Decentralized Anomaly Detection in Network Traffic using Federated Learning and Graph Neural Networks

Abstract

This paper proposes a novel approach to network anomaly detection leveraging the strengths of federated learning and graph neural networks (GNNs). Traditional centralized anomaly detection systems suffer from single points of failure and privacy concerns. Our decentralized system addresses these limitations by enabling each network device to independently train a GNN model using federated learning. Each device learns a local representation of its network traffic data, and anomalies are detected by comparing the device's local representation with the global model. The use of GNNs allows for the modeling of complex network topologies and relationships between devices, improving the accuracy of anomaly detection. The federated learning component ensures privacy preservation and reduces the need to transmit sensitive data to a central server. This approach offers a robust and scalable solution for real-time anomaly detection in dynamic network environments. The key contribution lies in the synergistic combination of these techniques to create a resilient and privacy-conscious system.