Robust and Invisible Watermarking Algorithm for Relational Databases Copyright Protection Using Arnold Transform

Abstract

Abstract—Digital watermarking for relational databases appeared as a candidate solution to provide copyright protection, tamper detection, and maintaining integrity of relational data. In this paper, an efficient database watermarking algorithm is proposed that based on embedding the binary image and secret text message in non-numeric mutli-word attributes of all database tuples. The binary image is scrambled first using Arnold transform to guarantee security of watermark even when the database is hacked. After that the tuples are partitioned logically using MD5 hash function using secret parameters. The algorithm takes advantage of the fact, that modern text encoding systems like Unicode offer multiple different characters describing whitespaces. Normal space and thin space characters are selected for representing and embedding watermark bits 0 and 1, respectively. These will not affect the size of database and ensure that the non-whitespace characters aren’t modified. The scheme is non-intrusive, resilient, blind, reversible and suitable for databases of any size with reasonable performance on embedding and extraction. Moreover, it supports incremental watermarking to manage with the dynamic nature database systems are subject to. Experimental results demonstrated the efficiency and robustness of the algorithm against common database attacks and is free distortion and imperceptible, as proved by the analysis.