一种基于非均匀离散傅立叶变换的鲁棒音频水印算法

摘要

基于离散傅立叶变换(DFT)的音频水印算法对常规信号处理操作具有较高的鲁棒性,然而,在DFT域的固定频率点嵌入水印信息易受频域攻击,导致此类水印算法存在安全隐患.为进一步说明这种安全隐患,本文描述了一种新颖的频域攻击方法,仿真结果表明采用该方法可以在不影响含水印音频信号听觉感知质量的条件下有效去除水印信息.针对上述问题,本文提出了一种基于非均匀离散傅立叶变换(NDFT)的鲁棒音频水印算法.该算法基于NDFT可以任意选择频率点的特性,利用混沌映射随机选取NDFT域的水印嵌入频率点,以实现水印嵌入位置的随机性.此外,引入另一个混沌映射置乱加密待嵌入的水印信息以提高算法抵抗拷贝攻击的能力.理论分析和实验结果表明该算法不仅具有抗常规信号处理操作高的鲁棒性,而且能够抵抗频域的恶意攻击,大的密钥空间保证了系统高安全性.%DFT-based audio watermarking schemes have attracted many eyes because of high robustness against common signal processing operations. However, the inherent limitation of discrete Fourier transform(DFT), which is public and fixed frequency points, results in weak robustness against malicious attack and inferior systematic security. In this paper, the authors propose a novel attack method to hack DFT-based audio watermarking scheme. Attack results show that the watermark embedded into the audio signal could not be extracted while the sense of hearing is less degradation after attacking. In order to overcome the drawback, a novel secure audio watermarking scheme based on nonuniform discrete Fourier transform(NDFT) is proposed in this paper. In the method, the authors utilize nonuniform discrete Fourier transform(NDFT) to randomly select sampling points in frequency domain, which could make the embedding points private. Furthermore, the authors also use two different chaotic maps to improve security. One chaotic map is used to choose the NDFT-domain embedding points for making them unpredictable;the other is used to scramble the watermark to be embedded with purpose of resisting so-called copy attack. Theoretical analysis and simulation results show obviously that the proposed scheme can successfully resist the designed attack method. Moreover, it has higher robustness against malicious attack and superior security.