A Fairer Client Selection Framework for Federated Internet of Things: Equality, Equity, and Trade-off Perspectives

The proliferation of the Internet of Things (IoT) devices has led to the generation of vast amounts of data containing users' private information. Federated Learning (FL) has emerged as a transformative approach to harness large volumes of data generated by IoT devices for intelligent insights while maintaining individual privacy. In FL-based IoT environments, where devices are highly diverse and operate under varying conditions, performance can fluctuate due to multiple factors, i.e., mobility, network congestion, and energy constraints. This variability contributes to the inherent heterogeneity, thereby posing significant challenges in maintaining performance consistency and fairness in resource distribution among participants in FL training. Through an extensive review of the state-of-the-art literature, we identify gaps in fairness-aware client selection strategies, particularly with dual dimensions of fairness: equality and equity. This paper explores a conceptual framework for fairness-aware client selection, formulates three research questions based on existing research gaps, discusses preliminary insights, and outlines key directions for future research, including algorithm design and performance evaluation under fairness constraints.