Adaptive Threat Intelligence through Federated Learning for Secure Distributed Network Environments with Heterogeneous Nodes

Abstract

The rise of distributed networks with heterogeneous devices has brought increased vulnerability to cyber-attacks, particularly in environments where data centralization is infeasible due to privacy or infrastructure constraints. This paper proposes a Federated Learning (FL) approach to enable adaptive threat intelligence across decentralized, multi-node systems. Our framework leverages collaborative learning among heterogeneous nodes to detect threats in real-time without sharing raw data. We evaluate the system across multiple threat categories, demonstrating an increase of 13.7% in detection accuracy over traditional centralized systems while reducing response latency by 22%. The proposed model exhibits resilience to data and device heterogeneity, offering a secure, scalable solution for modern cyber-defense architecture.