GAN-based watermarking for encrypted images in healthcare scenarios

Nowadays, it is very common for healthcare professionals or staff to transmit digital data in the form of images over public channels or store it on hard drives or third-party clouds. However, unauthorised users and cloud-service providers may view or abuse these sensitive images. This research proposes a generative adversarial network (GAN)-based watermarking for encrypted images to prevent data leakage in healthcare scenarios. The technique uses a combination of a chaotic map and randomised singular value decomposition (RSVD) to encrypt the image first. Subsequently, a GAN model is developed for watermark generation by hiding multiple marks within an image. Later, the encrypted image is marked by embedding the generated watermark for copyright protection and authentication. This fundamentally solves the problem of copyright violation and privacy leakage of medical data. Experimental results have demonstrated that the proposed method is imperceptible and successfully resists various attacks. The obtained results confirmed the superiority of this method over other techniques, which makes it more suitable for healthcare applications.