Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cellular Traffic Prediction via Byzantine-robust Asynchronous Federated Learning

Hui Ma, Kai Yang, Yang Jiao

TL;DR

This work tackles privacy-preserving cellular traffic prediction in the presence of Byzantine clients by proposing BAFDP, an asynchronous differential privacy-enabled federated learning framework. It fuses distributionally robust optimization with a Wasserstein-ball uncertainty set to guard against distributional shifts induced by DP noise and Byzantine perturbations, while employing an L1-regularized augmented Lagrangian for robust aggregation. The authors provide convergence guarantees with an $O(1/oldsymbol{}^2)$ rate and validate the approach on Milano, Trento, and LTE datasets, showing superior accuracy and robustness vs. state-of-the-art baselines and a favorable privacy-accuracy-distributiveness trade-off. The results suggest practical impact for privacy-conscious, large-scale cellular traffic forecasting in distributed environments where adversarial behavior or latency variability may otherwise degrade performance.

Abstract

Network traffic prediction plays a crucial role in intelligent network operation. Traditional prediction methods often rely on centralized training, necessitating the transfer of vast amounts of traffic data to a central server. This approach can lead to latency and privacy concerns. To address these issues, federated learning integrated with differential privacy has emerged as a solution to improve data privacy and model robustness in distributed settings. Nonetheless, existing federated learning protocols are vulnerable to Byzantine attacks, which may significantly compromise model robustness. Developing a robust and privacy-preserving prediction model in the presence of Byzantine clients remains a significant challenge. To this end, we propose an asynchronous differential federated learning framework based on distributionally robust optimization. The proposed framework utilizes multiple clients to train the prediction model collaboratively with local differential privacy. In addition, regularization techniques have been employed to further improve the Byzantine robustness of the models. We have conducted extensive experiments on three real-world datasets, and the results elucidate that our proposed distributed algorithm can achieve superior performance over existing methods.

Cellular Traffic Prediction via Byzantine-robust Asynchronous Federated Learning

TL;DR

This work tackles privacy-preserving cellular traffic prediction in the presence of Byzantine clients by proposing BAFDP, an asynchronous differential privacy-enabled federated learning framework. It fuses distributionally robust optimization with a Wasserstein-ball uncertainty set to guard against distributional shifts induced by DP noise and Byzantine perturbations, while employing an L1-regularized augmented Lagrangian for robust aggregation. The authors provide convergence guarantees with an rate and validate the approach on Milano, Trento, and LTE datasets, showing superior accuracy and robustness vs. state-of-the-art baselines and a favorable privacy-accuracy-distributiveness trade-off. The results suggest practical impact for privacy-conscious, large-scale cellular traffic forecasting in distributed environments where adversarial behavior or latency variability may otherwise degrade performance.

Abstract

Network traffic prediction plays a crucial role in intelligent network operation. Traditional prediction methods often rely on centralized training, necessitating the transfer of vast amounts of traffic data to a central server. This approach can lead to latency and privacy concerns. To address these issues, federated learning integrated with differential privacy has emerged as a solution to improve data privacy and model robustness in distributed settings. Nonetheless, existing federated learning protocols are vulnerable to Byzantine attacks, which may significantly compromise model robustness. Developing a robust and privacy-preserving prediction model in the presence of Byzantine clients remains a significant challenge. To this end, we propose an asynchronous differential federated learning framework based on distributionally robust optimization. The proposed framework utilizes multiple clients to train the prediction model collaboratively with local differential privacy. In addition, regularization techniques have been employed to further improve the Byzantine robustness of the models. We have conducted extensive experiments on three real-world datasets, and the results elucidate that our proposed distributed algorithm can achieve superior performance over existing methods.

Paper Structure

This paper contains 29 sections, 2 theorems, 25 equations, 8 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related work
  3. Traffic Prediction based on Federated Learning
  4. Differential Privacy
  5. Byzantine Robustness
  6. Distributionally Robust Optimization
  7. Problem Formulation
  8. Differential Privacy
  9. Federated Learning based on Differential Privacy
  10. Distributionally Robust Optimization
  11. Method
  12. Uncertainty Set
  13. BAFDP Algorithm
  14. Convergence Analysis
  15. Experimental Setup
  16. ...and 14 more sections

Key Result

Proposition 1

If Assumption ass.2 holds, we can derive the following formulation,

Figures (8)

  • Figure 1: The diagram of BAFDP.
  • Figure 2: Visualization of prediction and ground truth on two open-sourced datasets.
  • Figure 3: Visualization of privacy level changes during the training process on three real-world datasets.
  • Figure 4: Comparison of training loss for synchronous and asynchronous distributed algorithm on three real-world datasets.
  • Figure 5: Comparison of RMSE for synchronous and asynchronous distributed algorithm on three real-world datasets.
  • ...and 3 more figures

Theorems & Definitions (5)

  • Definition 1
  • Proposition 1
  • Definition 2
  • Definition 3
  • Theorem 1