一种实用的新型数据库水印方法

A New Practical Database Watermarking Method

摘要: 可逆的数据库水印方法是保护数据库版权的重要手段。然而，大多数现有的数据库水印方法会在很大程度上修改原始数据，导致数据质量下降；部分方法增加辅助数据导致存储空间增大，或是只能用于数值型数据库，实用性差。在本文中，我们提出了一种健壮且可逆的数据库水印技术，使用遗传模拟退火算法和直方图移位水印，用于数值关系数据，同时采用同义词替换和算术编码对文本数据进行水印添加。利用遗传模拟退火算法为分组数据库选择最佳密钥，将预测误差的直方图进行移位，以更好地嵌入水印。同义词替换和算数编码则可以有效地在文本信息中加入水印。分析显示，我们的方法不但可以增强数据库安全，并可以减少数据质量损失。

Abstract: Reversible database watermarking is an important method to protect database copyright. However, most existing database watermarking methods will modify the original data to a great extent, resulting in the decline of data quality; some methods increase the storage space due to the addition of auxiliary data, and some methods with poor practicability can only be used for numerical database. In this paper, we propose a robust and reversible database watermarking technique, which uses genetic simulated annealing algorithm and histogram shift watermarking for numerical relational data, and uses synonym substitution and arithmetic coding to add watermarking to text data. Genetic simulated annealing algorithm was used to select the best secret key for the grouping da-tabase, and the histogram of the prediction error was shifted to embed the watermark better. Synonym substitution and arithmetic coding can effectively add watermark to text information. Analysis shows that our method can not only enhance database security, but also reduce data quality loss.