Thesis title: Robust and Trustworthy Federated Learning in Distributed System
Date: Thursday, October 8, 2026, at 2:00 PM,
Place: Amphithéâtre Blaise at CERI.
The presentation will be held in English.
The jury members are:
Reviewers
• Mr Stefano Secci — Professor, Cnam, France
• Mr Jalel Ben Othman — Professor, University of Paris 13, France
Examiners
• Mr Mohamed El Kamili — Professor, Hassan II University of Casablanca, Morocco
• Ms Tamara Tosic — Research Engineer, Orange, France
• Ms Hajar Elhammouti — Senior Lecturer, Mohammed VI Polytechnic University, Morocco
Supervisors
• Mr Rachid Elazouzi — Professor, Avignon University, France
• Mr Abdellatif Kobbane — Professor, ENSIAS, Mohammed V University in Rabat, Morocco
Abstract:
Today, data is rarely available in a single place. It is produced and stored across personal devices, institutions, infrastructures, and services, often under constraints related to privacy, ownership, regulation, or technical accessibility. This situation has changed the way machine learning systems can be designed. In many applications, learning must take place without collecting all data in a central location, while still allowing several actors to contribute to a shared model. Federated Learning addresses this need by enabling collaborative training from local data. Yet, by avoiding direct access to the data, it also makes the learning process more uncertain. The server receives model updates from clients whose data distributions, training conditions, and reliability are not fully observable.
This thesis studies this challenge from the perspective of robustness and trust in Federated Learning. In practical deployments, clients may differ significantly in the nature and quality of their data, may participate irregularly, or may produce updates that do not reflect the intended learning objective. Such updates can result from natural heterogeneity, faults, corrupted environments, or adversarial behaviour. The central problem addressed in this work is therefore how to build a reliable global model when the contributions used for training cannot all be assumed to be equally trustworthy. More specifically, the thesis investigates how the server can infer the reliability of client updates and use this information during aggregation, without discarding useful knowledge arising from heterogeneous data.
The objective of this work is to develop methods that make this decision process more robust under uncertainty. The thesis explores how client reliability can be inferred from the behaviour of model updates, both individually and collectively, and how this information can guide the construction of the global model. Instead of treating robustness as a fixed rule applied after local training, this work considers it as an adaptive process embedded in the federated optimization loop. It studies how unreliable contributions can be limited while preserving the diversity needed for learning in heterogeneous environments.
Beyond the design of robust aggregation mechanisms, this thesis also considers their integration within a practical Federated Learning framework. This system-level perspective highlights the importance of making the training process observable and interpretable, so that aggregation decisions are not treated as black-box operations. Overall, the thesis contributes to the development of Federated Learning systems that can operate under partial observability and partial trust, and that are better suited to real-world distributed environments where data cannot be centralized and client reliability cannot be assumed.