A Federated Learning Service Ecosystem for Secure and Flexible Model Sharing in Multi-Cloud Environments

Naik, Dhyanesh (2024) A Federated Learning Service Ecosystem for Secure and Flexible Model Sharing in Multi-Cloud Environments. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

This research work proposes a relatively new serverless federated learning (FL) model-sharing mechanism for securing flexible multi-cloud environments. The research presents solutions to important issues in decentralized machine learning such as securing fine-grained access control and privacy-preserving aggregation. This proposed framework scales across multiple AWS accounts for secure data and model sharing using a decentralized architecture that resembles the federated learning process. It uses attributebased encryption (ABE) enabled by encapsulating the attributes into Advanced Encryption Standard (AES) to allow peer-to-peer model sharing across different AWS accounts without the need for central authorities. A masked-ring protocol is implemented for decentralized model aggregation to protect user privacy during training. We have implemented and evaluated our architecture on AWS Lambda to prove that it can be used on real-world serverless platforms. The framework is highly effective and scalable as demonstrated by experiments on multiple cloud accounts by training convolutional neural networks (CNN) on a subset of the MNIST datasets for local training and local model generation. The experimental framework is designed using AWS Lambda functions to distribute the dataset across clouds, begin local model training, encrypt, and use local masks before saving to S3 buckets. The host function can then be triggered to access these local models to perform masked ring aggregation for unmasking and decrypting the aggregated model. This research is part of the work that will advance privacy-encapsulated collaboration learning in multi-cloud, keeping focused on the pragmatic balance between security resourcing vs flexibility alongside performance.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Mijumbi, Rashid UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Cloud computing Q Science > QA Mathematics > Computer software > Computer Security T Technology > T Technology (General) > Information Technology > Computer software > Computer Security Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	04 Jul 2025 08:57
Last Modified:	04 Jul 2025 08:57
URI:	https://norma.ncirl.ie/id/eprint/8040

Actions (login required)

View Item