Efficient watermark detection by using the longest common substring technique

Large scale of watermarking methods is available in the literature. These methods differ in visibility, capacity, and robustness. In watermarking, the robustness against attacks is the most challenging issue. The desynchronization attacks are the most serious problems facing the watermarking process. The traditional correlation methods fail in watermark detection. Until today there is no widely used algorithm for solving the desynchronization attacks. In this paper, we will introduce a new algorithm for solving the watermark desynchronization attacks. The watermark embedding and detection models are introduced. So, these models are related to the attacker model by presenting four attacking scenarios. We show the effect of each attack scenario on bit rate, signal distortion, and robustness. We conclude that, the attacker could not distort a big part of the watermark. So, we suggest using a probabilistic embedding model combined with the longest common substring technique. This combination is efficient in solving the desynchronization attacks. Results show that, the proposed algorithm is powerful against the attacking scenarios. Moreover, the watermark is still be detected even if only 5% of the watermark is recovered.