A Blind Robust Reversible Watermark Scheme for Textual Relational Databases with Virtual Primary Key

In recent years, many researchers have studied reversible watermarking techniques for relational databases. Most of the developed schemes have been based on a primary key attribute in order to determine the selected tuples and attributes to carry the watermark bits. What happens, however, when the primary key attribute does not exist for a relational database? In this paper, we propose a blind robust reversible watermarking scheme for a textual relational database. This scheme does not rely on the primary key attribute. To avoid the absolute dependence on the primary key attribute, as in existing schemes, in the proposed scheme the content of textual attributes are used to generate the virtual primary attribute that is applied in tuple and attribute selections. Moreover, the selection of attributes does not depend on the order of attributes in the relational database. Model and robustness analysis demonstrate that our proposed scheme achieves a high resilience against different types of tuple attacks, i.e., tuple attacks and attribute attacks. The experimental results also confirm that the proposed scheme is more secure and robust than other existing schemes.