SECURITY-AS-CODE APPROACH FOR AUTOMATING PCI DSS COMPLIANCE USING AUTONOMOUS AGENTS BASED ON LARGE LANGUAGE MODELS

Read full text Summarise with VIRAT
Authors: Олександр Вахула, Dmytro Marchuk
Cited by: 0
References: 8

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper investigates the Security-as-Code approach for automating PCI DSS compliance in cloud-native environments using autonomous agents based on large language models. Based on an analysis of PCI DSS v4.0.1, which defines twelve principal requirements and emphasizes continuous assessment, evidence-based validation of controls, and proper scoping, a reference architecture is proposed that integrates declarative security policies, preventive configuration enforcement mechanisms, and runtime behavioral monitoring. The study formalizes a mechanism for transforming regulatory requirements into machine-verifiable policies through the use of control specifications (ControlSpec), a structured knowledge base, and a Retrieval-Augmented Generation approach, ensuring policy provenance and reducing the risk of incorrect generation. A model of autonomous agents is proposed to perform functions including requirements-to-policy transformation, configuration drift analysis, evidence collection and normalization, and the generation of remediation recommendations using controlled change mechanisms. Metrics for evaluating continuous compliance are defined, including control coverage, drift detection latency, Mean Time to Detect (MTTD), Mean Time to Respond (MTTR), completeness of evidence, and detection accuracy. An experimental validation plan is proposed using a test environment that simulates a Cardholder Data Environment. Particular attention is given to the analysis of risks associated with autonomous agents, including model hallucinations, prompt injection attacks, sensitive data leakage, and excessive tool privileges. Mitigation measures are defined, including deterministic policy enforcement, tool isolation, agent action logging, and the use of a human-in-the-loop approach for critical operations. The proposed approach enables continuous, verifiable, and scalable PCI DSS compliance in cloud-native environments.

Fields of Study & Subjects

Information and Cyber Security Cloud Data Security Solutions Security and Verification in Computing

Publication details

Journal / source
Cybersecurity Education Science Technique
Pages
154–154
Publication date
26 March 2026
Language
EN