基于指令统计分布的软件水印方案

提出了一种新的盲提取软件水印方案,水印信号分散在整个代码的指令统计分布上.在程序的指令集中随机抽取两个不相交的基数相等的子集,对特定子集的选取就是要嵌入的水印.改变程序使对应于某个子集的所有指令的出现次数都增加一个常数.检测时,分别计算相应两个子集指令出现次数的平均值,如果两个均值之差大于阈值,则认为包含水印.根据假设检验理论对方案的有效性进行了理论分析和证明.针对Java字节码实现了提出的方案,并通过实验进行了验证.结果表明新方案具有非常强的健壮性,能有效抵抗多种常用攻击.

Instruction statistic distribution based software watermark scheme

A novel blind watermark scheme with the mark spread over the whole instruction distribution was proposed. Two disjoint subsets of equal size were randomly selected from among instruction set, and this selection was the watermark superimposed by changing the elements of one subset through adding a positive integer factor. The means of two subsets were computed for detecting mark. The mark was found if the difference of two means exceeded the threshold. The effectiveness of the scheme was derived from hypotheses test theory. The scheme was implemented treating Java bytecode as targets and was validated by the experiments. The experimental results show that it is very robust and is immune to most known (attacks).