【摘要】 指出一种可抵抗敏感性攻击的水印方案仍然存在安全漏洞.攻击者对2个不同的临界信号分别实施敏感性攻击得到2个估计信号,然后将这2个信号做和或做差,便可得到有效的攻击信号.在含水印的媒体内容中迭加此攻击信号,便可去除合法水印.即使进一步改进水印方案、增加模板数量来增强水印安全性,攻击者依然可以通过多次尝试获得有效的攻击信号,进而删除水印.实验结果表明这种增强的敏感性攻击方法可以在保持高视觉质量的条件下去除水印.更多还原

【Abstract】 It is pointed out that a loophole still exists in a watermarking scheme capable of resisting sensitivity attack.After performing the sensitivity attacks on two different constructed signals close to the decision boundary and gaining two estimated signals,the attacker may add/subtract an estimated signal into/from another one to obtain an effective attack signal.By adding the attack signal into a legal watermarked content with a suitable strength,the embedded watermark can be removed.Even if numerous templates are introduced to improve the security of watermarking system,careful attempts for constructing two similar signals close to the decision boundary can still achieve an effective attack signal that can destroy the embedded watermark.Experimental results show that the embedded watermarks can be deleted when keeping a high perceptual quality of digital image.更多还原