Abstract
Most of data over the Internet today is hosted on outsourced third-party servers which are not trusted. Sometimes data is to be distributed to other organizations or individuals for pre-agreed use. In both of these scenarios data is susceptible to malicious tampering so there is a need for some mechanism to verify database integrity. Moreover, the authentication process should be able to differentiate between valid updates and malicious modifications. In this paper, we present a novel semi-fragile watermarking scheme for relational database integrity verification. Besides detection and localization of database tampering, the proposed scheme allows modifications to the data that need periodic updates, without requiring re-watermarking. Watermark embedding is distortion free, as it is done by adjusting the text case of selected data values resulting in retention of semantic meaning of data. Additionally, group-based embedding ensures the localization of tampering up to group level. We implemented a proof of concept application of our watermarking technique. Theoretical analysis and experiments show that even a single value modification can be detected with very high probability besides detection of attacks like tuple insertion and tuple deletion.