Federated Learning for Privacy-Preserving Internet of Things

Abstract

The rapid evolution of the Internet of Things (IoT) and relevant applications have been paving the way toward the fulfillment of smart cities. Smart cities are thought to come up with multiple crucial smart IoT applications i.e., smart manufacturing, smart transportation, autonomous driving/flight, smart buildings, smart healthcare, smart grid, and so many others [1]. Effective deployment of these IoT applications and services and making them available for users usually entail a vast number of IoT devices. Conferring to some global statistics, the total of IoT devices is believed to rises to 125 billion by 2030. This in turn reflects expected large growth in the amount of IoT-generated data. some other statistics show that, by 2025, the size of IoT-generated data will grow up to reach 79.4 zettabytes (ZB). That incredible growth in the size of IoT networks and associated data volume afford fascinating opportunities for deep learning in IoT environments. To this end, one can use centralized deep learning frameworks (discussed in Chaps. 4-7) to support the design of smart IoT applications. Nevertheless, centralized deep learning solutions struggle with the intrinsic problem of leakage of privacy of users owing to the requirement to transmit the local data at IoT devices data to a central third-party server (i.e., cloud server or datacenter) for training purposes. Additionally, centralized deep learning could not be viable in the case of big volumes of data (e.g., astronomical data) and placed at various geographic locations.To tackle the above-mentioned challenge of big and geographically dispersed data, distributed deep learning solution emerged to disseminate learning workload across manifold computing agents or devices. Those manifold devices could later train many models at distributed locations simultaneously to construct a cohesive deep learning model [2]. Distributed learning empowers more efficient computing through concurrent training of deep learning architecture. Two types of learning approaches, namely data-parallel and model-parallel approaches, could be exploited Electronic Supplementary Material The online version of this chapter

Towards Risk-Free Trustworthy Artificial Intelligence: Significance and Requirements

“…Some academic and commercial institutions have made eforts to explore and develop vehicular networks to support this vision provided by 3GPP. Focussing on the issues of secure communication, studies have explored security, privacy, and trust in vehicular ad hoc networks (VANETs) [212][213][214]. Despite that, cybersecurity fndings in the application of 5G in connected vehicles have remained scarce [213,[215][216][217].…”

“…Focussing on the issues of secure communication, studies have explored security, privacy, and trust in vehicular ad hoc networks (VANETs) [212][213][214]. Despite that, cybersecurity fndings in the application of 5G in connected vehicles have remained scarce [213,[215][216][217]. Malina et al [218] explored various IoT applications to categorise potential privacy leaks and then developed an innovative mechanism to protect privacy and maintain security within IoT services for actual applications, such as smart healthcare systems, intelligent vehicular networks, and smart cities.…”