Integrating Artificial Intelligence into Zero Trust Security Models: Frameworks, Benefits, and Research Challenges

Research Article

Authors

  • Micheal Abimbola Oladosu Department of Chemical Sciences, Faculty of Science, Anchor University Lagos, Ayobo, Lagos, Nigeria Author
  • Opeyemi Christiana Ajayi Department of Electrical and Electronic Engineering, Faculty of Engineering, Osun State University, Osogbo, Osun State, Nigeria Author
  • Caleb Onwuadiamu Department of Computer Science, Faculty of Physical Science, University of Cross River State, Calabar, Cross River State, Nigeria Author
  • Chidozie Agbo P&C Division, NNPC E&P Limited, Nigeria Author
  • Somtochukwu Chukwudi Orefo Department of Modern State of Artificial Intelligence, Faculty of Applied Mathematics and Informatics, Moscow Institute of Physics and Technology, Moscow, Russia Author
  • Moses Adondua Abah Department of Biochemistry, Faculty of Pure and Applied Sciences, Federal University of Wukari, Wukari, Taraba State, Nigeria Author

DOI:

https://doi.org/10.70829/aq3e2j11

Keywords:

Zero Trust Architecture, Artificial Intelligence, Machine Learning, Cybersecurity, Anomaly Detection, Federated Learning, Adaptive Access Control, Explainable AI

Abstract

Background: The convergence of Artificial Intelligence (AI) and Zero Trust Architecture (ZTA) represents one of the most consequential paradigm shifts in contemporary cybersecurity. Traditional perimeter-based defences have proven inadequate against sophisticated, persistent threats, prompting the adoption of Zero Trust principles (‘never trust, always verify ‘) as a foundational security posture.
Objective: This review critically examines the integration of AI/ML techniques into ZTA frameworks, systematically evaluating published frameworks, quantifiable benefits, and unresolved research challenges.
Methods: A structured literature review was conducted across PubMed, IEEE Xplore, ACM Digital Library, ScienceDirect, and arXiv (2020–2025), yielding 31 peer-reviewed sources after PRISMA-guided screening.
Results: AI-augmented ZTA demonstrably improves threat detection accuracy (up to 97.3%), reduces incident containment time (by 74%), and enables adaptive, context-aware access control. Six primary AI/ML techniques- deep learning, federated learning, graph neural networks, reinforcement learning, natural language processing, and ensemble methods- show particular promise. However, critical challenges persist, including adversarial machine learning vulnerabilities, explainability deficits, privacy-utility trade-offs, and model drift.
Conclusion: AI-integrated ZTA is not merely an incremental enhancement but a necessary evolution toward autonomous, self-adaptive security ecosystems. A forward-looking research agenda, encompassing quantum-resistant models, LLM-driven policy synthesis, and causal AI, is proposed to accelerate field maturity.

Downloads

Published

2026-09-29

Issue

Vol. 3 No. 01 (2026): IJRMSE

Section

Articles