COMMITTEE CHAIR: Dr. Cajetan Akujuobi
CO-COMMITTEE CHAIR: Dr. Justin Foreman

TITLE: MONETIZATION OF CROWD-SOURCED FOG NODE SERVICES USING BLOCKCHAIN AND SMART CONTRACTS AND THE ADAPTATION OF ML FOR DATA REDUCTION

ABSTRACT: T Fog computing is increasingly becoming the building block for the explosive growth in edge computing as it affords the edge all the capabilities of the cloud with low latency and more decongested internet traffic. It accounts for the limitations found in IoT and other edge devices regarding memory, CPU, and bandwidth. While firms are providing these fog nodes, there remains the issue of data ownership, pricing fairness, and the amounts charged to customers based on the quality of services received. Our work proposes a decentralized blockchain-based fog paradigm that addresses these issues and provides a platform for users to contribute to the fog network and get incentives when their contributed nodes are used for fog services. Our experiment shows that fairness can be achieved by the users and fog nodes submitting reports of the services received at the end of every connection. An independent smart contract reviews this report, and the proper charge is levied on the user based on the services received. We provided a mechanism that punishes offenders and dishonest participants by assigning a trust score to the users and a fog node score to the fogs, and deductions or additions are made to the scores based on honest or dishonest reports. The scores are reviewed before a participant gets involved in another connection, with a lower score receiving lower task priority assignments for both fog and user devices. A zero-trust score kicks the device out of the network. The cost of implementing the system also shows that the system is cost-effective, and the evolution of the Ethereum network moving from proof of work to proof of stake greatly reduced the gas fees. The system met the security Confidentiality, Integrity, Authorization, Availability, and Nonrepudiation (CIAAN) requirement. IoT devices dispense data at a rate that exceeds its capacity to process. This has led to the introduction of fog nodes to support the limited capacity of IoT devices. More than introducing fog nodes is required; this data must be further reduced to sizes that these devices can handle for processing analysis and economic overheads. Our work introduced a system that efficiently handled this by building a machine learning model that utilized the math of Principal Component Analysis and Singular Value Decomposition (PCA/SVD) for data reduction. The unique value of this combination of data reduction and feature selection methods shows that while the data was greatly decreased, the feature of the data was retained. This was verified using standard benchmark datasets and a large private IoT dataset to verify the system’s effectiveness.to verify the system’s effectiveness.

Keywords: Internet of Things, Blockchain, Smart Contract Ethereum, Data Reduction, Edge Computing,  Machine Learning, Principal Component Analysis, Singular Value Decomposition

Room Location: Electrical Engineering Building Conference, Room 315D.