【作者】 刘绪崇；

【导师】 陈建二； 王建新；

【作者基本信息】 中南大学 ， 计算机应用技术， 2010， 博士

【摘要】 在信息安全和版权保护的驱动下,信息隐藏技术在近几年得到很大的发展,已经在人类生活的许多方面得到广泛应用。目前信息隐藏技术的研究主要集中在隐秘术和数字水印两个方面,隐秘术是研究如何将秘密信息隐藏在不太容易引起注意的消息中,从而使得秘密通信不被觉察；数字水印源于数字媒体作品的版权保护,都已成为信息安全领域的最新研究热点。本文从分析传统信息隐藏算法的缺陷和不足出发,有针对性地提出新的信息隐藏算法,并从算法的设计、隐藏信息的检测、容量大小和抗攻击性等方面做了很多研究,主要研究工作包括：由于传统的数字水印算法试验都是采用逐点比对隐藏信息来实现,将相似度达到某个阈值作为水印存在的判据,容易导致误判和降低水印识别率,而且算法过于复杂难以实现。针对这些缺陷,提出了基于二代Bandelet图像认证水印算法和基于方向小波几何图像认证算法,并对基于局部统计模型自适应噪声强度算法进行改进。前者应用第二代Bandelet变换生成图像的几何方向流来刻画图像的纹理和几何特征,后者是利用方向小波变换生成图像的几何方向流来刻画图像纹理和几何特征,上述两个算法都是利用改进的统计模型自适应噪声强度算法来筛选出纹理丰富的子图,用于信息隐藏。实验表明上述两个算法不仅能将图像的方向流信息隐藏并检测出,而且还能够抗模糊、JEPG压缩、噪声等典型攻击。针对目前已有的纹理图像隐藏DEM数据的算法会对纹理图像产生破坏的缺点,提出了基于小波变换的纹理无损隐藏DEM数据算法。该算法应用基于整数小波变换和极低比特率压缩编码算法,采用直方图平移的无损隐藏算法将压缩的高程数据隐藏到纹理图像中,在提取隐藏信息时采用直方图平移算法可逆地将隐藏在纹理图中的DEM数据提取,并实现了纹理图的无损恢复。根据DEM数据的特点,提出了基于方向小波的无损三维地形数据水印签名技术和基于DEM数据可见三维水印技术。前者是利用方向小波变换来检测三维地形的趋势和走向,将其作为认证水印并隐藏其中,通过修改广义直方图的平移算法实现水印信息的隐藏。后者是将三维水印与DEM数据进行三维融合,并记录被三维水印遮蔽的DEM数据信息,利用广义直方图平移方法隐藏了被三维水印遮蔽的信息,然后通过广义直方图平移算法可逆地提取出被三维水印遮蔽的信息,用来替换三维水印恢复出原始DEM数据。实验表明方向小波流能很好地刻画出三维DEM数据地形的特征,在保证较高的峰值性噪比(>40dB)时能嵌入较大隐藏信息量(>16kb),在保证DEM地形数据纹理足够丰富时,可见三维水印越大,隐藏的信息量也就越大。为了解决水印与载体融合后视觉不太和谐、信息隐藏量小、可见水印容易破坏载体,以及算法本身安全的问题,提出了基于小波系数相对模糊关系的水印算法和基于模糊关系的无损可见水印。前者是采用带参数的整数小波变换获取载体的小波系数,应用Rabin方法构造的函数生成模糊关系矩阵,通过近似分量和对角分量的量化噪声余量矩阵的相对关系修改对角分量小波系数,最后对小波系数做小波逆变换得到隐含水印的图像。实验证明,该算法在嵌入水印后图像质量仍然很高,具有较强抗攻击能力,在水印提取过程中不需要原始图像的参与。后者是先将载体与可见水印进行模糊融合,然后利用RH算法对遮蔽子图进行加密,最后采用改进的直方图平移方法实现信息隐藏和恢复。该算法具有可见水印的无损自恢复性,不需要可见水印信息就可完全无损地恢复原始图像。由于算法是采用Rinjindael构造的Hash算子加密遮蔽子图,将1位扰动扩散到整个图像,具有较高的安全性。更多还原

【Abstract】 To protect security and copyright of information, the information hiding technology has been developed greatly in recent years. It has been used widely in many aspects of human life. The research on it currently focused on two things:steganography and digital watermark. The first one is to study how to insert confidential information into messages secretly which will not draw people’s attention and make the secret communication hard to be detected. The latter one is proposed for the copyright privacy of digital media. Both of them are the latest hotspot of information security fields.Based on the analysis of the defects of the traditional information hiding algorithms, new methods for solving these problems are proposed in this paper. A series of researches on the design of algorithms, detection of the hidden information, its capacity and the ability for defending attacks and so on, have been done. The main original works include:As all tests of the traditional digital watermark method are executed by contrasting embedded information point by point. The watermark is detected based on the calculation of the similarity. A watermark is proved to be existed when the similarity reaches a certain threshold. However, it easily leads to wrong judgment and reduction of the discrimination. Furthermore, it is too complex to be realized. Considering these shortcomings, the watermark algorithm for image authentication based on the second generation Bandelet and the one for geometric image authentication based on directional wavelet are proposed here. Moreover, the adaptive noise intensity algorithm based on the statistical models is improved. The former depicts the texture and geometric features of the image by adopting the second generation Bandelet transformation to generate geometric direction flow of the image. The latter realizes the goal by using directional wavelet transformation to generate the flow. Both of the two methods take advantage of the adaptive noise intensity algorithm, to filter out subgraphs owning rich texture for hiding information. The experimental results show that these two algorithms are not only able to embed the information of directional flow of image confidently and then detect it, but also withstand the typical attacks such as fuzzy, JEPG compression, noise and others.To fix the weakness that the present texture DEM data hiding algorithms damage that texture, the texture DEM data hiding algorithm without data lost, which is based on wavelet, is proposed. It uses the wavelet-based method of 3d DEM data compression of extreme small bit rate, adopts histogram-shift non-destructive hidden algorithm to embed the compressed elevation data into the texture image, and extracts the hidden DEM data from the image with the reversible histogram-shift algorithm. Moreover, a lossless recovery of the texture image is realized.According to the characteristic of DEM data, the non-destructive three-dimensional terrain data watermark signature technology based on directional wavelet, and visible three-dimensional watermark technology based on DEM data are developed. The former is to detect the trend and direction of the 3d terrain by the means of directional wavelet transformation, then hides them as authentication watermark, and inserts the data by modifying the translation algorithm of generalized histogram. The latter is the three-dimensional integration of three-dimensional watermark and DEM data, which embeds the three-dimensional watermark information into DEM data by the translation algorithm of histogram, meanwhile recording the information marked by three-dimensional watermark, replaces three-dimensional watermark with the information marked by three-dimensional watermark, which is extracted by the reversible algorithm of generalized histogram, and then recovers the original DEM data. The experimental results show that the wavelet can describe the feature of three-dimensional DEM data terrain well, when ensuring a high peak noise ratio (>40dB) and a great amount of hidden information can be embedded (>16kb) with the guarantee that the DEM ground texture data is plenty enough. The larger the visible three-dimensional watermark is, the greater the amount of hidden information is.To improve visual effects of fused image, and provide illegal user removing the visible watermark, the relative fuzzy relation based on wavelet coefficient and the lossless visible watermark based on fuzzy relation are proposed. Fuzzy relationship matrix is constructed by applying Rabin’s function. And diagonal components and approximate components modulo the weight of the quantization noise of wavelet. Relationship of remainder is used to embed watermark. The experiments show that, for the former, its quality is still very high when watermarks are embedded, what’s more, the watermarking algorithm has strong attack-resistance ability and the original image doesn’t need to be participated in the process of watermark extraction. The later, carrier fused with visible watermark firstly. Secondly, encrypt the block subgrah by applying RH function. Finally, an improved histogram shifting method achieves information hiding and recovery. Because the algorithm adopts Hash operator encipher shading sub-graph which is constructed by Rinj indael, it lets 1 bit diffuse into the whole image, thereby it can protect the image data, so this algorithm has highly security.更多还原