Cornelia Ifeoma Ejoh

Critical infrastructure operators increasingly deploy artificial intelligence (AI) to detect degradation, predict failures, optimise maintenance, identify cyber anomalies and support operational decisions. These capabilities create a plausible route to higher availability, reliability and quality, yet they also couple asset-performance decisions to data integrity, model uncertainty, software supply chains, human oversight and operational technology security. This critical narrative review examines how secure AI can contribute to asset performance without weakening safety or resilience. Literature spanning predictive maintenance and prognostics, industrial AI, digital twins, industrial cyber-physical security, adversarial machine learning, explainability, uncertainty, machine-learning operations and human factors was critically synthesised. The evidence is strongest for AI as a decision-support layer for condition monitoring, fault diagnosis and targeted maintenance where data provenance is controlled, failure modes are sufficiently represented and recommendations remain bounded by engineering constraints. Evidence for fully autonomous optimisation in high-consequence infrastructure is less mature because benchmark accuracy does not directly establish operational utility, transferability or safe behaviour under distribution shift and hostile manipulation. Digital twins can improve contextual diagnosis and testing, but their value depends on fidelity, synchronisation, governance and protection of the data-model-actuation pathway. Cybersecurity and safety therefore cannot be treated as independent assurance domains: poisoned training data, adversarial inputs, compromised updates or unavailable models can become physical reliability and quality risks. Explainability alone is also insufficient for assurance; uncertainty calibration, abstention, independent validation, auditability and meaningful human authority are required. The synthesis proposes a secure asset-performance loop in which AI recommendations are evaluated against safety, security, service and quality constraints, with outcomes fed back into monitored model and process improvement. The central implication is that asset performance should be maximised as a constrained socio-technical objective rather than as an unconstrained prediction or utilisation metric.

Artificial intelligence systems increasingly mediate consequential decisions in credit allocation, healthcare triage, employment screening and public administration, yet the data underpinning these decisions is frequently incomplete, mislabelled, stale or quietly altered as it moves through long and opaque pipelines. This review examines governance frameworks intended to preserve the integrity of algorithmic decision data and to align organisational practice with an increasingly dense regulatory landscape spanning the European Union, the United States and international standard-setting bodies. It synthesises literature on data quality theory, documentation artefacts such as datasheets and model cards, blockchain-based provenance mechanisms, algorithmic auditing regimes and sector-specific compliance obligations in finance and healthcare. The review finds that technical solutions for data quality monitoring have matured considerably faster than the institutional arrangements needed to make such monitoring auditable, contestable and legally enforceable, producing a persistent gap between what is technically feasible and what is organisationally practised. It further finds that regulatory instruments, notably the General Data Protection Regulation and the Artificial Intelligence Act, converge on transparency and documentation obligations but diverge on enforcement mechanics, creating compliance friction for organisations operating across jurisdictions. The review proposes a layered governance model integrating data-level controls, documentation practices, human oversight and independent auditing, and identifies future research priorities around interoperable provenance standards, cross-border regulatory harmonisation and the measurement of data integrity as a continuous rather than a point-in-time property.

This study evaluated the adversarial robustness of machine learning-based fraud detection systems by comparing classifier vulnerability profiles and assessing adversarial training as a mitigation strategy. Using the IEEE-CIS Fraud Detection dataset, comprising 590,540 transactions with a fraud incidence of 3.5%, four classifiers—logistic regression, random forest, gradient boosting, and a feed-forward neural network—were trained under identical preprocessing and class-weighting conditions and then subjected to Fast Gradient Sign Method and Projected Gradient Descent attacks at a perturbation budget of 0.02. Adversarial examples were constructed directly using closed-form and backpropagated gradients for the differentiable classifiers and using a logistic regression surrogate for the non-differentiable ensembles, before adversarial training was applied as a post-attack mitigation stage. Logistic regression proved the most adversarially vulnerable architecture, sustaining a 31.12-percentage-point recall loss under Projected Gradient Descent, while adversarial training subsequently restored its recall from 0.39 to 0.999 at an accuracy cost of 0.10 percentage points. Random forest and gradient boosting were not degraded by the surrogate-based attack, indicating that comparative robustness claims for tree-based ensembles require attack methods suited to their non-differentiable structure rather than transfer-based evaluation alone. Within the scope of this single-dataset evaluation, the findings support the adoption of adversarial training for gradient-based fraud detection models and suggest that robustness claims should be accompanied by disclosure of the attack methodology used to establish them.

This study examines the ethical and secure deployment of generative artificial intelligence in enterprise environments, with emphasis on innovation, data privacy, and risk governance. It addresses the gap between the rapid organisational adoption of generative AI systems and the slower development of institutional mechanisms for managing ethical, privacy, security, and regulatory risks. The study identifies the absence of a validated, integrated governance instrument that combines ethical, privacy, security, and enterprise risk controls for the specific characteristics of generative systems. A desk-based mixed-methods approach was used, combining systematic literature review, document analysis, thematic synthesis, comparative evaluation of governance frameworks, and Analytic Hierarchy Process weighting. Evidence was drawn from peer-reviewed literature and authoritative international governance instruments. Eight governance dimensions were assessed: scope and coverage, risk classification, data privacy, ethics and accountability, security controls, legal enforceability, enterprise applicability, and adaptability to generative AI. The findings show that existing governance instruments provide useful but fragmented coverage when applied independently. Ethics and accountability emerged as the highest-weighted dimension, followed by data privacy, security controls, and enterprise applicability. The proposed framework integrates five pillars: ethics and accountability, privacy and data governance, security governance, enterprise risk management, and innovation and compliance alignment. The study concludes that responsible enterprise deployment of generative AI requires coordinated, multi-layered governance rather than reliance on isolated ethical, technical, or legal controls.

Background: The COVID-19 pandemic accelerated the digital transformation of mental health and rehabilitation counseling services in the United States. This shift has expanded access to care, but it has also exposed persistent inequities that affect underserved populations. Objective: This scoping review maps policy pathways for equitable and culturally responsive mental-health support within digital rehabilitation counseling, and quantifies the evidence base supporting reform. Methods: Using the Arksey and O'Malley framework and the PRISMA Extension for Scoping Reviews (PRISMA-ScR), we identified, screened, and synthesized 88 sources published between 2020 and 2026 from PubMed, PsycINFO, Scopus, Google Scholar, and grey-literature repositories (WHO, FCC, HRSA, OECD, UNDP, NRHA). Descriptive statistics summarized source characteristics, a thematic coding matrix mapped eight policy-relevant themes, and a Pearson chi-square test examined associations between source type and thematic focus. Results: Most sources (84.1%, n = 74) were peer-reviewed journal articles, and 10.2% (n = 9) were institutional or governmental reports; the remainder were preprints, textbooks, or news commentaries. Ninety-six percent of dated sources appeared in 2022 or later, with a median publication year of 2025. The most frequently addressed themes were tele-rehabilitation and telehealth (27.3%), AI and digital mental health (20.5%), disability and rehabilitation (15.9%), and digital equity (14.8%). Source type and dominant theme were significantly associated (χ²(8) = 18.42, p = .018): institutional reports clustered around digital-equity and workforce issues, while peer-reviewed studies emphasized clinical and AI-related questions. Conclusions: The evidence base for digital rehabilitation counseling is growing rapidly but remains unevenly distributed across equity-critical themes. Sustainable reform calls for coordinated investment in broadband infrastructure, multilingual and disability-inclusive platform design, culturally responsive workforce development, ethical AI governance, and inclusive workforce reintegration, anchored in a “human-in-the-loop” model of care.

This study presents a hybrid federated learning and explainable artificial intelligence framework, termed RobustFL, designed to mitigate adversarial attacks in medical imaging while incorporating privacy-preserving mechanisms. Focusing on chest X-ray analysis, the research systematically examined adversarial vulnerabilities and data privacy challenges in distributed healthcare systems. A simulation-based methodology was implemented using a 500-sample dataset partitioned across multiple clients in a non-independent and identically distributed setting. The framework extends federated averaging by integrating projected gradient descent-based adversarial training, differential privacy stochastic gradient descent, and SHAP-based interpretability. Experimental results demonstrate that RobustFL maintains stable performance under adversarial conditions, achieving 70% accuracy against FGSM and PGD attacks, with an adversarial success rate of approximately 30–31%. A balanced privacy budget of ε = 1.0 produced an F1-score of 0.412, indicating a trade-off between privacy and model utility. Interpretability consistency, measured via SHAP outputs, served as a proxy for trust assessment, while privacy risk indicators were used to infer potential user concern. The framework provides a practical pathway toward secure, transparent, and privacy-aware healthcare AI systems.

The rapid growth of multimodal wearable devices has enabled continuous monitoring of physiological and behavioral patterns for home-based health applications. However, centralized data processing raises serious privacy concerns and limits real-time, interpretable insights. This study proposes PEX-FEL, a privacy-preserving and explainable federated edge learning framework for stress detection and activity recognition in decentralized environments. The framework combines federated learning with differential privacy (ε ≤ 1.0) and secure aggregation to protect user data. Low-Rank Adaptation (LoRA) is applied for efficient local training on resource-constrained edge devices, while SHAP is used to provide interpretable, user-centric explanations of predictions. Experiments were conducted using WESAD, PPG-DaLiA, and SWELL datasets under non-IID conditions to simulate real-world heterogeneity. Results show that a hybrid CNN-LSTM model achieved 0.85 accuracy, 0.85 F1-score, and 0.90 AUC-ROC, outperforming centralized approaches by 8–10%. The framework also maintained strong privacy (membership inference < 0.52) and low latency (~45 ms). SHAP analysis identified heart rate variability and electrodermal activity as key stress indicators. Overall, this work demonstrates a balanced approach to accuracy, privacy, efficiency, and interpretability in wearable health monitoring systems.

The rapid adoption of artificial intelligence (AI) in healthcare analytics has raised significant concerns regarding patient data privacy, security, and regulatory compliance. This study develops a Privacy-by-Design (PbD) data governance model tailored for AI-driven healthcare analytics in remote research environments. The proposed framework integrates privacy-preserving technologies to enable secure data analysis while preventing the centralization of sensitive health records. The novelty of this research lies in the design of a unified governance architecture that combines federated learning, differential privacy, and blockchain-enabled consent management to support transparent, accountable, and regulation-compliant healthcare AI systems. A simulation-based research approach was adopted using publicly available and synthetically generated healthcare datasets distributed across multiple virtual clients to emulate decentralized clinical data environments. The framework was evaluated based on model performance, privacy guarantees, fairness, and system convergence under heterogeneous data conditions. Experimental results demonstrate that the proposed PbD–federated architecture achieves reliable predictive performance while maintaining strong privacy protection and auditability without direct data sharing. Overall, the findings confirm that integrating privacy-preserving learning with governance mechanisms can enable responsible and scalable healthcare analytics in remote research contexts. The study provides a practical blueprint for developing trustworthy AI systems that align with modern data protection principles and ethical healthcare data management.

This study proposes and quantitatively evaluates a secure-by-containment architectural framework for deploying agentic artificial intelligence (AI) within operational technology and industrial control system (OT/ICS) environments. Agentic AI systems introduce autonomous planning and decision-making capabilities into cyber-physical infrastructures, creating new risks when deployed in safety-critical control environments that require deterministic and predictable behavior. To address this challenge, the study introduces containment as a first-class architectural security property that structurally limits an AI agent’s decision authority, operational scope, and propagation influence within industrial control architectures. The framework is evaluated using publicly available industrial control system datasets, including the ICS Gas Pipeline dataset, the Oak Ridge PowerCyber testbed dataset, the NIST Special Publication 800-82 control catalog, and the Tennessee Eastman Process dataset. A four-phase analytical methodology is implemented consisting of: cyber-physical risk quantification using the Process Impact Deviation Index (PIDI), Risk Severity Score (RSS), and Multi-Variable Escalation Index (MVEI); graph-theoretic containment modeling using betweenness centrality and the Containment Strength Score (CSS); standards alignment analysis through the Standards Coverage Quantification Index (SCQI); and scenario-based resilience evaluation using Mean Time to Hazard (MTH), Process Deviation Area (PDA), and Safety Threshold Violations (STV). The results demonstrate that secure-by-containment architectures substantially reduce systemic exposure in OT/ICS environments. Full containment reduced structural influence by 75.1% (CSS = 0.751), achieved 84.1% coverage of applicable NIST SP 800-82 security controls, increased mean time to hazard by 212%, and reduced cumulative process deviation by 73% during simulated attack scenarios. These findings demonstrate that architectural containment can materially improve cyber-physical resilience while maintaining compatibility with established industrial security standards. The study contributes to the emerging field of AI governance for cyber-physical systems by translating high-level governance principles into enforceable architectural constraints. The results support embedding decision-authority containment mechanisms into OT/ICS security architectures and recommend integrating containment requirements into industrial cybersecurity standards, regulatory validation frameworks, and responsible AI governance models for critical infrastructure environments.

Healthcare systems globally face persistent inefficiencies in clinical and administrative workflows due to complex, variable processes and limited visibility into real executions. This study developed an AI-enabled process mining framework designed for the automated discovery and improvement of these workflows through remote analysis of publicly accessible event logs. Using the Sepsis Cases - Event Log.xes as the primary dataset, the study applied AI-assisted discovery techniques including Heuristics Miner, Fuzzy Miner, and trace clustering to model clinical pathways and administrative activities. Conformance checking via optimal alignments assessed adherence to expected guidelines, highlighting deviations and bottlenecks. Predictive optimization was achieved using a Long Short-Term Memory (LSTM) neural network for next-activity forecasting and bottleneck detection, producing simulated throughput time reductions of approximately 18%. Performance metrics indicated moderate but informative conformance results: fitness of 0.609, precision of 0.443, composite F1-score of 0.64, and an AUC-ROC of 0.931 for predictive monitoring. The framework effectively enabled remote, reproducible analysis without direct data access, providing a practical blueprint for data-driven workflow improvement in resource-constrained environments. Despite limitations involving dataset specificity and noise sensitivity, the approach advances process mining applications in healthcare by integrating artificial intelligence for actionable insights. Future work should emphasize multi-dataset validation, explainability, and real-time deployment to further strengthen scalability and clinical decision support impact.