Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Learning for Resilient Adversarial Decision Fusion in Byzantine Networks

Kassem Kallas

TL;DR

The paper tackles resilient decision fusion under Byzantine attacks by deriving a unified Bayesian framework that covers i.i.d., Markovian, synchronized/unsynchronized, and unbalanced-prior scenarios. It then introduces a deep learning solution that directly maps the reports matrix $\mathbf{R}$ to the system state $\hat{\mathbf{s}}$, trained on a globally constructed adversarial dataset to generalize across diverse attack models without explicit parameter knowledge. Extensive simulations show the DL approach outperforms traditional fusion rules in accuracy and robustness while remaining computationally efficient for real-time use. The work highlights the potential of deep learning to transform Byzantine-resilient decision fusion and outlines pathways for scalable, edge-friendly deployments and improved interpretability.

Abstract

This paper introduces a deep learning-based framework for resilient decision fusion in adversarial multi-sensor networks, providing a unified mathematical setup that encompasses diverse scenarios, including varying Byzantine node proportions, synchronized and unsynchronized attacks, unbalanced priors, adaptive strategies, and Markovian states. Unlike traditional methods, which depend on explicit parameter tuning and are limited by scenario-specific assumptions, the proposed approach employs a deep neural network trained on a globally constructed dataset to generalize across all cases without requiring adaptation. Extensive simulations validate the method's robustness, achieving superior accuracy, minimal error probability, and scalability compared to state-of-the-art techniques, while ensuring computational efficiency for real-time applications. This unified framework demonstrates the potential of deep learning to revolutionize decision fusion by addressing the challenges posed by Byzantine nodes in dynamic adversarial environments.

Deep Learning for Resilient Adversarial Decision Fusion in Byzantine Networks

TL;DR

The paper tackles resilient decision fusion under Byzantine attacks by deriving a unified Bayesian framework that covers i.i.d., Markovian, synchronized/unsynchronized, and unbalanced-prior scenarios. It then introduces a deep learning solution that directly maps the reports matrix to the system state , trained on a globally constructed adversarial dataset to generalize across diverse attack models without explicit parameter knowledge. Extensive simulations show the DL approach outperforms traditional fusion rules in accuracy and robustness while remaining computationally efficient for real-time use. The work highlights the potential of deep learning to transform Byzantine-resilient decision fusion and outlines pathways for scalable, edge-friendly deployments and improved interpretability.

Abstract

This paper introduces a deep learning-based framework for resilient decision fusion in adversarial multi-sensor networks, providing a unified mathematical setup that encompasses diverse scenarios, including varying Byzantine node proportions, synchronized and unsynchronized attacks, unbalanced priors, adaptive strategies, and Markovian states. Unlike traditional methods, which depend on explicit parameter tuning and are limited by scenario-specific assumptions, the proposed approach employs a deep neural network trained on a globally constructed dataset to generalize across all cases without requiring adaptation. Extensive simulations validate the method's robustness, achieving superior accuracy, minimal error probability, and scalability compared to state-of-the-art techniques, while ensuring computational efficiency for real-time applications. This unified framework demonstrates the potential of deep learning to revolutionize decision fusion by addressing the challenges posed by Byzantine nodes in dynamic adversarial environments.

Paper Structure

This paper contains 17 sections, 18 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Unified Derivation for Adversarial Decision Fusion Rules
  3. Deep Learning Approach
  4. Simulation results and discussion
  5. Simulation Setup
  6. Dataset Generation and Experiment Overview
  7. DNN Model Architecture
  8. Comprehensive Evaluation of Model Performance and Robustness
  9. Performance Comparison with State-of-the-Art Methods
  10. Evaluation of i.i.d and Markovian Data
  11. Performance under Different Adversarial Scenarios
  12. Global Dataset: Comprehensive Adversarial Settings
  13. Impact of Varying Observation Window and Network Size
  14. Effect of Number of Samples Per Class
  15. Execution Time Analysis
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Problem Setup.
  • Figure 2: Model accuracy vs. $\alpha$ for i.i.d and Markovian data.
  • Figure 3: Performance comparison for Unbalanced Priors, Adaptive Strategies, and Fixed Byzantine Nodes.
  • Figure 4: Model performance under varying $m$ and $n$.
  • Figure 5: Model performance vs. number of samples per class.