Abstract: Digital fingerprinting is a technology which can protect the multimedia content from illegal use by unauthorized users.Collusion attack is an effective attack with digital fingerprinting where several copies with the same content but different fingerprints are combined to remove the original fingerprints.This paper proposes a new nonlinear collusion attack method based on the independent Gaussian fingerprint.The maximum value，the minimum value and the average value are chosen from every position of the fingerprints by colluders.A new illegal copy is built by the sum of maximum value and minimum value subtracted the average value.The result shows that the attack performance is more effective than other methods under the same perceptual condition.

Key words: digital fingerprinting, nonlinear collusion attack, spread-spectrum embedding

摘要： 数字指纹是一种保护多媒体数字产品防止未授权用户使用的技术。目前，针对数值指纹的一种强大的攻击方法是合谋攻击。合谋者将各自的拷贝联合起来，通过移除或消减拷贝中原始指纹构建一个新的非法拷贝。以独立同分布的高斯指纹为研究对象，提出了一种新的非线性合谋攻击方法。该方法通过选取拷贝序列中各位置的最大值与最小值以及平均值，将最大最小值相加后减去平均值的方法产生出一个新的非法拷贝。结果显示该攻击方法在同等的感知条件下，能够对高斯指纹获得更有效的攻击。

关键词: 数字指纹, 非线性合谋攻击, 扩频嵌入