Abstract: A fragile watermarking algorithm is proposed to detect malicious modifications of relational databases.In the proposed algorithm，all tuples in relational databases are divided into different groups，and by sorting secretly the tuples in each group，group watermarking matrix made up of attribute watermark and tuple watermark is generated for localizing modifications of relational databases in a group.The watermark is generated dynamically by using one-way hash function and relational data for watermark security and blind extraction.Theoretical analysis and experimental results show that the proposed method can detect tuple insertion，attribute value modification，tuple deletion and attribute variation made to relational database for providing authenticity verification of relational data.

Key words: fragile watermarking, relational database, watermarking matrix, group, modification

摘要： 为了检测对关系数据库的恶意篡改，提出了一种脆弱性数字水印算法。该算法将数据库的元组划分到不同的分组中，在对每个分组内的元组进行秘密排序的基础上，生成由属性水印和元组水印构成的分组水印矩阵，因此可以将对数据库的篡改定位在分组范围内。利用单向哈希函数及关系数据动态生成水印，不但保证了水印信息的安全性，而且也实现了水印的盲检测。理论分析和实验结果表明，该方法能够有效探测攻击者对关系数据库进行元组添加、属性值修改、元组删除和属性变化四类操作，从而为关系数据的真实性认证提供依据。

关键词: 脆弱性水印, 关系数据库, 水印矩阵, 分组, 篡改