A new approach to anti-forgery using saliency guided image watermarking

Abstract

Using various image characteristics as a secret key, the cryptanalytic watermarking method is known for enhancing the resilience of authentication systems and protecting against forgery attempts when concealing information within a host image. This study introduces a new approach that leverages saliency features to establish a secret key, subsequently utilizing this key as a parameter for embedding and extracting watermarks. Despite the alteration of image features during watermark embedding, we suggest employing learning techniques in conjunction with saliency models to ensure the robustness of watermark extraction. The proposed image watermarking technique incorporates SVM learning and multiple saliency models. Our findings demonstrate the effectiveness of the cryptanalytic watermarking method in maintaining the watermark's invisibility and stability. The benefits of the saliency feature-based approach for anti-forgery are evident through experiments conducted on a standard dataset.