Privacy-Preserving Threat Intelligence Sharing in 6G and IoT Networks Using Trust-Aware Federated Learning

Abstract

The proliferation of Internet of Things (IoT) devices in 6G cellular ecosystems demands robust, decentralized cybersecurity frameworks capable of handling high-velocity threat data without compromising user privacy. Centralized machine learning approaches risk data exposure and violate strict regulatory privacy standards. Federated learning (FL) enables collaborative model training across distributed edge nodes; however, traditional FL aggregation protocols remain highly vulnerable to client dropouts, statistical heterogeneity (non-IID data distributions), and malicious model-poisoning attacks. This paper introduces a trust-aware federated learning framework integrated with differential privacy and secure aggregation to defend 6G-enabled IoT environments against sophisticated poisoning attempts. Comprehensive evaluations across large-scale IoT benchmark datasets demonstrate that the proposed trust-aware mechanism maintains high convergence stability and preserves test accuracy under severe adversarial conditions where standard FedAvg experiences catastrophic degradation.