A Weight-based Semi-Fragile Watermarking Scheme for Integrity Verification of Relational Data

In recent years, digital watermarking has been proposed as an efficient method for protecting database integrity. However, prior authentication watermarking schemes are all totally fragile, i.e., they do not allow any legitimate data modification. Moreover, with most of these schemes, innocent distortions such as tuple and/or attribute sorting may be considered as tampering. In this paper, we propose a novel semi-fragile watermarking technique for tamper detection of relational databases. First, the proposed scheme associates a weight to each database attribute that reflects its sensitivity to benign updates. Then, the tuples are securely distributed into groups in which a watermark is embedded independently. As a result, our semi-fragile scheme can detect and localize malicious modifications at group level, while allowing authorized updates. Theoretical analysis and experimental results demonstrate that common database attacks can be detected with high rate.