Robust adaptive watermarking with deep learning and multi-chaotic encryption for secure NIfTI neuroimaging

This paper presents a Robust Adaptive Watermarking framework using Deep Learning and Multi-Chaotic Encryption (RAWDL-MCE) for securing NIfTI neuroimaging data. The pro-posed system integrates CNN-based adaptive embedding in the DWT domain with multi-chaotic key encryption to ensure high imperceptibility, robustness, and data security. The adaptive embedding mechanism dynamically adjusts watermark strength based on local image features, preserving diagnostic quality while enhancing resilience against various attacks, including noise, compression, and geometric distortions. Experimental results demonstrate that the proposed framework achieves PSNR values exceeding 49 dB, SSIM greater than 0.99, and NC above 0.95, ensuring minimal image distortion and reliable watermark recovery. Moreover, the system attains high key sensitivity and low computational time, making it suitable for real-time applications. Overall, RAWDL-MCE offers a secure, efficient, and clinically viable solution for protecting sensitive medical imaging data in healthcare and telemedicine environments.