Authors: Dr Devi Mani, Dr. Saley Seetharaman
Abstract: The breakdown of the network perimeter has made static trust models ineffective, although implementing the zero-trust model on a large scale is limited by privacy concerns and latency. This paper presents a dynamic framework for zero-trust threat engineering based on federated anomaly detection, risk-based verification, and automatic containment. The four-layer architecture allows for collaboration between models without sharing the telemetry 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 latency and ensures consistency in detection under non-identical distribution of the data. Comparison with the centralized and signature-based frameworks shows significant enhancements in lateral movement detection and reaction time without increased computation costs.
DOI: http://doi.org/
