Date of Award

6-2-2025

Degree Type

Thesis

Degree Name

Master of Science (M.S.)

Department

Electrical & Computer Engineering

First Advisor

Kamrul Hasan

Abstract

In modern times, electric utility companies are utilizing on-ground and aerial drone imagery to identify physical defects on transmission and distribution infrastructure to plan for more effective maintenance. To deploy faster and more cost-efficient maintenance decisions, utility companies are considering developing automation tools powered by AI. However, for AI technology to be effectively trained for real-world implementation, there is a need for cross-collaboration among a vast number of utility companies to share their imagery and analysis to improve the global AI model. The primary challenge of this collaboration comes at the expense of requiring data cleaning, a manual process to ensure sensitive data is removed from shared imagery. In this paper, we present a HE-FL framework that leverages homomorphic encryption and federated learning techniques to allow this secured collaboration among utility companies in the effort to improve upon an universal defect detection system. The proposed HE-FL global model was successful in benefitting, at minimum, the worst performing individual client or was able to perform on par or better than the other clients. This was achieved while maintaining local client data privacy through the federated learning scheme, as well as maintaining data security with communication to the central aggregator by applying partial homomorphic encryption to the model updates and newly updated HE-FL global model.

Download

Share

COinS