E-CIS: Edge-based classifier identification scheme in green & sustainable IoT smart city. (December 2021)

• Record Type: Journal Article
• Authors: Sun, Yi; Liu, Jie; Bashir, Ali Kashif; Tariq, Usman; Liu, Wei; Chen, Keliang; Alshehri, Mohammad Dahman
• Is Part Of: Sustainable cities and society. Volume 75(2021)
• Publication Date: 2021-12
• Subjects: IoT -- Edge computing -- Classifier identification -- Massive IoT device management -- IoT Device security & cyber defense; Sustainable urban development -- Periodicals; Sustainable buildings -- Periodicals; Urban ecology (Sociology) -- Periodicals; 307.76
• DOI: 10.1016/j.scs.2021.103312
• Languages: English
• ISSNs: 2210-6707

Abstract

Highlights: E-CIS changes the traditional centralized architecture and realizes low time delay and efficient identification of IoT devices based on edge computing. The types of IoT devices E-CIS can identify keeps growing. All gateways of E-CIS have the continuous learning ability synchronously. Abstract: Smart city has brought the unprecedented development and application of Internet of things (IoT) devices. Meanwhile, since both of the quantity and the type of IoT devices are growing rapidly, how to quickly identify the type of IoT devices is of paramount importance, especially in the fields of IoT Device Security, IoT Forensics, Cyber Defense, and Cyber Threats Intelligence Sharing, to make the IoT smart city green and sustainable. Traditional identification mode based on device or gateway often suffers from their limited computing and storage resources. Our work is motivated by the observation of the emergence of edge computing, in which computing and storage servers are placed in close proximity to IoT/mobile devices. In this paper, we propose an Edge-based Classifier Identification Scheme (E-CIS) for IoT Devices, where the neighboring edge servers provide powerful computing and storage capabilities. E-CIS changes the traditional centralized architecture and realizes low time delay and efficient identification of IoT devices based on edge computing. Experiments show that the average identification accuracy is as high as 99.2%. Besides, the optimization and security of the classification model can be maintained by the edge servers at the same time.