An Improved Secure Semi-fragile Watermarking Based on LBP and Arnold Transform

Abstract

In this paper, we analyze a recently proposed semi-fragile watermarking scheme based on local binary pattern (LBP) operators, and note that it has a fundamental flaw in the design. In this work, a binary watermark is embedded into image blocks by modifying the neighborhood pixels according to the LBP pattern. However, different image blocks might have the same LBP pattern, which can lead to false detection in watermark extraction process. In other words, one can modify the host image intentionally without affecting its watermark message. In addition, there is no encryption process before watermark embedding, which brings another potential security problem. To illustrate its weakness, two special copy-paste attacks are proposed in this paper, and several experiments are conducted to prove the effectiveness of these attacks. To solve these problems, an improved semi-fragile watermarking based on LBP operators is presented. In watermark embedding process, the central pixel value of each block is taken into account and Arnold transform is adopted to guarantee the security of watermark. Experimental results show that the improved watermarking scheme can overcome the above defects and locate the tampered region effectively.