Merit: an on-demand IoT service delivery and resource scheduling scheme for federated learning and blockchain empowered 6G edge networks with reduced time and energy cost

Abstract

Federated learning (FL) can improve the privacy-preserving issue of users' IoT devices, in which users complete the local training and transfer the updated model data to the central server for a global update. Due to high latency, the central server-based FL may suffer from huge energy loss at local user devices. MEC-based FL can improve the model accuracy and energy consumption at user devices via edge server-based task execution. Along with FL, blockchain can improve data security via permission-based access. Existing works explored only single type of IoT task without any appropriate resource scheduling for multiple tasks with different preferences, FL, and blockchain operations. This paper provides a merit-based resource scheduling scheme for different tasks with preferences, blockchain, and FL operations by checking resources, deadlines, delays, and resource costs. The simulation results verify that 45% running time and 53% cost gain is achieved in proposed scheme over the baseline schemes.