EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

An Exploratory Survey Study Toward a Conceptual Framework for Structuring the Integration of Privacy-Enhancing Technologies in Federated Learning and Analytics

Hammam Abu Attieh, Mehmed Halilovic, Raouf Kerkouche, Shadi Rahimian, Hui-Po Wang, Raphael Rehms, Georgios Kaissis, Alexander Ziller, Jonas Kuntzer, Karen Otte, Thierry Meurers, Mario Fritz und 2 weitere

IEEE Access · 2026 · 14 · 41508-41528

DOI: 10.1109/access.2026.3672991 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Hammam Abu Attieh und 13 weiteren Autor:innen. Die Arbeit nennt 222 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: German Federal Ministry of Research, Technology, and Space (BMFTR) through the PrivateAIM Project.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The increasing relevance of distributed data in medical research has accelerated the need for privacy-preserving technologies that enable secure analysis without centralizing sensitive information. Federated learning and analytics have emerged as promising approaches that allow multi-institutional collaboration while maintaining data locality. This paper aims to 1) consolidate key biomedical data types, representative analytical methods, and typical outputs relevant to federated settings across modalities, 2) synthesize which privacy-enhancing technologies (PETs) have been proposed to protect privacy in these contexts by mapping them to distinct FL/FA workflow stages and highlighting stage-specific assumptions and trade-offs, and 3) develop a modular conceptual framework to structure integration decisions in complex research networks. Drawing on interdisciplinary expertise, we map representative PETs, such as pseudonymization, anonymization, differential privacy, secure computation, and statistical disclosure controls to the Five Safes Framework, encompassing safeguards across trusted users, governance structures, data handling, analytical processes, and outputs. Our synthesis highlights how privacy risks can be mitigated through a layered combination of technical, legal, and organizational measures. The resulting framework supports the modular composition of PETs tailored to specific data protection needs and illustrates how trade-offs between privacy and analytical utility can be balanced. Overall, this work provides a conceptual foundation for integrating PETs in federated biomedical research, aligning technical implementation with compliance and ethical accountability.

Förderung

  • German Federal Ministry of Research, Technology, and Space (BMFTR) through the PrivateAIM Project

Zitieren

APA

Attieh, H. A., Halilovic, M., Kerkouche, R., Rahimian, S., Wang, H., Rehms, R., Kaissis, G., Ziller, A., Kuntzer, J., Otte, K., Meurers, T., Fritz, M., Rüeckert, D., & Prasser, F. (2026). An Exploratory Survey Study Toward a Conceptual Framework for Structuring the Integration of Privacy-Enhancing Technologies in Federated Learning and Analytics. IEEE Access, 14, 41508-41528. https://doi.org/10.1109/access.2026.3672991

Vancouver

Attieh HA, Halilovic M, Kerkouche R, Rahimian S, Wang H, Rehms R, et al. An Exploratory Survey Study Toward a Conceptual Framework for Structuring the Integration of Privacy-Enhancing Technologies in Federated Learning and Analytics. IEEE Access. 2026;14:41508-41528. doi:10.1109/access.2026.3672991

BibTeX
@article{Attieh2026,
  author = {Attieh, Hammam Abu and Halilovic, Mehmed and Kerkouche, Raouf and Rahimian, Shadi and Wang, Hui-Po and Rehms, Raphael and Kaissis, Georgios and Ziller, Alexander and Kuntzer, Jonas and Otte, Karen and Meurers, Thierry and Fritz, Mario and Rüeckert, Daniel and Prasser, Fabian},
  title = {An Exploratory Survey Study Toward a Conceptual Framework for Structuring the Integration of Privacy-Enhancing Technologies in Federated Learning and Analytics},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {41508--41528},
  doi = {10.1109/access.2026.3672991}
}
Kostenlos starten →