Application of Robust Zero-Watermarking Scheme Based on Federated Learning for Securing the Healthcare Data

Area of Science:

• Medical Imaging
• Cybersecurity
• Machine Learning

Background:

• Internet of Medical Things (IoMT) frameworks face significant privacy and data security challenges.
• Teledermatology relies on smartphone-acquired medical images, which are vulnerable during transmission, risking tampering and data breaches.
• Existing solutions often fail to protect image integrity and patient confidentiality simultaneously.

Purpose of the Study:

• To propose a robust zero-watermarking scheme for teledermatology to address privacy and security vulnerabilities.
• To ensure the integrity of dermatology medical images while preventing unauthorized data disclosure.
• To develop a watermarking solution suitable for the unique requirements of medical image transmission.

Main Methods:

• A federated learning approach was used to train a sparse autoencoder network.
• The trained network extracts image features from dermatology medical images.
• Two-dimensional Discrete Cosine Transform (2D-DCT) was applied to select low-frequency coefficients for zero-watermark creation.

Main Results:

• The proposed scheme demonstrates enhanced robustness against conventional and geometric attacks.
• It achieves superior performance compared to existing zero-watermarking methods.
• Experimental results validate the scheme's effectiveness in protecting medical image data.

Conclusions:

• The developed zero-watermarking scheme effectively safeguards teledermatology data.
• It preserves the crucial information within medical images without modification.
• The federated learning-based approach ensures both privacy protection and data security in teledermatology frameworks.