Neutrosophic Multi-Criteria Decision Framework for CI/CD Selection in Public Security Institutions: A Chain of Experts Approach

Abstract

Public security institutions face critical challenges when adopting Continuous Integration and Continuous Deployment (CI/CD) automation frameworks due to conflicting requirements: accelerating software delivery while maintaining strict security compliance, operational resilience under resource constraints, and regulatory adherence. This study addresses the real institutional decision problem of selecting an optimal CI/CD framework for a public security organization operating under national sovereignty restrictions, legacy system constraints, and multi-stakeholder governance. The decision environment is characterized by incomplete information, contradictory expert opinions, and indeterminate risk assessments—conditions where classical multi-criteria decision-making (MCDM) approaches fail to capture epistemic uncertainty. We propose a neutrosophic AHP-TOPSIS framework that explicitly models truth, indeterminacy, and falsity components of expert judgments. The decision process was supported by a Neutrosophic Chain of Experts implemented through Large Language Models, where specialized roles (domain contextualization, neutrosophic modeling, consistency validation, aggregation synthesis, and academic documentation) systematically refined criteria weights, alternative evaluations, and consensus mechanisms. Four realistic alternatives were evaluated: on-premises CI/CD, hybrid CI/CD, secured cloud-based CI/CD, and minimal automation baseline, against five criteria reflecting institutional priorities: implementation cost, deployment speed improvement, security compliance, operational resilience, and maintainability. Results demonstrate that the Chain of Experts methodology reduced ranking inconsistencies by 34% compared to single-expert neutrosophic evaluation and improved decision robustness across sensitivity scenarios involving weight perturbations (±20%) and indeterminacy level variations (I ∈ [0.1, 0.5]). The hybrid CI/CD model emerged as the optimal choice with a final closeness coefficient of 0.742, balancing security requirements with modernization objectives. Comparative analysis against classical weighted-sum MCDM and fuzzy TOPSIS revealed that neutrosophic modeling captured 23% more uncertainty variance in security compliance assessments and 18% in resilience evaluations. The Chain of Experts architecture demonstrated measurable improvements in decision quality through systematic contradiction detection, consensus-driven refinement, and methodological tr

Keywords