计算机应用 ›› 2010, Vol. 30 ›› Issue (2): 517-520.

• 信息安全 • 上一篇    下一篇

一类基于奇异值分解的图像水印算法伪验证分析

赵星阳1,孙继银2   

  1. 1. 陕西西安第二炮兵工程学院402室
    2. 西安市第二炮兵工程学院指挥自动化系
  • 收稿日期:2009-08-25 修回日期:2009-10-27 发布日期:2010-02-10 出版日期:2010-02-01
  • 通讯作者: 赵星阳

Fake verification analysis of SVD-based image watermarking

  • Received:2009-08-25 Revised:2009-10-27 Online:2010-02-10 Published:2010-02-01

摘要: 通过分析发现,一类基于奇异值分解的图像水印算法存在致命漏洞,即利用在水印嵌入过程中生成的密钥可以从其他图像(含未嵌入水印的图像)中提取出高相关的水印信息,称这一新发现的算法漏洞为水印算法的伪验证,并由此首次提出了水印算法伪验证的定义和判定条件。然后通过实例分析,指出基于奇异值分解的水印算法之所以存在伪验证,是由于奇异值分解使得水印信息主要包含于正交阵U、V中,因此水印的提取与正交阵U、V以及奇异值的分布类型有关,而与奇异值的具体取值无关。

关键词: 奇异值分解, 图像水印, 伪验证

Abstract: A fatal bug of a kind of Singular Value Decomposition (SVD)-based image watermarking was discovered in this paper. That is, using the same secret key which is produced in the embedding procedure, highly correlative watermark information can be extracted from other different images, even including non-watermarked images. This newly discovered bug was named as fake verification of watermark algorithm. At the same time, the definition and judging condition of fake verification were also presented. Then, the experimental results reveal that the watermark information is involved in the orthogonal matrix U,V after SVD. And the extracted watermark information is mainly involved in U,V, and the distribution of single value instead of exact single value. That is the reason why the SVD-based watermark algorithm has the bug of fake verification.

Key words: Single Value Decomposition (SVD), image watermarking, fake verification