Cheon-Ho Min

Artificial intelligence (AI) systems have become operational infrastructure whose value is increasingly dominated by trained models, behavioral configurations, and decision-making logic rather than by software binaries alone. As a result, ransomware threats against AI systems cannot be adequately addressed by conventional recovery strategies that assume service availability can be restored through file and code recovery. In AI environments, assets such as model parameters, training data, inference pipelines, and safety policies constitute primary attack targets, and their compromise can invalidate system behavior even when files are successfully restored. This study re-examines ransomware threats against AI systems from an asset-based protection perspective and demonstrates why traditional recovery assumptions structurally fail in AI-centric environments. Based on this analysis, we show that protection mechanisms limited to file integrity are insufficient and must be extended to include behavioral consistency and decision-making reliability. To address this gap, we propose a behavior-aware ransomware protection methodology, implemented as the Behavior-Aware Integrity Protection System (BIPS). BIPS augments existing ransomware response processes by redefining protection targets, establishing behavioral baselines, verifying post-recovery behavioral integrity, and supporting risk-based operational decisions. This work contributes by reframing ransomware threats against AI systems as an issue rooted in protection scope and recovery assumptions rather than isolated attack techniques, thereby extending ransomware response for AI systems toward a reliability- and risk-oriented protection framework.

This study addresses the challenge of evaluating artificial intelligence (AI) systems across heterogeneous regulatory frameworks. Although the NIST AI RMF, EU AI Act, and ISO/IEC 23894/42001 define important governance requirements, they do not provide a unified quantitative method. To bridge this gap, we propose the Cross-Assessment & Verification for Evaluation (CAVe) Framework, which maps shared regulatory requirements to four measurable indicators—accuracy, robustness, privacy, and fairness— and aggregates them into a Cross-Compliance Index (CCI) using normalization, thresholding, evidence penalties, and cross-framework weighting. Two validation scenarios demonstrate the applicability of the approach. The first scenario evaluates a Naïve Bayes-based spam classifier trained on the public UCI SMS Spam Collection dataset, representing a low-risk text-classification setting. The model achieved accuracy 0.9850, robustness 0.9945, fairness 0.9908, and privacy 0.9922, resulting in a CCI of 0.9741 (Pass). The second scenario examines a high-risk healthcare AI system using a CheXNet-style convolutional model evaluated on the MIMIC-CXR dataset. Diagnostic accuracy, distribution-shift robustness, group fairness (finding-specific group comparison), and privacy risk (membership-inference susceptibility) yielded 0.7680, 0.7974, 0.9070, and 0.7500 respectively. Under healthcare-oriented weighting and safety thresholds, the CCI was 0.5046 (Fail). These results show how identical evaluation principles produce different compliance outcomes depending on domain risk and regulatory priorities. Overall, CAVe provides a transparent, reproducible mechanism for aligning technical performance with regulatory expectations across diverse domains. Additional metric definitions and parameter settings are provided in the manuscript to support reproducibility, and future extensions will incorporate higher-level indicators such as transparency and human oversight.

Recent data center incidents have revealed that certification under ISO 22301 and ISO/IEC 27001 does not guarantee real operational resilience. This study presents the Availability Assurance Framework (RMF-A), an extension of the NIST Risk Management Framework that introduces an Availability Assurance Phase. RMF-A combines ISO-based management controls with NIST’s evidence-driven assessment using the Availability Evidence Model (AEM) and the Availability Assurance Index (AAI). AEM defines measurable indicators—recovery rate (RR), recovery time (RTO), and Detection Effectiveness (DET)—and AAI aggregates them into a quantitative assurance score. Validation using three open datasets—Google Cluster Trace, Azure Cloud Trace, and LANL HPC Logs—showed consistent assurance results: Google (AAI = 0.758, ATO-Conditional), Azure (AAI = 0.720, ATO-Conditional), and LANL HPC (AAI = 0.744, ATO-Conditional). The results confirm that RMF-A provides a reproducible, evidence-based approach to quantify operational resilience and ensure continuous availability.

As smart manufacturing environments continue to evolve, operational technology systems are increasingly integrated with external networks and cloud-based platforms. However, many manufacturing facilities still use legacy systems running on end-of-support/life operating systems with discontinued security updates. It is difficult to mitigate the cyber threats and risks for these systems using perimeter-based security models that isolate them from other networks. To address these constraints, a Zero Trust-based security architecture tailored for legacy manufacturing environments with practical field applicability is proposed. Our architecture builds upon the six core functions outlined in National Institute of Standards and Technology Cybersecurity Framework 2.0—identify, protect, detect, respond, recover, and govern—adapting them specifically to manufacturing environment security challenges. To achieve this, the architecture combines asset identification, policy-driven access control, secure SMB gateway transfers, automated anomaly detection and response, clean image recovery, and organizational governance procedures. This study validates the effectiveness and scalability of the proposed architecture through scenario-based simulations. When combining the EoSL defense hardening and gateway-based perimeter control, the architecture achieves approximately 99% overall threat suppression and a 98% reduction in critical-asset infection rates, demonstrating its strong resilience and scalability in large-scale legacy OT environments.