Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HierSFL: Local Differential Privacy-aided Split Federated Learning in Mobile Edge Computing

Minh K. Quan, Dinh C. Nguyen, Van-Dinh Nguyen, Mayuri Wijayasundara, Sujeeva Setunge, Pubudu N. Pathirana

TL;DR

HierSFL addresses privacy and resource constraints in mobile edge computing by introducing a three-tier hierarchical split federated learning framework that leverages edge servers and a cloud server for model aggregation. It integrates Local Differential Privacy at both the client and edge levels and provides guidelines for optimal aggregation intervals to balance computation and communication costs. Empirical results on CIFAR-10 and MNIST show HierSFL outperforms traditional FL, SFL, and HFL in accuracy and convergence speed, while reducing training time and central server load. This approach enables faster content delivery and improved mobile service quality in MEC environments, with future work focusing on adaptive privacy budgets to further optimize the privacy-utility trade-off.

Abstract

Federated Learning is a promising approach for learning from user data while preserving data privacy. However, the high requirements of the model training process make it difficult for clients with limited memory or bandwidth to participate. To tackle this problem, Split Federated Learning is utilized, where clients upload their intermediate model training outcomes to a cloud server for collaborative server-client model training. This methodology facilitates resource-constrained clients' participation in model training but also increases the training time and communication overhead. To overcome these limitations, we propose a novel algorithm, called Hierarchical Split Federated Learning (HierSFL), that amalgamates models at the edge and cloud phases, presenting qualitative directives for determining the best aggregation timeframes to reduce computation and communication expenses. By implementing local differential privacy at the client and edge server levels, we enhance privacy during local model parameter updates. Our experiments using CIFAR-10 and MNIST datasets show that HierSFL outperforms standard FL approaches with better training accuracy, training time, and communication-computing trade-offs. HierSFL offers a promising solution to mobile edge computing's challenges, ultimately leading to faster content delivery and improved mobile service quality.

HierSFL: Local Differential Privacy-aided Split Federated Learning in Mobile Edge Computing

TL;DR

HierSFL addresses privacy and resource constraints in mobile edge computing by introducing a three-tier hierarchical split federated learning framework that leverages edge servers and a cloud server for model aggregation. It integrates Local Differential Privacy at both the client and edge levels and provides guidelines for optimal aggregation intervals to balance computation and communication costs. Empirical results on CIFAR-10 and MNIST show HierSFL outperforms traditional FL, SFL, and HFL in accuracy and convergence speed, while reducing training time and central server load. This approach enables faster content delivery and improved mobile service quality in MEC environments, with future work focusing on adaptive privacy budgets to further optimize the privacy-utility trade-off.

Abstract

Federated Learning is a promising approach for learning from user data while preserving data privacy. However, the high requirements of the model training process make it difficult for clients with limited memory or bandwidth to participate. To tackle this problem, Split Federated Learning is utilized, where clients upload their intermediate model training outcomes to a cloud server for collaborative server-client model training. This methodology facilitates resource-constrained clients' participation in model training but also increases the training time and communication overhead. To overcome these limitations, we propose a novel algorithm, called Hierarchical Split Federated Learning (HierSFL), that amalgamates models at the edge and cloud phases, presenting qualitative directives for determining the best aggregation timeframes to reduce computation and communication expenses. By implementing local differential privacy at the client and edge server levels, we enhance privacy during local model parameter updates. Our experiments using CIFAR-10 and MNIST datasets show that HierSFL outperforms standard FL approaches with better training accuracy, training time, and communication-computing trade-offs. HierSFL offers a promising solution to mobile edge computing's challenges, ultimately leading to faster content delivery and improved mobile service quality.
Paper Structure (13 sections, 7 equations, 5 figures, 2 tables, 1 algorithm)

This paper contains 13 sections, 7 equations, 5 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Motivation and Key Contributions
  3. Paper Organization
  4. System Design
  5. Learning Problem Overview
  6. Federated Learning (FL)
  7. Split Federated Learning (SFL)
  8. HierSFL Framework
  9. Local Differential Privacy (LDP) in $\mathsf{HierSFL}$
  10. Simulations and Evaluation
  11. Simulation Settings
  12. Simulation Results
  13. Conclusion

Figures (5)

  • Figure 1: The workflow of $\mathsf{HierSFL}$ framework.
  • Figure 2: Fine-tuning MES and client numbers for optimal CIFAR-10 accuracy.
  • Figure 3: Fine-tuning MES and Client Numbers for Optimal MNIST Accuracy.
  • Figure 4: Loss values comparison of FL, SFL, HFL and $\mathsf{HierSFL}$ on MNIST dataset.
  • Figure 5: Impact of privacy budget $\varepsilon$ on training accuracy and aggregation time for $\mathsf{HierSFL}$ with 4 MESs and 20 clients.