Kabir, Md Humayun and Chowdhruy, Sajidul Haque and Akter, Shakila (2024) An AI-Driven Predictive Framework for Zero-Day Threat Mitigation in Cloud-Native Enterprise Workloads. American Journal of Economics and Business Management, 7 (1). ISSN 2576-5973

Text AJEBM_Md+Humayun+Kabir+et+al_An+AI-Driven+Predictive+Framework+for.pdf Download (486kB)

Abstract

Cloud-native architectures improve delivery speed and scalability, but their distributed, ephemeral, and identity-dependent structure also enlarges the number of places where a previously unknown exploit can hide. Signature-based controls remain necessary, yet they are structurally limited when the vulnerable component, payload, or attacker pathway has not been observed before. This article proposes PREDICT-ZD, an AI-driven predictive defense framework for critical enterprise workloads running across containerized, microservices, and service-mesh environments. The framework combines control-plane audit events, service-to-service flows, runtime behavior, software-supply-chain provenance, threat-intelligence signals, and workload criticality in a temporal event graph. A calibrated ensemble of anomaly detection, temporal forecasting, and graph-context models produces an uncertainty-aware risk score that is translated into bounded zero-trust actions, including identity re-verification, egress restriction, micro-segmentation, quarantine, and analyst escalation. The article synthesizes evidence from peer-reviewed intrusion-detection, exploit-prediction, container-security, and trustworthy-machine-learning research and maps the design to relevant NIST guidance. Because no primary enterprise dataset was supplied, the Results section uses an explicitly labeled synthetic evaluation to demonstrate the analysis protocol; its values are not empirical claims. The contribution is therefore a testable research architecture and evaluation plan rather than a claim of completed field validation. The paper closes with limitations, governance requirements, and a research agenda for cross-cloud, privacy-preserving, adversarially robust deployment.

Item Type:	Article
Uncontrolled Keywords:	cloud-native security, zero-day threats, predictive analytics, anomaly detection, zero trust, critical workloads
Subjects:	H Social Sciences
Depositing User:	admin eprints
Date Deposited:	04 Aug 2026 04:14
Last Modified:	04 Aug 2026 04:14
URI:	http://eprints.umsida.ac.id/id/eprint/16902

Actions (login required)

View Item