FoG-FL: Federated learning-aided FoG computing for honest node selection and efficient data aggregations

Federated learning (FL) has grown in popularity among universities and industrial businesses in processing and analyzing large amounts of data from IoT devices. Maintaining FL efficiency by preserving the privacy of IoT device input data is problematic due to its unreliability. To overcome these difficulties, this proposed method offers a secure aggregation mechanism using effectual additive secret allotment in the fog computing (FC) environment. FL training requires regular secure aggregation; hence, the protocol should minimize communication latency and processing time. We deploy a fog node (FN) as intermediary units to deliver local request that help the cloud aggregate the sum of tasks. We designed a FogFL mechanism to ensure our proposed method is robust against dropped user tasks. Our protocol offers the selection of new fog aggregator nodes in a fog environment to reduce the latency. We analyze and compare the performance of the proposed method with that of existing systems. The proposed FoG-FL outperforms other methods and achieved an increment up to $8.5 \%$ of global accuracy as compared with reported methods.