An algorithm resistive to invertibility attack in watermarking relational databases

As a copyright protection technique, digital watermarking has arisen more and more attention from the public. It has been widely applied to copyright protection of multimedia data (images, video and audio). The increasing use of relational databases also calls for the need of applying the technique to databases in order to protect the copyright. With the development of digital watermarking technique for relational databases, some studies go deep into watermarking attacks. Study on attacks will obtain algorithm's improvement and technique's development. Although there are several relational database digital watermarking algorithms, they cannot resist some attacks like invertibility attack. Invertibility attack is such an attack that the attacker finds a fictitious watermark in the watermarked database, and he fabricates an original database to announce his copyright possession. Thus it will result in the confusion of copyright confirmation. After analyzing the invertibility attack and studying the current watermarking algorithms on relational databases, we propose an algorithm by constructing relating characteristics according to the content of the database, which can resist invertibility attack efficiently. The simulative experiments also demonstrate that the algorithm is resistive to subset attacks as well.