Advancements in federated fog architectures for enhanced quality-of-service in IoT and IoV applications

Abstract

Network delays cause disturbance and reduction in the Quality-of-Service (QoS) for Internet-of- Things (IoT) and Internet-of-Vehicles (IoV) applications while end-users are running critical real-time services. Federated fog computing emerged as a viable solution to overcome such a problem. By merging resources from multiple fog providers and agreeing on a service level agreement, the federated infrastructure can offer the opportunity to adapt to dynamic environmental changes and facilitate vehicle-to-vehicle communication. This thesis aims to contribute to the comprehensive and secure fog federation architecture to enhance the QoS for IoT in general, and autonomous driving applications in IoV in specific. It focuses on several challenges, including a lack of investigation into a comprehensive federated fog architecture, potential instability in fog federations, lack of support for mobility in IoV, complications in federated learning for IoV, and the impact of untrustworthy fog providers. The main objectives of this thesis include developing a comprehensive and efficient federated fog computing architecture, creating a robust formation mechanism to limit providers from switching federations, extending fog federation formation to support mobility, supporting vehicular federated learning applications, and ensuring that fog federation formation considers trust and reputation during the formation of the architecture and its maintaining phases. We complement the current research progress about federated fog in the literature by adding various modules to facilitate the goal of enhancing the QoS of Federated Learning applications in IoV. First, we investigate a novel architecture for the federated fog concept and propose an adaptive, intelligent, and dynamic federation formation approach using Machine Learning and Genetic Algorithms. Moreover, we address the problem of instability within fog federations by proposing a decentralized algorithm based on evolutionary game theory. Furthermore, we expand our area to cover a more dynamic environment; the Internet-of-Vehicles. To satisfy mobile users, mobility should be considered, thus, we rely on a fog federation formation mechanism on the fly where we consider the mobility of the devices to provide them a good service quality using game theory. In addition, we present a horizontal-based federated learning architecture empowered by fog federations to support on-device training with the requested QoS. Finally, we extend the formation mechanism used for federated fog computing by introducing a Blockchain infrastructure to handle the federations’ administrative tasks and secure the formation process. Real datasets are used to evaluate the proposed architecture and formation mechanisms. The results show a notable improvement in the throughput and a decrease in the response time for the services requested, in addition to stabilizing the fog federations.

Date	12 Oct 2023
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Zbigniew Dziong (Supervisor), Hadi Otrok (Co-supervisor) & Azzam Mourad (Co-supervisor)