Federated learning based caching in fog computing for future smart cities. Issue 1 (28th September 2020)

• Record Type: Journal Article
• Authors: Sharma, Sushant; Gupta, Nitin
• Is Part Of: Internet technology letters. Volume 5:Issue 1(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-09-28
• Subjects: caching -- cloud -- edge computing -- federated learning -- fog computing -- machine learning; Internet -- Periodicals; 004.67805
• DOI: 10.1002/itl2.225
• Languages: English
• ISSNs: 2476-1508

Abstract

Abstract : Edge devices in the Internet of Things (IoT) networks are responsible for the continuous generation and communication of massive chunks of transient data to the fog devices for data processing. However, resources at the fog nodes are limited. Therefore, objective of the work is efficient caching, which is achieved using Federated Learning (FL), where the data used for learning is not gathered centrally but remains where it is produced. Therefore, the heavy transmission of the data to the fog nodes for learning is not required. Simulation results depict the advantages of using this approach over other centralized approaches. Abstract : In the future smart cities, everything will be connected through intelligent IoT and edge devices. These edge and IoT devices produce huge data, which needs to be communicated to the central cloud for storing, processing and analyzing which leads to lot of issues like congestions, delays and privacy issues. Fog Computing is a new method of computing which brings the contents, resources and services provided by the cloud closer to the edge of the network. It reduces the delays as the transmission of data is first to be done to the fog nodes which are closer to the edge. The fog nodes do not have the enough resources to store all the data communicated to it. Due to limited cache size, it is important to know that which data is useful and required to be cached and which is not. This work considers Federated learning in which, instead of sending raw data to the server, the edge devices send model updates to the server where multiple models are aggregated to form a global model.