【作者】 李岭；

【导师】 路文焕；

【作者基本信息】 天津大学 ， 软件工程， 2020， 硕士

【摘要】 近年来,随着计算机技术的发展,隐蔽通信已经成为一个比较活跃的话题。在本研究中,我们通过利用数字音频水印技术来实现隐蔽通信。数字水印是一种有效的军事领域秘密通信的技术,可以解决传统加密中存在的问题。它是将具有特定含义的信息嵌入到原始信号中而不影响其正常使用的过程。在传输时只需要传输嵌入水印的信号即可。由于信号在传输过程中,可能会遭遇不同程度的有意或者无意的攻击,因此在嵌入水印时,水印算法需要满足一定的鲁棒性来抵抗这些攻击。扩展频谱(Spread Spectrum,SS)算法由于其较好的隐蔽性和抗干扰能力可以很好地满足隐蔽通信过程的要求。然而,传统的SS算法严格基于位置来提取水印信息,因此对于一些可以改变音频长度从而改变水印嵌入位置的去同步攻击存在天然的劣势。本研究在SS算法的基础上,为了应对去同步攻击,提出了一种使用原始音频信号的二阶导数来提取特征点,以提取的特征点为中心的音频段来嵌入水印的方法。去同步攻击包括例如裁剪,均匀裁剪(抖动)和非均匀裁剪,音高不变的拉伸或者缩放以及音高变化的拉伸或者缩放攻击等。实验结果表明,当水印信号经历裁剪攻击时,本文提出的算法可以准确的提取出水印信息。当经过音高不变的拉伸或缩放攻击时,相对于传统的SS算法,本文提出的算法在提取水印的准确率上有较大的提升。此外,与现有的SS算法相比,所提出的方法可以在不影响音频质量的前提下实现较高的嵌入容量。更多还原

【Abstract】 Recently,with the development of computer technology,covert communication has become an active topic.In this study,we use digital audio watermarking technology to implement covert communication.Digital watermarking is an effective technique for covert communication in military area which can solve the problem in traditional encryption.It is the process of embedding hidden information with specific meaning into original signals without affecting its normal usage.During transmission,only the embedded watermark signal needs to be transmitted.Since the signal may encounter different degrees of intentional or unintentional attacks during the transmission process,the watermarking algorithm needs to meet certain robustness to resist these attacks when embedding.Spread Spectrum(SS)algorithm can satisfy the requirements of covert communication process because of its good concealment and anti-interference ability.However,the traditional SS algorithm is based on the position to extract watermark information,so there is a natural defect for against desynchronization attacks that can change the audio length and the watermark embedding position.Based on the SS algorithm,in order to deal with the desynchronization attack,this paper proposes to extract the feature points using the second derivative of the original audio signal,and embed the watermark with the extracted audio points centered on the feature points.Desynchronization attacks include,for example,cropping,uniform cropping(jittering)and non-uniform cropping,pitch-invariant time scale modification and pitch-shifting time scale modification.The experimental results show that the proposed algorithm can accurately extract the watermark information when under the cropping attacks.Compared with the traditional SS algorithm,when experiencing the pitch-invariant time scale modification,the proposed scheme in this paper has a great improvement in the accuracy of watermark extraction.In addition,compared with the existing SS algorithm,the proposed method can achieve higher embedding capacity without affecting the audio quality.更多还原