【作者】 谈华斌；

【导师】 朱晓冬；

【作者基本信息】 吉林大学 ， 计算机软件与理论， 2004， 硕士

【摘要】 随着网络和多媒体技术的广泛应用，需要进行加密、认证和版权保护的声像数据也越来越多，传统密码学中的认证方法对多媒体内容的保护已无能为力。数字水印技术在这种背景下提了出来。数字水印技术是指利用信号处理的方法在数字声音、图像或是视频等数字化的多媒体的数据中嵌入隐蔽的信息。通过这些隐藏在多媒体内容中的信息，可以达到确认内容的创建者，购买者或是判断内容是否真实完整的目的。本论文在分析了数字音频水印的特点和方法的基础上，提出了一种数字音频水印应用方案，实现了三种音频水印算法，并分析了这些方案和算法的特点和优缺点。最后对硕士期间所做工作作了总结，并对今后的研究方向作了展望。首先概述了数字音频水印技术，让大家了解什么是数字音频水印技术、数字音频水印的特点和应用等。总结了人类听觉系统的特点，特别是人类听觉的感知特征以及心理声学模型。这主要是考虑到有些数字音频水印算法中用到了这些声学特点，以及从理论的角度考虑音频水印的不可感知性。由于音频水印具有不可感知、密不可分、相同改变等优点，数字音频水印有着广泛的应用。本文在总结音频水印的主要应用的基础上，提出了一个音频水印结合密码学实现版权保护和拷贝控制的应用方案。该方案通过计算音乐作品的版权标示E、序列号S，将E加密为,使用作为密钥对拷贝状态C加密为，接着用,S产生水印嵌入的密钥。最后将,S,分别写回到Wav文件头文件中的INAM,IKEY,ICOP中去。在提取水印时，先读取Wav头文件中的INAM,IKEY,ICOP，得到,S, 。若此文件不含有这三项信息，或者无法解密，则判断此文件为盗版文件，禁止播放。否则利用解密，得到此视频文件的拷贝状态C。再利用,S算出水印密钥，进行水印提<WP=78>取。在对已有的多种数字空域水印的空域嵌入算法研究的基础上，本论文提出了一种新的基于双重密钥空域音频水印的嵌入和提取算法。该算法采用了线性移位寄存器序列作为水印信息，使用扩频、调制和同步的手段来增加水印的鲁棒性。该算法的特点在于：1.使用了公共密钥和私有密钥双重密钥和位置密钥，使得攻击者在没有密钥或仅仅取得一个密钥的情况下无法取得水印信息，增强了保密性；2.使用双重水印，通过公共水印保护了私有水印，增加了私有水印的鲁棒性； 3.使用同步控制，提高水印的抗剪裁、重采样等攻击的鲁棒性；4.仅仅改变音频数据的最低位，保证了良好透明性。试验结果表明，实现的水印算法具有很好的透明性，而且对于常见的滤波和A/D转换等攻击具有较好的鲁棒性。在脆弱水印算法中，我们引入了MD5算法，实现了对音频作品的内容认证。该算法嵌入水印时抽取原音频采样的（S-256）个字节最后一位的数据，组成采样值序列。使用MD5算法对上述采样值序列做运算，生成256位的索引值。以此索引值作为脆弱水印加入到余下的256原音频采样值的最低位中。提取时直接使用与嵌入方法相同的过程，首先获得(S-256)个采样值序列。使用MD5算法对上述序列计算索引值Index。在余下的256个采样值的最低位中提取索引值。若，则原音频信息未受到篡改或攻击。反之，原音频信息一定受到篡改或攻击。在音频频域水印算法中，本文探讨了基于离散小波变换（DWT）的音频水印嵌入和提取算法。由于混沌序列具有良好的平衡性和类似高斯噪声的自相关特性，而且数目众多，易于产生，所以使用混沌序列作为水印序列，水印的随机性和透明性较好。该算法着重叙述使用混沌序列作为水印在DWT的细节分量中嵌入和提取。这一算法具有以下几个特点： <WP=79>1.在提取水印的时候不需要原始音频,完全实现了盲检测。2.采用受控的水印信息分量嵌入方法，保证水印具有很好的鲁棒性。3.对方案中参数的设置合理性进行了充分的试验验证。试验结果表明，该算法有较好的透明性和鲁棒性。为了评价上述算法的性能，我们引入了常用的平滑滤波，低通滤波，高通滤波，高斯噪声，重采样等音频攻击方法。通过这些攻击的试验结果来对水印算法进行检验,其中包括了对水印算法进行透明性性能评价（信噪比和主观平均判分法）和可靠性检测评价（归一化相关度）。在文章的结尾对本论文所做的工作进行了总结，并对今后的研究工作作了展望。更多还原

【Abstract】 The wide application of the multimedia technology has resulted in a strong demand for encryption, authentication and copyright protection for multimedia data. As the authentication ways of tradition encryption is helpless to the protection of multimedia content, the watermark technique is put forward under this condition. Digital watermarking is the technology of embedding secret information into digital multimedia data such as digital audio, image or video using signal processing technology. Making use of the information hiding in the multimedia data can verify the author, purchaser or confirm the multimedia content. In this paper, we advance an applied scheme of digital audio watermarking and implement three audio watermark algorithms on the base of analyzing the characters and method of the audio watermark technology. And we analyze their specialty, merit and deficiency. At last, we conclude the work during my master period, and make a prospect for the future.First, we give a general concept of the audio watermarking. We introduce the audio watermark character and application. And we summarize the character of the human auditory system, especially the psychoacoustic modal. The analysis of this can give the inaudible, redundant elements of audio. In this paper we employ it to guarantee the inaudibility of the embedded watermark.Because of the good properties of intangibility, impartibility and common change, the audio watermark has a comprehensive application. We advance an applied scheme of digital audio watermarking for copyright protection and the copy controlling with the cryptology on the base of summarizing the audio watermark application.First, compute the audio works’ copyright mark E, sequence number S, then we compute the encrypted , and encrypt the copy controlling mark C to whose key is , get watermark key by using and S. And at the last, we write , S and into the INAM, IKEY, ICOP information which are located in the head of Wav file. When extracting the watermark, we first get the , S and . If they are existed or not to be decrypted, then we judge this file as pirate and prohibit its play; else we use to decrypted and get the copy state of this file. Then we compute the watermark key using and S and extract it.A new spatial digital audio watermark algorithm is presented by analyzing the merit and deficiency of all spatial domain watermark algorithm. In the algorithm, we take the linear offset register sequence as the watermark information, and enhance the <WP=81>robustness of watermark by using spread spectrum, mix and synchronization. This algorithm has features as followed：Using public key and private key makes the attacker fail without any of them, enhancing the secrecy; Using double watermarks enhances the robustness of the private watermark by the protection of the public watermark;Using synchronization enhances the robustness of the watermark to resist the attack of compression and resampling. Only changing the least significant bit enhances the transparence.The experiment results tell us that the algorithm has a good transparence and robustness to resist the attacks. In the fragile watermark algorithm, we use MD5 algorithm and implement the authentication of the content. While inserting, we use the last bit of each byte to compose the sampling sequence whose length is the number of primary samples subtracts 256. Then we use MD5 to compute the sampling sequence and get the index value, which is 256 bits. At the last we use the index as fragile watermark to insert the last bit of the residuary 256 primary samples.While extracting, we use the same method as inserting; first we get the sampling sequence whose length is (S-256). Then use MD5 to compute the index and to extract the from the last bit of the residuary 256 samples. If , the audio hasn’t been attacked or tampered.In the audio watermark algorithms of the frequency domain, we discuss the watermark algorithm based on the low frequency of DWT. Because the chaotic sequences have good properties of更多还原