Abstract
Recent research brought up numerous techniques for copyright protection and tamper proofing of relational databases along with proof of robustness, etc. However, these techniques are short of presenting a generalized method for susceptible key-based attacks. In this research, we proposed a framework for the analysis of watermarking system against susceptibility to key attacks. We identified two primary concepts of attack models, SKMDs (Single Key Multiple Datasets) and MKsSD (Multiple Keys Single Dataset). These attack models make variants of single and multiple datasets by the usage of single and multiple keys for watermark insertion. The relationship between various pairs of original and watermarked datasets is then statistically analyzed to determine the linearity among datasets. The strength of the attack models is measured by multivariate and discriminant analysis methods like Wilks’ lambda, Pillai’s trace test, and Box’s M test. The empirical analysis shows that MKsSD model in a watermarking system has high significance as compared to SKMDs. We conclude that SKMDs model is more vulnerable to key-based attacks than MKsSD model even by varying watermarking system parameters
