Journal of Next-Generation Cybersecurity and Risk Engineering

The breakdown of the network perimeter has made static trust models ineffective, although implementingthe zero-trust model on a large scale is limited by privacy concerns and latency. This paper presents a dynamicframework for zero-trust threat engineering based on federated anomaly detection, risk-based verification, andautomatic containment. The four-layer architecture allows for collaboration between models without sharing thetelemetry data, whereas a gradient response system varies the degree of enforcement based on the aggregate risk score.The quantitative assessment indicates that the proposed framework offers a sub-three-second containment latencyand ensures consistency in detection under non-identical distribution of the data. Comparison with the centralized andsignature-based frameworks shows significant enhancements in lateral movement detection and reaction time withoutincreased computation costs.