Lightweight dual-watermarking framework for medical image authentication and integrity preservation

Area of Science:

• Medical Imaging
• Digital Watermarking
• Image Security

Background:

• Medical image authentication is crucial for secure healthcare and reliable clinical decisions.
• Ensuring the integrity and authenticity of diagnostic images is paramount.
• Existing methods may lack robustness or efficiency for real-world medical applications.

Purpose of the Study:

• To propose a robust dual watermarking framework for medical image authentication.
• To develop a lightweight Convolutional Neural Network (CNN) decoder for efficient watermark extraction.
• To enhance the security and integrity of medical images against various attacks.

Main Methods:

• A hybrid frequency-domain watermarking technique combining Discrete Wavelet Transform (DWT) and Discrete Cosine Transform (DCT).
• Embedding a machine-readable QR code and a hospital logo as dual watermarks.
• Utilizing a lightweight CNN decoder optimized with an enhanced loss function (MSE, SSIM, Sobel edge loss).

Main Results:

• High watermark extraction fidelity with Peak Signal-to-Noise Ratio (PSNR) from 64.87 to 68.75 dB and Normalized Correlation (NC) reaching 1.0.
• Demonstrated robustness against Salt & Pepper noise, median filtering, rotation, and cropping.
• Achieved significant improvements in PSNR (28-35%) and NC (12-15%) compared to baseline methods.
• The lightweight CNN decoder (0.65 MB) offers real-time inference for resource-constrained devices.

Conclusions:

• The proposed dual watermarking method effectively maintains medical image visual quality and structural integrity.
• The framework provides robust security and efficient watermark retrieval, suitable for medical devices.
• This approach enhances the overall security and trustworthiness of digital medical images.