A DCT-Based Semi-Fragile Image Watermarking Scheme with Tamper Detection Algorithm

Abstract

This research proposes two level tamper detection in a semi-fragile image watermarking. The primary challenge in semi-fragile image watermarking is to achieve the balance between the visual imperceptibility of the watermarked image and the effectiveness of tamper detection. This study investigates the best embedding location for embedding authentication data against malicious, incidental, and hybrid attacks. The best embedding location in the semi-fragile can be achieved by evaluating coefficient pairs in the DCT coefficients. In the first step, the proposed tamper detection will check each surrounding block of non-tampered blocks. In the second step detection, the proposed method will re-evaluate the corresponding tampered block, it is being tampered with or not. The experimental results present the proposed embedding scheme produced an imperceptibility PSNR about 40.9807 dB and SSIM about 0.9790 for various RGB images from USC-SIPI, UCID-1338, Kodak-PCD0992, and Stanford Dogs image datasets. The proposed scheme effectively localizes malicious attacks across different tampering rates and maintains an average accuracy of 0.9964 under malicious attack, and an average value of 0.6155 under hybrid attack.