Rachinger B, Meier S, Franke J, Ockel M (2026)
Publication Type: Journal article
Publication year: 2026
Original Authors: Ben Rachinger, Sven Meier, Jörg Franke, Manuela Ockel
Book Volume: 200
Article Number: 108291
DOI: 10.1016/j.infsof.2026.108291
Context:
Federated learning (FL) enables multiple organizations to collaboratively train machine learning models without centralizing raw data. In practice, however, realizing FL systems requires coordination across autonomous stakeholders, heterogeneous technical environments and cross-organizational governance, while existing FL-specific process guidance remains fragmented and often limited to selected life cycle activities.
Objective:
This paper proposes STEER-FL, a Structured End-to-End Reference process for Federated Learning, supporting the systematic realization of FL projects.
Methods:
We conducted a dual-perspective systematic literature review covering both FL-specific process-oriented publications and established machine learning, data science and data mining process models. The identified approaches were evaluated using complementary rubrics for process documentation quality and FL suitability and then synthesized into an integrated process design. The synthesized model was then evaluated by eleven domain experts through a structured rating instrument and semi-structured interviews, and refined accordingly.
Results:
The review identified recurrent deficiencies in existing approaches, including incomplete life cycle coverage, missing progression control through explicit decision points, limited treatment of cross-organizational governance and weak guidance for transferring FL initiatives from simulation to deployment. In response, STEER-FL defines six phases from project definition to deployment & monitoring, 30 activities assigned to four consolidated roles, and decision gates for suitability, feasibility and production readiness. Experts rated the model positively across all acceptance, quality and mechanism-efficacy constructs, confirming its perceived usefulness and informing the refinements. The model is tool-agnostic and applicable to both cross-organizational and enterprise-internal FL settings.
Conclusion:
STEER-FL provides a literature-grounded, process-level model for structuring the socio-technical realization of FL systems, contributing guidance for distributed artificial intelligence engineering and a foundation for more systematic planning, coordination and operationalization of FL projects.
APA:
Rachinger, B., Meier, S., Franke, J., & Ockel, M. (2026). STEER-FL: A process model for federated learning derived from a dual-perspective systematic literature review. Information and Software Technology, 200. https://doi.org/10.1016/j.infsof.2026.108291
MLA:
Rachinger, Ben, et al. "STEER-FL: A process model for federated learning derived from a dual-perspective systematic literature review." Information and Software Technology 200 (2026).
BibTeX: Download