Secure Federated Learning Framework for Privacy-Preserving Medical Image Sharing Using Zero-Watermarking & ECC

Abstract

The rapid advancement of digital healthcare has amplified the need for secure medical image sharing. This paper presents a comprehensive framework integrating Digital Zero-Watermarking based on Fractional Racah Orthogonal Moments (FrROMs), Federated Learning (FL), and Elliptic Curve Cryptography (ECC) to deliver privacy-preserving, tamper-resistant medical image management. Patient identity is imperceptibly embedded into images using the FrROMs-based zero-watermarking scheme, which demonstrates robust resistance to Gaussian noise, JPEG compression, salt-and-pepper attacks, and cropping (BER ? 10?³, NC ? 0.99). Disease prediction is performed via decentralized federated learning — only ECC-encrypted model updates are shared — eliminating raw data exposure. Access control, audit logging, and real-time alerts complete the security stack. Experimental results confirm superior performance over existing DWT- and SVD-based zero-watermarking methods

Introduction

The text proposes a secure medical image sharing framework designed to protect patient privacy in digital healthcare systems.

It combines three key technologies:

• Digital zero-watermarking using Fractional Racah Orthogonal Moments (FrROMs) to embed patient identity into medical images without altering them. This ensures strong resistance against attacks like noise, compression, and cropping while maintaining high accuracy.
• Federated Learning (FL) for disease prediction, where models are trained in a decentralized way so that raw medical data never leaves hospitals.
• Elliptic Curve Cryptography (ECC) to encrypt model updates, ensuring secure communication between participants.

The system also includes access control, audit logging, and real-time alerts to enhance security and monitoring.

Experimental results show that the approach outperforms traditional methods like DWT and SVD-based watermarking in terms of robustness and reliability, making it a strong solution for privacy-preserving medical image management.

Conclusion

In this project, a Secure Federated Learning Framework for Privacy-Preserving Medical Image Sharing and Diagnosis has been successfully proposed and analyzed. The system addresses critical challenges in modern healthcare, particularly data privacy, security, and collaboration between multiple medical institutions. By leveraging federated learning, the framework enables hospitals to collaboratively train machine learning models without sharing sensitive patient data, thereby ensuring confidentiality and compliance with privacy regulations. The implementation of local model training at client nodes, combined with secure communication and centralized aggregation using the Federated Averaging algorithm, enhances both data security and model performance. The integration of deep learning techniques for medical image analysis further improves the accuracy and reliability of disease diagnosis. Additionally, the system reduces data silos and enables efficient knowledge sharing across healthcare providers. Experimental results demonstrate that the proposed approach achieves high diagnostic accuracy while maintaining strong privacy protection compared to traditional centralized methods. The use of encryption and secure aggregation ensures that model updates are protected from potential threats, making the system robust and trustworthy. Overall, the proposed framework provides an effective balance between privacy, accuracy, and scalability, making it a promising solution for real-world healthcare applications. With further enhancements such as blockchain integration, advanced AI models, and large-scale deployment, this system has the potential to significantly improve the quality and efficiency of healthcare services.

References

[1] El-Khanchouli, K., et al. (2025). Protecting Medical Images Using a Zero-Watermarking Approach Based on Fractional Racah Moments. IEEE Access, 13, 16978–17001. [2] Rani, A., et al. (2015). A zero-watermarking scheme using discrete wavelet transform. Procedia Computer Science, 70, 603–609. [3] Huang, T., et al. (2022). Robust zero-watermarking algorithm for medical images using double-tree complex wavelet transform and Hessenberg decomposition. Mathematics, 10(7), 1154. [4] Tu, S., et al. (2023). Application of zero-watermarking for medical image in intelligent sensor network security. CMES, 136(1), 293–321. [5] Taj, R., et al. (2024). A reversible-zero watermarking scheme for medical images. Scientific Reports, 14(1), 17320. [6] ] Abualigah, L., et al. (2022). Reptile search algorithm (RSA): A nature-inspired metaheuristic optimizer. Expert Systems with Applications, 191, 116158. [7] Daoui, A., et al. (2022). Stable analysis of large-size signals and images by Racah’s discrete orthogonal moments. Journal of Computational and Applied Mathematics, 403. [8] Thabit, F., et al. (2022). A novel effective lightweight homomorphic cryptographic algorithm for data security in cloud computing. Int. Journal of Intelligent Networks, 3, 16–30. [9] Singh, A.K., & Saxena, D. (2022). A cryptography and machine learning based authentication for secure data-sharing in federated cloud services. J. Applied Security Research, 17(3), 385–412. [10] Xia, Z., et al. (2019). Geometrically invariant color medical image null-watermarking based on precise quaternion polar harmonic Fourier moments. IEEE Access, 7, 122544–122560.

Copyright

Copyright © 2026 Narmatha S., Nemika R., Rithika M., Jeyasri S. R., Mrs. R. Subha. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.