Metraipolis › archive › dc_ff925c202dfd

An Open-Agent Framework for Intelligent Compliance Analysis in Infrastructure as Code Using Large Language Models (2026) · DOI 10.21203/rs.3.rs-10907487/v1

Source: Metraipolis archive
Original: https://doi.org/10.21203/rs.3.rs-10907487/v1
License (as archived): CC-BY-4.0
Archived: 2026-10-20T10:29:07.470Z
SHA-256 of the archived text: 36b0e1996bbbbfd469bcfb1a271bbb47d8b6015da84868dd647dedb6aedd2791
Stable URL of this page: https://metraipolis.com/documento/dc_ff925c202dfd

How to cite

An Open-Agent Framework for Intelligent Compliance Analysis in Infrastructure as Code Using Large Language Models (2026) · DOI 10.21203/rs.3.rs-10907487/v1 — Metraipolis archive, https://doi.org/10.21203/rs.3.rs-10907487/v1 (CC-BY-4.0); archived copy: https://metraipolis.com/documento/dc_ff925c202dfd (Metraipolis archive dc_ff925c202dfd, sha256 36b0e1996bbbbfd469bcfb1a271bbb47d8b6015da84868dd647dedb6aedd2791)

BibTeX

@misc{dcff925c202dfd,
  title = {An Open-Agent Framework for Intelligent Compliance Analysis in Infrastructure as Code Using Large Language Models (2026) · DOI 10.21203/rs.3.rs-10907487/v1},
  howpublished = {Metraipolis archive},
  url = {https://doi.org/10.21203/rs.3.rs-10907487/v1},
  note = {License CC-BY-4.0; archived copy https://metraipolis.com/documento/dc_ff925c202dfd; sha256 36b0e1996bbbbfd469bcfb1a271bbb47d8b6015da84868dd647dedb6aedd2791},
  year = {2026}
}

Archived text

Reproduced under CC-BY-4.0. Attribution: Metraipolis archive, https://doi.org/10.21203/rs.3.rs-10907487/v1. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Articolo recente con licenza aperta: scheda dai metadati Crossref (posted-content).
Titolo: An Open-Agent Framework for Intelligent Compliance Analysis in Infrastructure as Code Using Large Language Models
Autori: Ayah Al-Jabali, Mu'awya Al-Dala'ien
Pubblicato: 2026-10-19 · editore: Springer Science and Business Media LLC
DOI: https://doi.org/10.21203/rs.3.rs-10907487/v1
Licenza dell'articolo: CC-BY-4.0

Abstract (testo degli autori, CC-BY-4.0):
Abstract Cloud misconfiguration is a leading cause of data breaches in production cloud environments, and Infrastructure as Code (IaC) has amplified this risk by letting a single insecure configuration propagate across an entire cloud estate in one deployment. This paper presents an open-agent framework for intelligent compliance analysis of IaC configurations, integrating Open Policy Agent (OPA)-based Policy-as-Code enforcement with Large Language Model (LLM)-driven contextual reasoning in a single hybrid pipeline. Prior work has treated deterministic policy enforcement and LLM-based analysis independently; the proposed framework instead combines the precision of OPA-based Rego rule evaluation with the explanatory and remediation capabilities of an LLM to produce structured, audit-ready compliance evidence aligned to a formal control framework. The framework is instantiated with a custom 16-rule Rego policy library spanning three NIST SP~800-53 control families Access Control (AC), System and Communications Protection (SC), and Audit and Accountability (AU) across eight AWS resource types, and evaluated on 231 labeled Terraform configurations derived from the IaC-Eval benchmark. The OPA enforcement layer achieves 96.9\% precision, 90.5\% recall, and 93.6\% F1 score. The LLM enrichment layer, guided by a deterministic safeguard against LLM recall errors and explicit prompt-level format constraints, produces reliable, audit-ready control mappings and remediation guidance across 16 violation types. The performance advantage over an industrial static-analysis baseline (Checkov) is confirmed statistically significant via McNemar's exact test ($\chi^2=13.0$, $p<0.001$), demonstrating that the hybrid approach delivers compliance analysis outcomes that neither paradigm achieves alone.

Fonte: https://api.crossref.org/works/10.21203%2Frs.3.rs-10907487%2Fv1 — metadati Crossref, consultati il 2026-10-20. Interrogazione: «multi-agent systems large language models».