Adaptive AI-Driven Zero-Trust Security for Cloud-Native Enterprise Systems

Abstract

Background: The traditional model of network perimeter, where traffic within organizational borders may be trusted, is structurally incapable of cloud-oriented and distributed enterprise systems. Combined with containerized microservices, multi-cloud implementation, remote work, and interconnectedness of supply chains enabled by the convergence of all these factors has made perimeter-based defenses irrelevant to the current adversary.Goal: The proposed systematic literature review (SLR) synthesizes peer-reviewed articles on the topic of zero-trust security frameworks and AI-based identity and access management (IAM) to identify the state of the art, common themes, the technologies applied, and unresolved research problems.Procedures: We searched IEEE Xplore, Springer Link, and Wiley online library according to the PRISMA 2020. Out of 642 records of the first phase of records, 20 main studies were then selected and subjected to two-reviewer screening and quality evaluation to reach final eligibility conditions.Findings: It identified six themes that included (1) zero-trust architecture design, (2) AI-based identity checks, (3) federated learning to perform distributed threat detection, (4) user and entity behavior analytics, (5) multi-cloud policy enforcement, and (6) IoT/edge adaptions. The access intelligence can computerize accent anomalies by 2334 percentage points over rule baselines which are created using ML. Federated training attains the centralized accuracy within a range of 4% and maintains the data locality. Graph neural networks also identify lateral movement 2.3 days before signature tools.Conclusion: Zero-trust and AI-driven IAM go hand in hand: zero-trust includes the enforcement model, whereas AI includes the intelligence to make the decision on the cloud scale. Three gaps exist, including explainability of AI decisions, adversarial robustness, and lightweight zero-trust threat riskiness of constrained edge devices.