the security of steganography system and design the embedding algorithms with high-capacity, high-security and low distortion. This dissertation studies deeply the evaluation of high-order statistical security of steganography, the design of steganalysis algorithm, the design of high-order statistical secure steganography algorithm, and the application of cover-codes. The main contributions can be enumerated as follows:Firstly, for the evaluation of the high-order statistical security of steganography, a high-order Markov chain model for digital image steganography is proposed in order to characterize the correlative of the spatial adjacent pixels of cover-image. And the effect of several general image scanning methods for the model is compared. Based on the model a high-order statistical measure between the cover-image and stego-image is proposed. Moreover, the relationship of the measure and the traditionalε-secure criterion is analyzed.Secondly, the effect of four kinds of additive and multiplicative spread spectrum steganography in the cover-images’spatial and DCT domain on the image high-order Markov chain model is studied. Then a kind of steganalysis algorithm, against the spread spectrum image steganography is proposed which extracts image pixels’statistical features based on the empirical matrix of the high-order Markov chain model and uses support vector machine for classification.Thirdly, a LSB match steganography algorithm with second-order statistical security is proposed. It optimizes the embedding algorithm based on the statistical measure of image spatial pixels’Markov chain model. Furthermore, for large numbers of steganography, after analyzing the effect on the statistical distribution of the cover image spatial pixels’Markov chain model causing by LSB match embedding, a fast LSB match steganography algorithm with second-order statistical security is proposed based on the dynamic compensation idea.Fourthly, the image high-order Markov chain model is extended to the prediction-error domain of image pixels. In the prediction-error domain of cover image pixels, a second-order statistical secure optimized quantization steganography algorithm which based on the Markov statistical measure is proposed. Moreover, for large numbers of embedding, a fast compensated quantization embedding algorithm which maintains the second-order statistical distribution of the pixel prediction-error of cover-object is proposed.Finally, a self-adaptive algorithm for image steganography which reasonably uses a cluster of cyclic cover codes based on the local pixels complexity of cover image is proposed. Furthermore, a self-adaptive image steganography algorithm which takes account of the perceptual distortion and second-order statistical security is proposed. The algorithm combines the dynamic compensation method based on image high-order Markov chain model with the self-adaptive method based on cover codes.At last, the deficiencies in the dissertation are summarized, and some open issues in information hiding as well as the future work are presented.更多还原