Federated Quantum Kernel Learning for Decentralized Anomaly Detection in Zero Trust Edge Networks

The proliferation of Industrial IoT (IIoT) and 5G/6G edge networks has expanded the attack surface for critical infrastructure. Zero Trust Architecture (ZTA) addresses this through continuous verification, but implementation at the edge faces some challenges: the sensitivity-privacy paradox requiring local data while detecting subtle anomalies; computational asymmetry between resource-constrained devices and complex threat detection; and trust calibration gaps lacking epistemic uncertainty quantification. In this paper, we propose Federated Quantum Kernel Learning (FQKL), a framework for decentralized, privacy-preserving anomaly detection in Zero Trust edge networks to address these challenges to improve detection performance. FQKL leverages Parameterized Quantum Circuits (PQCs) to encode data into high-dimensional Hilbert spaces, capturing non-linear correlations that classical kernels cannot efficiently represent. Edge nodes compute local quantum kernels and transmit only top-m support vectors with differential privacy noise. We introduce uncertainty-aware trust scoring and cross-segment mutual information analysis to detect lateral movement attacks. Experiments demonstrate that increasing PQC depth significantly improves anomaly detection with minimal communication overhead.