A Custom GAN-Based Robust Algorithm for Medical Image Watermarking

In recent years, digital healthcare and deep learning have seen substantial advancements, with medical imaging emerging as a crucial component. Despite this progress, the security of these images remains a significant challenge, particularly during network transmission where they are at risk of unauthorized access and tampering. To mitigate these risks, we introduce a novel and robust watermarking algorithm, rooted in advanced deep learning technologies and tailored specifically for medical images. Employing a custom Generative Adversarial Network (GAN), our innovative approach embeds QR codes—encoding confidential, patient-specific data—directly into medical images. Rigorous experimental evaluations confirm the resilience of our solution against a wide array of adversarial attacks and various image distortions, achieving an exceptional average Peak Signal-to-Noise Ratio (PSNR) of 37.12 and an extraction accuracy above 99%. Our algorithm not only enhances the security and integrity of medical images but also fortifies the protection of patient privacy. Importantly, this work fills a research gap by applying Generative Adversarial Networks (GANs) to the domain of end-to-end medical image watermarking.