4th Workshop on Advancements in Federated Learning

AI-based systems, particularly those grounded in machine learning, have become central to modern society. At the same time, growing awareness of privacy risks has made users increasingly reluctant to share sensitive data, while regulatory frameworks such as the GDPR and the European AI Act (2024) impose stricter requirements on how data is collected, processed, and protected.

Federated Learning (FL) has emerged as a key paradigm for privacy-preserving machine learning, enabling collaborative model training across distributed data sources without sharing raw data. While it offers strong potential to reconcile data utility with privacy, FL also introduces new challenges, including vulnerability to adversarial behavior, bias across heterogeneous participants, and limited interpretability and accountability.

These issues raise important concerns about the reliability and societal impact of FL, particularly in sensitive domains such as healthcare and finance. In this context, Trustworthy Federated Learning (TFL) extends the principles of Trustworthy AI to decentralized settings, addressing challenges such as fairness under non-IID data, robustness, privacy guarantees, and transparency.

Given the considerable potential of FL in enabling privacy-preserving data-driven innovation, the 4th WAFL workshop aims to provide a focused forum for researchers and practitioners to discuss recent advances and open challenges in Federated Learning, with a particular emphasis on Trustworthy FL. Leveraging the broad expertise of the ECML-PKDD community, the workshop welcomes contributions spanning algorithmic, theoretical, and system-level perspectives, including fairness-aware FL, privacy- and security-enhancing techniques, interpretability, regulatory compliance in decentralized learning, and real-world deployment experiences.

TOPICS AND THEMES

The WAFL workshop will be centered on the theme of improving and studying the Federated Learning setting. It will welcome applicative and theoretical contributions as well as contributions about specific settings and benchmarking tools.

Topics of interest include, but are not limited to:

• Algorithmic and theoretical advances in FL

• Federated Learning with non-iid data distributions

• Security and privacy of FL systems (e.g., differential privacy, adversarial attacks, poisoning attacks, inference attacks, data anonymization, model distillation, secure multi-party computation ...)

• Fairness, Interpretability and explainability of FL models

• Decentralized and peer-to-peer FL

• Transparency and accountability in decentralized learning

• Compliance with regulatory frameworks (e.g., the European AI Act)

• Real-world applications of FL (e.g., FL for healthcare, FL on edge devices, advertising, social network, blockchain, web search ...)

• Tools and resources (e.g., benchmark datasets, software libraries, ...)

SUBMISSION GUIDELINES

We invite submissions of original research on all aspects of Federated Learning (see the not complete list of topics above). Each accepted (short and long) paper will be included in the workshop proceedings (published by Springer Communications in Computer and Information Science). All papers will be presented in the talk session. Authors of short and long papers will have the faculty to opt-in or opt-out from the publication in the proceedings.

We accept the following types of submissions:

• Short Papers (6 pages + references+supplementary): Work-in-progress, position papers, or open problems. Accepted short papers will be included in the Springer Workshop Proceedings of ECML-PKDD 2026. Short papers must follow the ECML-PKDD 2026 formatting guidelines (see here).

• Long Papers (12 pages + references+supplementary): Novel, original research not published elsewhere. Accepted long papers will be included in the Springer Workshop Proceedings of ECML-PKDD 2026. Long papers must follow the ECML-PKDD 2026 formatting guidelines (see here).

• Non-archival Submissions: Papers recently accepted or under review at other venues. These submissions have no formatting restrictions but must be accompanied by a cover letter explaining their relevance to the workshop. Non-archival submissions will not be included in the Springer Workshop Proceedings and will undergo a different selection process by the workshop organizers.

Short and long papers need to be "best-effort" anonymized. We strongly encourage making code and data available anonymously (e.g., in an anonymous GitHub repository via Anonymous GitHub or in a Dropbox folder). The authors may have a (non-anonymous) pre-print published online, but it should not be cited in the submitted paper to preserve anonymity. Reviewers will be asked not to search for them.

Submissions will be evaluated by at least two reviewers on the basis of relevance, technical quality, potential impact, and clarity. The reviewing process is double-blind (reviewers and area chairs are not aware of the identities of the authors; reviewers can see each other’s names). Papers must not include identifying information of the authors (names, affiliations, etc.), self-references, or links (e.g., GitHub, YouTube) that reveal the authors’ identities (e.g., references to own work should be given neutrally like other references, not mentioning ‘our previous work’ or similar). However, we recognize there are limits to what is feasible with respect to anonymization. For example, if you use data from your own organization and it is relevant to the paper to name this organization, you may do so.

Papers must be submitted in PDF format via Microsoft CMT at this link. Choose "WAFL" from the drop-down when creating the submission.

Important dates are reported here.