Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Jamming Detection in Cell-Free MIMO with Dynamic Graphs

Ali Hossary, Laura Crosara, Stefano Tomasin

TL;DR

This work addresses jamming detection in cell-free MIMO networks where topologies evolve over time. It introduces a dynamic-graph representation and a GCN-Transformer pipeline that jointly captures spatial connectivity and temporal patterns to detect jamming as anomalies. Key findings show that mixed-time-pattern training markedly improves generalization across fading and non-fading channels, while persistent jamming reveals a baseline-shift effect that challenges traditional anomaly detection. The approach offers a robust, scalable framework for secure, dynamic wireless networks and informs training strategies for real-world deployment.

Abstract

Jamming attacks pose a critical threat to wireless networks, particularly in cell-free massive MIMO systems, where distributed access points and user equipment (UE) create complex, time-varying topologies. This paper proposes a novel jamming detection framework leveraging dynamic graphs and graph convolutional neural networks (GCN) to address this challenge. By modeling the network as a dynamic graph, we capture evolving communication links and detect jamming attacks as anomalies in the graph evolution. A GCN-Transformer-based model, trained with supervised learning, learns graph embeddings to identify malicious interference. Performance evaluation in simulated scenarios with moving UEs, varying jamming conditions and channel fadings, demonstrates the method's effectiveness, which is assessed through accuracy and F1 score metrics, achieving promising results for effective jamming detection.

Jamming Detection in Cell-Free MIMO with Dynamic Graphs

TL;DR

This work addresses jamming detection in cell-free MIMO networks where topologies evolve over time. It introduces a dynamic-graph representation and a GCN-Transformer pipeline that jointly captures spatial connectivity and temporal patterns to detect jamming as anomalies. Key findings show that mixed-time-pattern training markedly improves generalization across fading and non-fading channels, while persistent jamming reveals a baseline-shift effect that challenges traditional anomaly detection. The approach offers a robust, scalable framework for secure, dynamic wireless networks and informs training strategies for real-world deployment.

Abstract

Jamming attacks pose a critical threat to wireless networks, particularly in cell-free massive MIMO systems, where distributed access points and user equipment (UE) create complex, time-varying topologies. This paper proposes a novel jamming detection framework leveraging dynamic graphs and graph convolutional neural networks (GCN) to address this challenge. By modeling the network as a dynamic graph, we capture evolving communication links and detect jamming attacks as anomalies in the graph evolution. A GCN-Transformer-based model, trained with supervised learning, learns graph embeddings to identify malicious interference. Performance evaluation in simulated scenarios with moving UEs, varying jamming conditions and channel fadings, demonstrates the method's effectiveness, which is assessed through accuracy and F1 score metrics, achieving promising results for effective jamming detection.
Paper Structure (18 sections, 11 equations, 5 figures)

This paper contains 18 sections, 11 equations, 5 figures.

Table of Contents

  1. Introduction
  2. System Model
  3. User Assignment Rule
  4. Jammer Behavior
  5. Jamming Detection By Dynamic Graph
  6. Jamming Detection Technique
  7. Model Training
  8. Numerical Results
  9. Dataset Generation
  10. GNN Implementation
  11. Performance Metrics
  12. Simulation Results
  13. $\tau=10$ Specialist Training Analysis
  14. Mixed-$\tau$ Training Performance
  15. Channel Fading Effects on Detection Performance
  16. ...and 3 more sections

Figures (5)

  • Figure 1: Architecture of the proposed jamming detection model.
  • Figure 2: Accuracy and F1 score vs $\tau$, for the deterministic scenario. Training performed with a dataset having $\tau =10$.
  • Figure 3: Accuracy and F1 score vs $\tau$, for the fading scenario. Training performed with a dataset having $\tau=10$.
  • Figure 4: Accuracy and F1 score vs $\tau$, for the deterministic scenario. Training performed with a dataset having a mixture of attacks with different values of $\tau$.
  • Figure 5: Accuracy and F1 score vs $\tau$, for the random fading scenario. Training performed with a dataset having a mixture of attacks with different values of $\tau$.