On the possibility of non-invertible watermarking schemes

Recently, there are active discussions on the possibility of non-invertible watermarking scheme. A non-invertible scheme prevents an attacker from deriving a valid watermark from a cover work. Recent results suggest that it is difficult to design a provably secure noninvertible scheme. In contrast, in this paper, we show that it is possible. We give a scheme based on a cryptographically secure pseudo-random number generator (CSPRNG) and show that it is secure with respect to well-accepted notion of security. We employ the spread spectrum method as the underlying watermarking scheme to embed the watermark. The parameters chosen for the underlying scheme give reasonable robustness, false alarm and distortion. We prove the security by showing that, if there is a successful attacker, then there exists a probabilistic polynomial-time algorithm that can distinguish the uniform distribution from sequences generated by the CSPRNG, and thus contradicts the assumption that the CSPRNG is secure. Furthermore, in our scheme the watermark is statistically independent from the original work, which shows that it is not necessary to enforce a relationship between them to achieve noninvertibility.