FedHome: Cloud-Edge based Personalized Federated Learning for In-Home Health Monitoring

Abstract: In-home health monitoring has attracted great attention for the ageing population worldwide. With the abundant user health data accessed by Internet of Things (IoT) devices and recent development in machine learning, smart healthcare has seen many successful stories. However, existing approaches for in-home health monitoring do not pay sufficient attention to user data privacy and thus are far from being ready for large-scale practical deployment. In this paper, we propose FedHome, a novel cloud-edge based federated learning framework for in-home health monitoring, which learns a shared global model in the cloud from multiple homes at the network edges and achieves data privacy protection by keeping user data locally. To cope with the imbalanced and non-IID distribution inherent in user's monitoring data, we design a generative convolutional autoencoder (GCAE), which aims to achieve accurate and personalized health monitoring by refining the model with a generated class-balanced dataset from user's personal data. Besides, GCAE is lightweight to transfer between the cloud and edges, which is useful to reduce the communication cost of federated learning in FedHome. Extensive experiments based on realistic human activity recognition data traces corroborate that FedHome significantly outperforms existing widely-adopted methods.

• The paper introduces FedHome, a framework that leverages a cloud-edge architecture and a generative convolutional autoencoder to address non-IID and imbalanced data challenges.
• It personalizes federated learning by tailoring models at the client level, achieving up to 95.41% accuracy in imbalanced data scenarios.
• By integrating homomorphic encryption and local data processing, the framework preserves data privacy and reduces communication overhead for in-home health monitoring.

FedHome: Cloud-Edge based Personalized Federated Learning for In-Home Health Monitoring

The paper introduces FedHome, an innovative framework designed for in-home health monitoring that leverages personalized federated learning within a cloud-edge architecture. Recognizing the growing demands for healthcare for the ageing global population, FedHome aims to integrate machine learning, data privacy, and Internet of Things (IoT) capabilities to offer an efficient, privacy-preserving solution for health data processing and analysis.

Key Contributions

1. Cloud-Edge Computing Integration: FedHome employs a synergistic architecture combining cloud and edge computing. By processing data at or near the source (edge level), FedHome not only ensures timely response for health monitoring but also adheres to privacy constraints by reducing the need for centralized data aggregation.
2. Generative Convolutional Autoencoder (GCAE): Central to FedHome’s architecture is the GCAE, which addresses non-IID and imbalanced data distribution challenges inherent in decentralized health data. The GCAE is lightweight, allowing it to operate effectively within the constraints of communication bandwidth and computational capabilities typical of edge environments. By generating class-balanced datasets locally, GCAE enhances both model accuracy and personalization.
3. Federated Learning with Personalization: Unlike traditional federated learning models that train a single global model, FedHome incorporates mechanisms for personalizing models at the client level. This personalization is crucial in accommodating the unique health profiles of individuals while leveraging the shared model’s generalizability.
4. Privacy Preservation: Through federated learning, FedHome retains data privacy by keeping user data locally at the edge level. The use of homomorphic encryption furthers the platform’s capability to protect model parameters during network transfers, ensuring that personal health data remains secure.

Experimental Results

FedHome’s efficacy is demonstrated using a realistic dataset for human activity recognition. The research presents comprehensive evaluations under various data distributions:

• In both balanced and imbalanced data scenarios, FedHome achieves high accuracy, notably outperforming both conventional centralized learning algorithms and other federated learning models. Specifically, in imbalanced data distribution, FedHome achieves an accuracy of 95.41%.
• Compared to other federated models like FL-MLP and FL-CNN, FedHome offers improved accuracy with reduced model parameter sizes, suggesting both computational and communicational efficiency.

Implications and Future Directions

The introduction of FedHome lays a crucial foundation for future advancements in decentralized health monitoring scenarios. The framework’s scalability and adaptability to various health monitoring tasks illustrate its potential to become a significant tool in smart healthcare systems. The integration of advanced encryption techniques with federated learning models marks a progressive step toward sustainable privacy-aware AI solutions.

As IoT devices and smart healthcare applications continue to evolve, FedHome’s principles could be extended to other areas of personalized health analytics. Its framework can adapt to incorporate other neural network architectures suitable for specific health monitoring tasks, offering a flexible and robust solution for diverse environments.

Conclusion

In summary, FedHome represents an advancement in the domain of privacy-preserving health monitoring solutions. By addressing key challenges in federated learning through its innovative GCAE design and personalized approach, the framework paves the way for deploying effective, secure, and tailored health monitoring systems in home environments. As smart healthcare technologies advance, FedHome’s underlying framework is well-positioned to support the needs of a rapidly ageing population.