A Novel Relational Database Watermarking Algorithm Based on Clustering and Polar Angle Expansion

Digital watermarking has been widely applied to relational database for ownership protection and information hiding. But robustness and reversibility are two key challenges due to the frequently database maintaining operators on those tuples. This paper proposes a novel relational database watermarking scheme based on a fast and stable clustering method on database tuples, which adopts Mahalanobis distance as the similarity measurement. Before the process of watermark embedding and detecting, the databases tuples are adaptively clustered into groups according to the length of binary watermark. Moreover the watermark segments are respectively embedded into or detected from those groups according to the numeric field's Lowest Significant Bit (LSB) and polar angle expansion. The majority decision strategy is used to determine the value of watermark bit in blind detection process. The experiment results indicate that the proposed watermarking scheme has higher robustness and reversibility under blind detection against the database maintaining operators.