A privacy-aware federated learning framework for collaborative healthcare IoT security

Building reliable AI systems for healthcare remains challenging due to the sensitive and highly regulated nature of medical data. Healthcare institutions often manage fragmented datasets, making data sharing difficult. Federated Learning offers a privacy-preserving solution by enabling model training without transferring raw data; however, many existing FL approaches do not fully address key challenges in healthcare IoT settings, such as non-IID data, device heterogeneity, and unstable communication, limiting their effectiveness in real-world deployment. In this paper, we propose a modular privacy-aware FL integration framework designed explicitly for healthcare IoT security and distributed sensing scenarios. The framework combines TensorFlow Federated (TFF), differential privacy, and homomorphic encryption to enhance data protection throughout the training process. We also introduce a controlled data-sharing strategy that enables participating healthcare nodes to adjust the trade-off between privacy and model performance based on their own constraints. We designed system-level integration that makes privacy-preserving FL practically deployable in heterogeneous IoT healthcare settings. To evaluate our approach, we conduct experiments on the IoT-Healthcare-Security dataset and compare our framework with widely used FL baselines, including FedAvg, FedProx, and FedSGD. Our method achieves 98.53% accuracy and consistently outperforms baseline models while maintaining strong privacy guarantees.