Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Federated Testing (FedTest): A New Scheme to Enhance Convergence and Mitigate Adversarial Attacks in Federating Learning

Mustafa Ghaleb, Mohanad Obeed, Muhamad Felemban, Anas Chaaban, Halim Yanikomeroglu

TL;DR

FedTest introduces a novel federated testing scheme where participants train locally and evaluate others, generating accuracy-based scores used for aggregation. By shifting testing responsibilities to edge users and using a weighted moving average over current and past rounds, FedTest mitigates non-IID data effects and reduces the influence of malicious contributors, while accelerating convergence. Empirical results on CIFAR-10 and MNIST show improved robustness to adversarial behavior and superior convergence speed compared with FedAvg and accuracy-based baselines. The approach relies on device-to-device testing and orthogonal resource blocks to preserve privacy and efficiency, with future work focusing on communication optimization and adaptive score weighting.

Abstract

Federated Learning (FL) has emerged as a significant paradigm for training machine learning models. This is due to its data-privacy-preserving property and its efficient exploitation of distributed computational resources. This is achieved by conducting the training process in parallel at distributed users. However, traditional FL strategies grapple with difficulties in evaluating the quality of received models, handling unbalanced models, and reducing the impact of detrimental models. To resolve these problems, we introduce a novel federated learning framework, which we call federated testing for federated learning (FedTest). In the FedTest method, the local data of a specific user is used to train the model of that user and test the models of the other users. This approach enables users to test each other's models and determine an accurate score for each. This score can then be used to aggregate the models efficiently and identify any malicious ones. Our numerical results reveal that the proposed method not only accelerates convergence rates but also diminishes the potential influence of malicious users. This significantly enhances the overall efficiency and robustness of FL systems.

Federated Testing (FedTest): A New Scheme to Enhance Convergence and Mitigate Adversarial Attacks in Federating Learning

TL;DR

FedTest introduces a novel federated testing scheme where participants train locally and evaluate others, generating accuracy-based scores used for aggregation. By shifting testing responsibilities to edge users and using a weighted moving average over current and past rounds, FedTest mitigates non-IID data effects and reduces the influence of malicious contributors, while accelerating convergence. Empirical results on CIFAR-10 and MNIST show improved robustness to adversarial behavior and superior convergence speed compared with FedAvg and accuracy-based baselines. The approach relies on device-to-device testing and orthogonal resource blocks to preserve privacy and efficiency, with future work focusing on communication optimization and adaptive score weighting.

Abstract

Federated Learning (FL) has emerged as a significant paradigm for training machine learning models. This is due to its data-privacy-preserving property and its efficient exploitation of distributed computational resources. This is achieved by conducting the training process in parallel at distributed users. However, traditional FL strategies grapple with difficulties in evaluating the quality of received models, handling unbalanced models, and reducing the impact of detrimental models. To resolve these problems, we introduce a novel federated learning framework, which we call federated testing for federated learning (FedTest). In the FedTest method, the local data of a specific user is used to train the model of that user and test the models of the other users. This approach enables users to test each other's models and determine an accurate score for each. This score can then be used to aggregate the models efficiently and identify any malicious ones. Our numerical results reveal that the proposed method not only accelerates convergence rates but also diminishes the potential influence of malicious users. This significantly enhances the overall efficiency and robustness of FL systems.
Paper Structure (10 sections, 5 figures, 1 algorithm)

This paper contains 10 sections, 5 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Traditional Federated Learning
  3. FedTest
  4. Results and Discussion
  5. Research Directions
  6. Optimizing the Communication Cost of FedTest
  7. Optimizing the Accuracy Score of FedTest
  8. Enhancing Models Evaluation by Identifying Malicious Users and Relying on Trusted Testers
  9. Mitigation of Adversarial Attack
  10. Conclusion

Figures (5)

  • Figure 1: FedAvg method where the models are aggregated based on the dataset sizes.
  • Figure 2: Flowchart of the proposed model, considering CIFAR-10 dataset.
  • Figure 3: The comparison between the FedTest and the accuracy-based procedures.
  • Figure 4: Convergence rate of the proposed FedTest compared to FedAvg and accuracy-based approaches using CIFAR-10 dataset.
  • Figure 5: Convergence rate of the proposed FedTest compared to FedAvg and accuracy-based approaches using MNIST dataset.