This paper highlights the need for research to improve model fairness in Federated Learning (FL) environments, focusing on addressing the issue of bias across diverse clients. It highlights the limitations of existing fairness-enhancing FL solutions, highlighting the problems of failing to consider multiple sensitive attributes or overlooking unfairness at the individual client level. To address these issues, we propose a comprehensive benchmarking framework for fairness-aware FL at both the global and client levels. This framework provides the \fairdataset library, which enables fairness assessment in client environments with varying biases, four bias-heterogeneous datasets, and fairness assessment functions for these datasets.
