Robust digital watermarking systems are important building blocks in applications such as fingerprinting, dispute resolving or direct proofs of authorship, where the presence of a watermark serves as evidence for some fact, e.g., illegal redistribution or authorship. A major drawback of (symmetric) watermarking schemes in this context is that proving the presence of a watermark requires disclosing security critical detection information (watermark, detection key, original data) to a (potentially malicious) verifying party. This may completely jeopardise the security of embedded watermarks once this information is revealed. To overcome this problem recent work on secure watermark detection proposes cryptographic proofs that perform the detection on concealed detection information. The proposed solutions focus on correlation-based detection and can be applied to any watermarking scheme whose detection criteria can be expressed as a polynomial relation between the quantities required for the detection.
In this paper, we present in-depth guidelines for the adoptions required to transform well-established watermarking schemes by Cox et al and Piva et al into secure cryptographic proofs in the non-interactive setting. Moreover, we present our implementation, its performance results and the corresponding tool we have developed for this purpose. Our results underpin the practicability of the cryptographic approach.