针对现有水印算法的不足，提出了一种新的脆弱性关系数据库水印算法。在本算法中，由水印密钥控制元组的划分和水印的生成，通过构造水印矩阵对关系数据遭受的篡改进行定位和描述。实验结果和理论分析表明，水印生成方法安全有效，可实现水印的盲检测。水印嵌入引起的数据误差较小，算法具有较好的可行性。关键词: 数据库水印;内容认证;分组;误差Abstract: A new fragile watermarking algorithm based on relational database is proposed in view of existing algorithm insufficiency. The tuples are divided into groups and the watermarks are constructed according to a embedding key in the algorithm. Watermarking matrix is constructed to detect and describe malicious modifications of relational data. Theoretical analysis show that the method of generating watermark is secure and effective, and may be used for blind extraction. Experimental results and theoretical analysis has proved the feasibility because of less permissible error of data caused by the embedding of watermarking.Key words: database watermarking, content authenticity, group, permissible error