Robust Federated Learning on Edge Devices is an emerging paradigm that
addresses the growing need for privacy-preserving, scalable, and dependable
Machine Learning (ML) across distributed and resource-constrained. Due to the
proliferation of edge devices like smartphones, IoT sensors, wearables, and
embedded systems, they become increasingly essential data generators and
processors due to their widespread use. Federated Learning (FL) offers a promising
solution by enabling collaborative ML across multiple decentralized devices while
keeping data local and private. The evolutionary growth of edge devices has further
accelerated the FL by enabling data processing near the point of generation.
However, implementing FL on edge devices is inherently challenging due to the non-
IID nature of edge data, varying computational capabilities, limited energy and
memory, intermittent network connectivity, and dynamic device availability.
Moreover, robustness in FL is often compromised by adversarial participants, data
and model poisoning attacks, and system-level vulnerabilities. A robust federated
system must be resilient against such threats while ensuring model accuracy,
fairness, and reliability across all participating edge devices.
This workshop provides a dedicated forum for researchers, practitioners, and
industry professionals to discuss innovations, best practices, and future perspectives
for allowing effective FL at the edges. The goal is to explore frameworks, algorithms,
and application-specific approaches that enhance FL's reliability and security in
diverse edge environments. The workshop seeks to bring together experts from ML,
distributed systems, edge computing, cybersecurity, and data privacy to foster
interdisciplinary discussions and collaborations. It encourages submitting original
research, case studies, and experimental findings that push the boundaries of what
is achievable with FL in edge contexts. The workshop aims to bring research ideas in
following areas but not limited to,
• Robust model aggregation techniques that resist adversarial updates
• Defensive mechanisms against poisoning, backdoor, and inference attacks
• Handling data heterogeneity and imbalance across distributed edge devices
• Energy-efficient and resource-aware FL algorithms
• Robust FL under partial device participation and dropout scenarios
• Personalized FL with robustness to user-specific variations
• Incentive and reputation mechanisms to encourage reliable participation
• Formal guarantees of convergence for robust FL
• Case studies and benchmarks in FL applications like healthcare, smart cities,
industrial IoT, etc.