图像的加密与数字水印技术的研究-智能与分布计算实验室

图像的加密与数字水印技术的研究

姓名	杨双远
论文答辩日期	2004.11.05
论文提交日期	2004.11.17
论文级别	博士
中文题名	图像的加密与数字水印技术的研究
英文题名	Research on Image Encryption and Digital Watermarking Technology
导师1	卢正鼎
导师2
中文关键词	数字图像信息安全;数字图象加密;信息隐藏;数字水印;非对称图像加密;非对称水印系统;多媒体认证
英文关键词	Digital Image Information Security;Digital Image Encryption;Information Hiding;Digital Watermarking;Asymmetric Image Encryption;Asymmetric Watermarking System;Multimedia Authentication
中文文摘	伴随着网络技术和多媒体技术的飞速发展，多媒体数据逐渐成为人们获取信息的重要来源，并成为人们生活的重要组成部分。因而，如何保护多媒体信息的安全成为国际上研究的热门课题。对于这个问题，多媒体加密技术以及水印技术给出了两种可能的解决方案。传统加解密系统，对于文件的处理是将其加密成密文再进行传送，使得在网络传递过程中的非法拦截者无法从中获取机密信息，从而达到保密的目的。对于数据量极为庞大的多媒体数据流而言，传统的加密系统难以进行快速加解密，所以应结合多媒体信息的本身特点，研究适合于多媒体信息的加密技术。依据图像信息的矩阵存储格式，提出了两种基于矩阵变换的非对称图象加密技术。充分利用混沌系统具有形式简单，初始条件的敏感性和具备白噪声的统计特性等诸多特点，将混沌理论引入图象加密技术中，这不仅能够改变图象像素的空间分布和灰度分布，并且可以容忍对加密图象进行的一定的失真处理，如叠加噪声、JPEG有损压缩等，与此同时还可以定位出可能篡改的位置，提高了图象加密技术的鲁棒性。数字水印技术和多媒体加密技术，是两种完全不同的多媒体保护方式。多媒体加密技术，是通过对多媒体信息进行随机化处理来保证信息的安全性，而数字水印技术是往多媒体信息中添加某些数字信息来确保多媒体信息的版权。在简要探讨了数字水印技术的一般模型和典型算法之后，给出了两种一般数字水印算法。结合目前的JPEG2000标准，充分利用小波变换的多尺度分解特点以及人的视觉特性，提出了一种基于小波的鲁棒水印算法。由于充分利用了图象压缩技术和视觉系统HVS的照度掩蔽特性与纹理掩蔽特性，在实现了嵌入大容量水印信息的同时，算法还具有较好的稳健性。相对于鲁棒性水印来说，半易损水印，并不过度强调水印算法的鲁棒性，而是如何在鲁棒性和易损性之间取得折中。结合分块变换，给出了一种半易损水印系统。本系统不但能够很好的对抗JPEG压缩，而且能够清晰的定位出可能被篡改的区域以及形状，很好的满足了半易损水印的要求。在水印技术发展的今天，已经有了很多优秀的算法。但是和鲁棒性不同，目前的对于安全性的需求，还没有多少算法给予足够的关注。事实上，目前还没有几个算法能够对抗针对水印安全性的攻击。在充分分析了水印安全性的特点以及可能的攻击类型之后，提出了一种新颖的非对称数字水印方案。数据所有者通过注册后，利用私钥向数据中嵌入版权信息，数据使用者可以利用公钥从水印数据中提取版权信息。我们提出的水印系统是安全的，用户无法从公钥中获取任何私钥信息和破坏或删除水印信息等。实验结果也表明本水印算法是有效性。
英文文摘	The digital media has become a main way for information communication along with the wide use of network. At the same time, multimedia information security problems including copyrights piracy emerged as side effects of the popularity of digital representation and distribution over network. Information hiding, also called Steganography, digital watermarking technology and multimedia information encryption methods provide novel ways to solve these security problems for multimedia information. Conventional encryption systems usually encode the original files into the secret information, and then transmit them on the network. Because only the authorized users can decrypt the secret files, the original files are secure against the unauthorized attacks. But, for the multimedia files with a large amount of information, these schemes usually are so slow for encryption and decryption process. So it is important for us to propose some multimedia encryption and decryption systems combining to the character of the multimedia informations themselves. The image information is usually stored in matrices. Two kinds of image encryption schemes base on chaotic sequence and matrix transformation are proposed. Chaotic sequences have several good properties including the ease of their generation, the sensitive dependence on their initial condition and the statistics property of noise. Because the proposed scheme uses block matrix transformation to encrypt images, it can scramble the grayscales and frequency domain. At the same time, it is very robust against distortion operations, for example, JPEG lossy compression, filtering, etc. It is another advantage for the proposed scheme with good property of localization for the possibly changed regions. The schemes satisfy the characters of convenient realization, less computation complexity and better security based on the theory analysis and vast test results. The multimedia encryption shemes usually randomize the original multimedia files to assure the security, but the digital watermarking technique is a different multimedia protection style, which embeds some digital labels or signatures into host multimedia data to guarantee that the copyright of host multimedia data isn’t infringed. After analyzing the ordinary model, typical schemes and security of the digital watermark in detail, two types of ordinary watermarking systems are proposed. One is the robust watermarking system, and the other is the semi-fragile watermarking system. In the former, considering the standard of JPEG2000, a new technique for the adaptive digital watermarking of still image is presented based on the concepts of multi-resolution wavelet fusion and Human Vision System. According to the characteristics of the multi-resolution decomposition of the wavelet transformation, a scheme with use of a wavelet transform to embed vast signature information in different frequency bands of an image is purposed. The semi-fragile watermarking system requests the tradeoff of the robustness and frangibility, not only thinking about the robustness. A novel semi-fragile watermarking technique is proposed for image authentication in this paper based on block-DCT transformation. The proposed sheme has very effective robustness against JPEG compression and good property of localization of possibly changed regions. Currently, many excellent watermarking schemes are proposed. But few authors pay attention to the security of the watermarking system. In fact, most of the proposed watermarking systems aren’t secure. After thinking over the characters of the secure watermarking system and the types of possible attacks, an asymmetric watermarking system is proposed. Data owner can embed the identification into the original data and retrieve it from the watermarked data using his private key, while data user can detect the identification using the public key. The data user can’t forge the private key and destroy or erase the identification that embedded into the original image by using public key. Experiment results show its security and validity.