Abstract: This paper proposed a new semi-fragile audio zero-watermarking algorithm which can be used to authenticate the copyright and content of digital data. This algorithm has the following features: (1) it extracted the low frequency components of host audio to construct zero-watermarking, ensured the imperceptibility of watermarking algorithm, and achieved blind detection; (2) it distributed the pixels of watermarking images rationally by adopting an adaptive audio segmentation frame method, and improved the capability of tampering localization and robustness of regular attacks; (3) it used multilevel scrambling technology to eliminate the correlation of the watermarking images, improved its safety and the robustness toward regular attacks. Meanwhile, this algorithm can not only conduct the integrity authentication, but also locate the tampering area accurately by tampering assessment. The experimental result shows that this algorithm has good robustness toward regular attacks and strong capability of tampering localization toward malicious attacks.

Key words: zero-watermarking, semi-fragile watermarking, content authenticity, tampering location, adaptive frame

摘要： 提出了一种可用于版权和内容认证的半脆弱音频零水印算法，该算法提取载体音频的中低频分量构造零水印，确保了水印算法的不可感知性，并可实现盲检测。采用自适应的音频分帧方法，合理地分配了水印图像的像素点，从而提高了算法的篡改定位能力和对于常规攻击的鲁棒性。同时利用多级置乱技术消除水印图像的相关性，提高了算法的安全性以及对于常规攻击的鲁棒性。算法不但可进行完整性认证，还可以通过篡改评估准确定位篡改区域。实验结果表明，该算法对于常规攻击具有较好的鲁棒性，对恶意攻击还体现了很强的篡改定位能力。

关键词: 零水印, 半脆弱水印, 内容认证, 篡改定位, 自适应分帧