Access Type

Open Access Dissertation

Date of Award

January 2022

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Electrical and Computer Engineering

First Advisor

Lubna Alazzawi

Second Advisor

Harpreet Singh

Abstract

The development of intelligent transportation systems (ITS) is aided by the advent of Internet-of-Vehicles (IoV), which is a decentralized network that allows connected vehicles and vehicular ad hoc networks to share data (VANETs). However, today's IoV networks face a number of challenges, including effective resource utilization, security and privacy, trust, information irregularity, etc. In addition, IoV applications have a wide range of Quality-of-Service (QoS) requirements, making it difficult to develop an efficient solution to deal with big data in IoV. Furthermore, the solution should be scalable and extendable, as well as lightweight and cost-effective to maintain. By outsourcing computationally-intensive operations to nearby situated fog nodes, fog computing tackles the fundamental weakness of centralized data processing in cloud computing. Furthermore, as the number of vehicles using the IoV architecture expands, new challenges and requirements emerge, such as scalability, resource efficiency, and secure communication.

In this research work, we look at load balancing, secure communication, privacy preservation, and trustworthy communication in SDN-enabled and fog-based IoV networks. Using reinforcement learning (RL) approaches, our propose a framework that efficiently distribute tasks in the fog-to-fog and vehicles-to-fog layers. Furthermore, since vehicle data is private and sensitive, further vigilance is required. Authentication of communicating devices is one example of a data security approach. Authentication is used to safeguard data transferred through public channels. Many protocols have been created; nevertheless, typical authentication methods cannot be readily applied to situations that need minimal latency. They're also ineffectual for two reasons: first, they can't keep up with the expanding volume of data collected, and second, they're vulnerable to cyber-attacks. As a result, we attempt to propose a viable solution that is totally resilient and solves the aforementioned issues in this work. We created a lightweight, fog-based authentication mechanism to protect data from IoV devices during transmission. Our method provides low communication costs while meeting high security requirements. Finally, we evaluate and compare the performance of our technique in terms of network parameters including throughput, end-to-end latency, and packet loss rate.

In addition, when private data is exchanged among fog nodes, privacy concerns arise, limiting the usefulness of IoV systems. We propose a Federated Learning-based (FL) and Blockchain-based system for privacy preservation in IoV to address this challenge. Traditional machine learning algorithms are not well suited for distributed and highly dynamic systems like IoV since they train on data with local features. As a result, FL is used to train the global model while preserving the privacy. In addition, our strategy is built on a reputation scheme that evaluates the reliability of vehicles participating in the FL training process. Furthermore, our solution makes use of blockchain technology to ensure trust across numerous communication nodes. All transactions, for example, take place on the blockchain when local learned model updates from vehicles and fog nodes are shared with the cloud to update the global learnt model. As a result of allowing reputable vehicles to update the global model, our proposed method improves the global model's accuracy, according to the results of our experimental study.

Download

Share

COinS