Zero-Watermarking Algorithm for Medical Image Based on VGG19 Deep Convolution Neural Network.

Area of Science:

• Medical Imaging
• Computer Science
• Information Security

Background:

• Medical information systems face security challenges in storing and transmitting sensitive patient data.
• Protecting lesion areas in medical images requires specialized security measures.
• Existing watermarking techniques may not adequately address the unique requirements of medical image security.

Purpose of the Study:

• To propose a robust zero-watermarking algorithm for enhancing medical image security.
• To ensure the integrity and confidentiality of medical images during storage and transmission.
• To develop a method resistant to local nonlinear geometric attacks.

Main Methods:

• Utilizing a pretrained VGG19 network to extract deep feature maps from medical images.
• Applying Fourier transform and selecting low-frequency coefficients to construct a feature matrix.
• Generating 64-bit binary perceptual hashing values using a mean-perceptual hashing algorithm.
• Employing a scrambled watermarking image and Hermite chaotic neural network for robust zero-watermarking and secondary protection.

Main Results:

• The proposed algorithm effectively extracts deep features for robust watermarking.
• The method demonstrates strong resistance to local nonlinear geometric attacks.
• Achieved robustness, security, and invisibility in medical image watermarking.
• The algorithm is simple to implement compared to existing methods.

Conclusions:

• The VGG19-based zero-watermarking algorithm provides a secure solution for medical images.
• The algorithm offers effective protection against various attacks, ensuring data integrity.
• This approach enhances the security of medical information systems through robust watermarking.