Implementing server-side federated learning in an edge-cloud framework for precision aquaculture

Ng, Bryan Jing Hong (2025) Implementing server-side federated learning in an edge-cloud framework for precision aquaculture. Final Year Project, UTAR.

PDF Download (6Mb)

Abstract

This project is about a growing need in the aquaculture industry which is precision aquaculture. Precision aquaculture involves the use of smart technologies such as sensors, cloud computing, and artificial intelligence to monitor and manage fish farming environments. However, small-scale farmers still face major challenges such as high implementation costs, poor internet connectivity, and concerns about data privacy. This project focuses on two key problems which are data privacy and system scalability. Most existing systems rely heavily on cloud connectivity and do not provide secure or efficient solutions for farms in remote areas with limited internet access. In this project, research and literature reviews were conducted to explore the current technologies in smart aquaculture, federated learning, and data privacy. Reviews include systems using IoT and AI-based models, along with analysis of different federated learning algorithms such as FedSGD, FedAvg, and FedProx, and privacypreserving methods such as DA, SA, HE and CKKS encryption. After identifying the gaps in current systems, this project proposes a secure and scalable server-side federated learning framework in an edge-cloud architecture for precision aquaculture. The system is designed to enable encrypted model training at the edge using IoT sensors and Raspberry Pi, while a federated learning server hosted on AWS aggregates updates without accessing raw data. The final product integrates edge computing, federated learning, encryption and cloud storage to create a privacy-preserving and cost-effective solution for monitoring aquaculture environments and supporting small-scale farmers.

Actions (login required)