一种新颖的用于攻击辨别的脆弱语音水印算法  

Novel Fragile Speech Watermarking Scheme for Tamper Discrimination

作　　者：陈攀[1] 何晨[1] 何迪[1] 戴华亮[1] 张理[1] 胡玲娜[1] 

机构地区：[1]上海交通大学电子工程系,上海200240

出　　处：《武汉理工大学学报》2009年第20期125-129,共5页Journal of Wuhan University of Technology

基　　金：国家自然科学基金(60772100)

摘　　要：提出了一种新颖的用于攻击辨别的脆弱语音水印算法。与传统脆弱语音算法相比,算法不仅能够实现内容完整性认证,攻击定位和原始语音恢复,而且还能够实现攻击辨别(Tamper Discrimination),即能够辨别攻击是针对所嵌入的水印,语音载体,还是对两者都进行了攻击。算法在保护录音证词方面具有很强的实用性,能够有效地验证上述3种攻击后录音证词的完整性,从而确保了证词内容的真实性。为保证原始语音信号能够被更精确地恢复,算法首先对原始语音信号的低频小波系数进行A律量化,然后由量化系数产生水印信号,最后将置乱后的水印嵌入在原始语音信号中随机选出系数的LSB(Least Significant Bits)位。仿真结果表明算法能够很好地实现攻击辨别和攻击定位,有效地证明了方案的可行性。In this paper, a novel fragile speech watermarking scheme for Tamper Discrimination is presented. Compared with the traditional fragile speech watermarking schemes, the proposed scheme can not only be used for content authentication, tamper localization and original speech signal reconstruction, but also for Tamper Discrimination. That is, it can distinguish whether the watermark, the speech or both of them are tampered. One important application of this scheme is testimony record protection. It can effectively authenticate the integrity of the testimony record and ensure the authenticity of the record content even if it is tampered in the above 3 ways. In order to recover the original speech signal more accurately, A-law algorithm is applied for the quantization of the DWT low-frequency sub-band coefficients of the original speech signal. Using these quantization coefficients, the watermark signal is generated. Finally, the scrambled watermark bits are embedded into the LSBs （Least Significant Bits） of the randomly selected coefficients of the original speech signal. Simulation results show the good performance of the proposed .scheme on tamper discrimination and tamper localization, which demonstrates the feasibility of the scheme.

关 键 词：脆弱水印 攻击定位 攻击辨别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]