Robustness Evaluation of Spread-Spectrum Watermarking against Collusion Attack
Robustness Evaluation of Spread-Spectrum Watermarking against Collusion Attack
复制标题
扩频水印对抗共谋攻击的鲁棒性评估
DOI：
10.1143/jpsj.81.035005
复制
发表时间：
2012
影响因子：
1.7
通讯作者：
Mimura,K.
中科院分区：
文献类型：
--
作者：
Sakai,Y.;Mimura,K.
随着网络技术的进步，许多数字内容，如音频，图像和视频，现在可以很容易地通过互联网访问。数字内容的版权保护已变得越来越重要。近年来，电子水印1）作为该问题的解决方案之一而引起关注。数字水印是一种将用户信息以不可见的方式嵌入到数字内容中的技术。到目前为止，已经提出了许多种方法作为水印。2、3）我们这里重点介绍基于频谱扩展的方法，4 - 6）它可以将信息嵌入到内容的整个区域中，并且可以将信息添加到带水印的内容中。所有从水印内容中移除嵌入信息的操作都被称为攻击，例如剪切、缩放、旋转、加性噪声和JPEG压缩攻击等。如果用户的信息没有被嵌入，恶意用户可以很容易地复制数字内容并重新分发它。因此，将用户的信息存储到内容中是非常重要的，特别是在指纹的背景下。最严重的攻击之一是共谋攻击，使得具有相同内容的几个用户联合收割机组合他们的副本以删除嵌入的信息。7，8）水印和指纹需要抵抗这种攻击的鲁棒性。即，鲁棒性是能够保持嵌入的信息不受攻击，并且通过从加水印或加指纹的内容中提取的信息的比特错误率来评估。在本文中，我们特别关注使用信噪比分析5，9）的共谋攻击的效果，并分析基于干扰抵消方法的检测算法的性能。9，10）我们在这里考虑由N个像素组成的256级灰度图像f º Ω f f 2FN作为内容，其中F º Ω 0：255 ä和Ω a：bä，fn 2 Zja n bg。假设一些K比特的消息被嵌入到相同的图像f和K º O N fi中。通过使用分配给每个消息的扩展码fs "; kg k2 Ω 1：K" f "; k" f "1; 1gN将第k比特消息b" b "; k" f "1; 1gK嵌入到原始图像f中来生成第"水印图像f "f" 2 FN。也就是说，第n个加水印图像的第n个像素值被给定为f ′ pPK
As the network technology advances, so many digital contents, such as audio, image and video, can be now easily accessed via the internet. It has become more important to protect the copyright of digital contents. In recent years, the electoronic watermarking 1) attracts attention as one of the solutions to this problem. The watermarking is a method to embed users’ information into the digital contents invisibly. Many kinds of methods have been proposed as the watermarking so far. 2, 3) We here focus on the method based on the spectrum spreading, 4–6) which can embed information into a whole region of the content and can add information into the watermarked content. All operations to remove the embedded information from the watermarked contents are called the attack, eg, the cutout, the scaling, the rotation, the additive noise, and the JPEG compression attacks, etc. If users’ information is not embedded, ill-intentioned users can easily copy digital content and redistribute it. It is therefore important to store users’ information into the contents especially in the context of the fingerprinting. One of the most severe attacks is the collusion attack such that several users having a same content combine their copies to delete the embedded information. 7, 8) The robustness against such attacks is required to the watermarking and the fingerprinting. Namely, the robustness is to be able to keep the embedded information against the attacks, and is evaluated by the bit error rate of the extracted information from the watermarked or fingerprinted contents. In this note, we especially focus on the effect of the collusion attack using the signalto-noise analysis 5, 9) and analyze the performance of the detection algorithm based on the interference cancelling method. 9, 10)We here consider a 256-level gray-scale image f º f fi2FN consisting of N pixels as the content, where F º Ω0: 255ä and Ωa: bä, fn 2 Zja n bg. Suppose that the some K-bit messages are embedded into the same image f and K º O Nfi. The ‘th watermarked image f ‘º f‘fi 2 FN is generated by embedding the ‘th K-bit message b ‘º b ‘; kfi 2 f¿ 1; 1gK into the original image f using spreading codes fs ‘; kg k2Ω1: Kä º f s ‘; kfig 2 f¿ 1; 1gN that are assinged to each message. Namely, the th pixel value of the ‘th watermarked image is given as f‘º f PK