In the modern digital healthcare ecosystem, the electronic transmission of medical imagery requires stringent data-provenance frameworks that safeguard sensitive patient information without compromising diagnostic fidelity. Traditional image watermarking methods often fail to balance visual imperceptibility with robustness against malicious cyber threats. To overcome these operational bottlenecks, this paper introduces a highly secure, computationally optimized hybrid deep learning-based image watermarking framework that integrates high-dimensional hyperchaotic encryption. The proposed architecture establishes a zero-leakage protection layer by introducing a discrete 6-Dimensional hyperchaotic cosine-sine map (6D-HCSM) driven by a host-derived SHA-256 seed vector. Moving beyond conventional isolated encryption, this model introduces a dynamic nonlinear feedback loop in which the map's real-time trajectories directly pace the search core of the Adaptive Osprey Optimization Algorithm (AOOA). Following optimal coordinate selection, a Depthwise Separable Convolutional Assisted Generative Adversarial Network (DWC-GAN) manages spatial feature embedding, significantly reducing computational complexity. Experimental simulations on standard clinical datasets demonstrate outstanding efficiency, achieving a peak signal-to-noise ratio (PSNR) of 68.9754 dB, a structural similarity index measure (SSIM) of 0.9925, and a watermark retrieval accuracy of 99.45% under aggressive attacks. |
