The Cybersecurity and Distributed Federated Learning project represents an international research exchange and collaborative program between the University of Murcia and the University of Zurich (UZH) (Switzerland). Conducted from November 2025 to January 2026, the project was hosted by the Communication Systems Group (CSG) at the UZH Department of Informatics, under the direction of Prof. Burkhard Stiller and sponsored by the Fundación Séneca (Región de Murcia).
Project Overview
In trust-constrained and decentralized cross-border settings, establishing resilient and secure collaborative learning requires robust mechanisms to identify and isolate adversarial anomalies. As participants train models over public and shared networks, they must protect their infrastructure against malicious model updates, communication attacks, and metadata poisoning.
This international collaboration investigated threat mitigation architectures for peer-to-peer federated learning. By combining Zurich's expertise in distributed ledgers and reputation monitoring with Murcia's expertise in cybersecurity and DFL, the project explored security layers to mitigate active network threats and poisoned clients.
Key Outcomes
- Reputation-based Client Isolation: Studied distributed reputation models that score clients based on the quality and validity of their local model updates.
- Ledger-assisted Aggregation: Explored distributed ledger primitives to record and audit aggregation steps, reducing opportunities for nodes to manipulate shared history.
- Cross-Border Security Simulation: Analyzed federated intrusion detection scenarios across UZH and UMU research contexts to assess DFL robustness against adversarial conditions.
This project was funded by UZH, UMU, and the Fundación Séneca under the international research exchange framework. For inquiries regarding ledger-assisted federated learning or international security exchanges, please contact me at enriquetomas@um.es.
