Relational data watermarking resilience to brute force attacks in untrusted environments

Watermarking techniques for relational data that aim at robustness face the challenge of guaranteeing watermark persistence despite aggressive attacks. Nevertheless, attackers cannot compromise watermark detection to their will since they must ensure the usability of the stolen data. Attacks that constitute a potential master-key for false ownership claims focus on leaking or compromising values used for watermark synchronization. For instance, brute-force attacks attempt to guess the values of parameters used for watermark embedding. The chances for brute-force attacks of successfully performing illegal watermark detection increases in untrusted environments where the leaks of parameters are common. In this paper, (i) we show how brute-force attacks succeed even if not all watermark detection parameters’ values are correctly guessed, (ii) we define the role played by each parameter in robustness, and (iii) we formalize a method for improving the resilience of existing watermarking techniques in scenarios lacking trust. According to the experimental results, watermarking approaches to which our proposal has been applied benefit in terms of robustness and security requirements, extending the deployment opportunities for other techniques.