On the Insecurity of Non-invertible Watermarking Schemes for Dispute Resolving

Robust watermarking is an important and powerful technology with several applications in the field of copyright protection. Watermarking schemes were proposed as primitives for dispute resolving schemes and direct proofs of authorship. However, it was shown that watermark-based dispute-resolving schemes are vulnerable to inversion attacks where an adversary exploits the invertibility property of the underlying watermarking scheme to lead the dispute-resolving mechanism to a deadlock. One of the proposed countermeasures is to make watermarking schemes “non-invertible” by incorporating cryptographic primitives, such as one-way functions, into the watermark embedding process. However, this solution ignores the fact that the security strongly depends on the false-positives probability of the underlying watermarking scheme, i.e., the probability to find a detectable watermark which has never been embedded into the considered content. In this paper, we analyze the impact of the false-positives rate on the security of dispute-resolving applications that use “non-invertible” watermarking schemes. We introduce a general framework for non-invertibility and show that previous constructions based on cryptographic one-way functions, such as [6,11], are insecure, if the false-positive rate is non-negligible.