ERAFL: Efficient Resource Allocation for Federated Learning Training in Smart Homes

Research output: Contribution to conferencePaperpeer-review


Federated learning (FL), due to its privacy-by-design feature, has been widely adopted in smart homes, from device anomaly detection to traffic characterization. However, with the growing number of FL applications, it becomes crucial to manage communication and computation resources within the smart home so that FL applications can complete their training on time. While computation offloading has relieved the challenge of timely completion of applications in case of high competition for local resources, privacy of the smart home data remains a critical concern. For a smart home with FL training tasks, this paper introduces ERAFL, a resource allocation and computation offloading algorithm running on a home gateway. Unlike privacy-oblivious prior works, ERAFL considers privacy-sensitivity level of FL training data in offloading decision and aims to locally process the data with higher privacy-sensitivity level, e.g., biological personal data. Moreover, in case of insufficient local resources, ERAFL offloads a part of data and accelerates training by leveraging parallel training on the cloud and the edge device. To guarantee that the final model can achieve a certain level of accuracy, ERAFL puts a limit on the amount of offloaded data or performs the training either locally or remotely. Our simulation results show that ERAFL can satisfy more FL training tasks, offload less data and reduce data privacy leakage in comparison to the baselines that do not consider partial offloading, privacy sensitivity of application data or resource allocation. Also, the accuracy obtained from our approach and the convergence speed are comparable with traditional FL.
Original languageEnglish
Publication statusAccepted/In press - 2024


  • Federated Learning
  • Computation Offloading
  • Smart Home
  • IoT


Dive into the research topics of 'ERAFL: Efficient Resource Allocation for Federated Learning Training in Smart Homes'. Together they form a unique fingerprint.

Cite this