Skip to content
Conference Open access

Explainable Agentic LLM Framework for Autonomous Cloud Security, Compliance Monitoring, and Self-Healing Operations

Oct 2026 · 2026 IEEE Conference on Generative AI for Secure Systems (GAISS) · pp. 1-5 · 0 citations · 16 references

Abstract

Today's multi-cloud infrastructures produce security telemetry at a rate and volume that exceeds the ability of a rule-based Security Information and Event Management (SIEM), a static compliance engine, and traditional AIOps platforms. These tools detect known patterns but cannot reason about novel attack chains, justify their decisions, or autonomously restore a degraded service, while emerging single-agent Large Language Model (LLM) assistants improve reasoning yet remain opaque, non-auditable, and structurally unable to coordinate detection, compliance, and remediation as one closed loop. This paper proposes X-ALSO (Explainable Agentic LLM for Security and Operations), a six-agent framework in which specialized LLM agents for security monitoring, compliance assessment, root-cause analysis, remediation planning, self-healing execution, and explainability collaborate over a structured message bus, grounded by a retrieval-augmented knowledge base of policies and benchmarks, and supervised by a human-in-the-loop governance layer. Every autonomous decision emits a traceable, human-readable explanation recorded in an immutable audit trail. We formalize four operational metrics-Security Risk Score, Compliance Confidence Score, Agent Reliability Score, and Healing Effectiveness Index-and a constrained optimization objective. Evaluated on AWS, Azure, and Kubernetes workloads using CloudTrail, CloudWatch, and audit logs, X-ALSO attains 96.3% detection accuracy, raises CIS compliance accuracy to 97.1%, reduces mean time to recovery by 26% over a single-agent LLM baseline, and achieves a 93.6 composite explainability score, substantially outperforming rule-based, SIEM, and single-agent baselines.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.