Implementing federated learning with privacy-preserving encryption to secure patient-derived imaging and sequencing data from cyber intrusions
Fuente:
Zenodo
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Recurso digital |
| Language: | English |
| Published: |
Zenodo
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866901120623837184 |
|---|---|
| author | Kalejaiye, Adebayo Nurudeen Shallom, Kigbu Chukwuani, Elvis Nnaemeka |
| author_facet | Kalejaiye, Adebayo Nurudeen Shallom, Kigbu Chukwuani, Elvis Nnaemeka |
| contents | <p>The growing adoption of artificial intelligence (AI) and data-driven analytics in healthcare has accelerated the integration of large-scale patient-derived imaging and genomic sequencing data into clinical workflows. However, this surge in biomedical data sharing has intensified cybersecurity challenges, particularly in protecting sensitive patient information from unauthorized access and cyber intrusions. Traditional centralized machine learning models, which aggregate data into a single repository, pose significant privacy risks and increase the attack surface for malicious actors. To address these challenges, federated learning (FL) has emerged as a transformative paradigm, enabling collaborative model training across decentralized nodes without transferring raw data. Yet, while FL mitigates some privacy concerns, it remains vulnerable to inference attacks, gradient leakage, and model inversion tactics. This paper explores the implementation of federated learning frameworks integrated with privacy-preserving encryption techniques, such as homomorphic encryption, differential privacy, and secure multiparty computation, specifically for safeguarding patient-derived medical imaging and sequencing datasets. These technologies ensure that sensitive genetic markers, radiographic scans, and multi-omic features remain encrypted throughout model training and aggregation processes. We examine recent advances in privacy-enhancing technologies, discuss system architectures suited for cross-institutional healthcare collaboration, and evaluate their performance trade-offs in terms of computational cost, model accuracy, and security guarantees. Furthermore, we propose a hybrid encryption-aware federated learning workflow tailored to radiogenomic applications, highlighting its resilience against adversarial threats while maintaining diagnostic precision. By narrowing focus to clinical implementations, this work provides a scalable and secure foundation for AI-driven biomedical research, enhancing trust and compliance in digital health ecosystems.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_17233165 |
| institution | Zenodo |
| language | eng |
| publishDate | 2025 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Implementing federated learning with privacy-preserving encryption to secure patient-derived imaging and sequencing data from cyber intrusions Kalejaiye, Adebayo Nurudeen Shallom, Kigbu Chukwuani, Elvis Nnaemeka Federated Learning Privacy-Preserving Encryption Medical Imaging Genomic Data Security Homomorphic Encryption Radiogenomics <p>The growing adoption of artificial intelligence (AI) and data-driven analytics in healthcare has accelerated the integration of large-scale patient-derived imaging and genomic sequencing data into clinical workflows. However, this surge in biomedical data sharing has intensified cybersecurity challenges, particularly in protecting sensitive patient information from unauthorized access and cyber intrusions. Traditional centralized machine learning models, which aggregate data into a single repository, pose significant privacy risks and increase the attack surface for malicious actors. To address these challenges, federated learning (FL) has emerged as a transformative paradigm, enabling collaborative model training across decentralized nodes without transferring raw data. Yet, while FL mitigates some privacy concerns, it remains vulnerable to inference attacks, gradient leakage, and model inversion tactics. This paper explores the implementation of federated learning frameworks integrated with privacy-preserving encryption techniques, such as homomorphic encryption, differential privacy, and secure multiparty computation, specifically for safeguarding patient-derived medical imaging and sequencing datasets. These technologies ensure that sensitive genetic markers, radiographic scans, and multi-omic features remain encrypted throughout model training and aggregation processes. We examine recent advances in privacy-enhancing technologies, discuss system architectures suited for cross-institutional healthcare collaboration, and evaluate their performance trade-offs in terms of computational cost, model accuracy, and security guarantees. Furthermore, we propose a hybrid encryption-aware federated learning workflow tailored to radiogenomic applications, highlighting its resilience against adversarial threats while maintaining diagnostic precision. By narrowing focus to clinical implementations, this work provides a scalable and secure foundation for AI-driven biomedical research, enhancing trust and compliance in digital health ecosystems.</p> |
| title | Implementing federated learning with privacy-preserving encryption to secure patient-derived imaging and sequencing data from cyber intrusions |
| topic | Federated Learning Privacy-Preserving Encryption Medical Imaging Genomic Data Security Homomorphic Encryption Radiogenomics |
| url | https://doi.org/10.5281/zenodo.17233165 |