Watermarking is now a well acknowledged means for protection of ownership rights on digital images and videos. The best watermarking algorithms perform well against the now standard benchmark tests included in the Stirmark package. However the Stirmark tests are not sufficient to accurately model the behavior of a skilled attacker with knowledge about the embedding algorithms. In this context, the goal of the Certimark project was twofold: to propose a European watermarking benchmark suite that better matches the most recent attacks as well as the main identified user needs, and to present the state of the art with respect to watermarking methods and watermark attacks. An important issue of benchmarking is a development of realistic operational scenarios and attacks that can be applied in practice against the watermarking technologies. Therefore, the analysis, the classification and the implementation of these attacks are our main contributions to Certimark project.
Our contributions to the Certimark project have in more details been the following:
· Requirements: categorization and description of intentional attacks;
· Tests and trials: description and testing of new attacks, and of the benchmark version BV1 for still images;
· Research and development: besides completing the work on the second generation of estimation-based attacks, we have contributed a novel framework for the capacity-security analysis of data hiding technologies, i.e., an important issue that was previously completely ignored by the watermarking community. We have demonstrated the importance of this analysis for the general concept of copyright protection and secret covert communications. We have also leaded the project deliverable D5.2 'Robustness and countermeasures', regarding the analysis of possible attacks previously described in the deliverable D4.1 and development of appropriate countermeasures against these attacks. A number of developed attacks were integrated into the first academic application-oriented benchmarking tool Checkmark that was publicly distributed from the University web site. Checkmark was presented to the Certimark consortium and positively evaluated as the first working concept of Certimark benchmark. Also in partnership with DCT, we have contributed with the implementation of several attacks in C;
· Dissemination: we have installed and maintained the Certimark Web site
http://www.certimark.org. Besides this 'service activity', we have published a number of articles related to Certimark (available from
http://vision.unige.ch), and co-organized the special session on benchmarking of data hiding technologies during ICME2002, 26-29 August, 2002, Lausanne, Switzerland that is considered by the Certimark consortium as the major CERTIMARK dissemination event at this stage. We are also members of the CAST forum (Competence Center for Applied Security Technologies, Technical Committee on Digital Watermarking,
http://www.cast-forum.de/activities/tcw/overview) and have participated to a workshop on September 25-26, 2002, which has presented major results in the field. Finally, we are co-editors of a forthcoming Signal Processing special issue on Security of Data Hiding Technologies.
In summary, the Certimark project has been very satisfactory. The goals have been reached. We have made many new contacts at the highest research level in the domain and have significantly advanced in the field. We are now recognized as belonging to the best groups in the domain, and are active partners in the preparation of several new European projects (6th Research Framework).
