Leakage Identification for Secret Relational Data Using Shadowed Watermarks

Various database watermarking techniques have been developed to protect the ownership of relational data, some provide extra utilities for pirate identification. In all the proposed schemes, the data owner controls the watermark key and the watermark generation process, which makes it possible for the owner to falsely accuse an innocent user of pirating. In this paper, we propose a novel watermark scheme to solve the problem of leakage identification for secret relational data, which is a small-scaled pirate identification problem. We devise a scheme that combines two keys to generate a shadowed watermark, and we design a protocol to bind the data distribution to the identities of both the owner and the user. A user is able to prove his innocence if falsely accused, so that both parties are fairly protected. We provide probabilistic analyses for the effectiveness and security of our scheme, and we test its performance to conclude that the scheme is qualified for practical use.