Ololade Zainab Adesokan

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.

Agentic AI has strong potential to boost productivity and growth, but current evidence suggests it is more likely reinforcing existing core–periphery inequalities than fundamentally reshaping the global economic order, due to uneven access, capabilities, and institutional readiness across countries. This study examines the economic value of Agentic Artificial Intelligence (AI), defined as advanced AI systems capable of autonomous decision-making and task execution, by analysing its impact on firm-level productivity, financial performance, and macroeconomic growth across developed and emerging economies. While existing literature has extensively examined general AI adoption, limited empirical evidence exists on how more autonomous, agent-like systems contribute to economic outcomes and whether their benefits are evenly distributed across different economic contexts. To address this gap, the study employs a quantitative panel data approach using data from the World Bank (World Development Indicators and Enterprise Surveys) and OECD AI indicators for the period 2015 to 2024. An AI Adoption Index is constructed using indicators of AI investment, business adoption, and innovation output, serving as a proxy for the diffusion of advanced AI capabilities, including agentic features. Fixed effects regression, mediation analysis, and quantile regression are used to estimate firm-level and macroeconomic relationships. The results show that AI adoption significantly improves firm-level productivity (β = 0.18, p < 0.01) and influences economic growth primarily through a productivity channel (β = 0.35, p < 0.01), with a comparatively weaker direct effect (β = 0.09). However, the magnitude of these effects varies across economic contexts, with developed economies experiencing substantially stronger growth impacts (approximately 0.33) than emerging economies (approximately 0.15). These findings suggest that while Agentic AI enhances economic performance, its benefits are mediated by structural conditions and are unevenly distributed across countries. The study contributes to the literature by providing an integrated micro to macro empirical framework and highlighting the role of advanced AI capabilities in reinforcing global economic disparities. Policy recommendations emphasise the need for investment in digital infrastructure, human capital development, and inclusive technology diffusion strategies to ensure more equitable distribution of AI-driven economic value.

The global financial system is being transformed by the integration of digital technologies such as artificial intelligence, big data, blockchain, and cloud computing, reshaping financial services, risk management, and market dynamics. This study examines how digital transformation is reshaping financial value creation and risk operations within contemporary financial systems. Despite rapid digitalisation, existing financial frameworks remain largely grounded in traditional valuation and risk assumptions, creating a gap in understanding the interaction between digital value, emerging risks, and governance. To address this, the study adopts a multi-method quantitative approach, combining fixed effects panel regression, Difference-in-Differences, Vector Autoregression, and Structural Equation Modelling on a balanced panel of eight countries (four developed and four emerging economies) over the period 2020–2024. The results show that digital transformation significantly enhances financial value, with R&D expenditure exerting a strong positive effect (β = 1.736), while digital adoption reduces non-performing loans (β = −0.967) and operational costs (β = −2.483). However, fintech adoption increases cyber risk (β = 0.341) and contributes to systemic financial stress. Structural modelling further reveals that digital value intensifies risk exposure (β = 0.534), whereas governance mitigates it (β = −0.418). The study advances financial theory by integrating value creation, risk dynamics, and governance into a unified framework. It recommends adaptive regulatory frameworks, strengthened institutional governance, and revised valuation models to ensure stability in digitally transformed financial 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 evolution of Industry 4.0 and emerging Industry 5.0 paradigms has accelerated the adoption of autonomous cyber-physical systems (CPS) in smart manufacturing, creating a need for robust runtime policy orchestration to ensure governance, regulatory compliance, and resilience under dynamic disruptions and cyber threats. This study addresses the lack of unified frameworks by proposing the Policy Resilient Orchestrator (PRO), a simulation-driven architecture integrating machine-readable policy cards, multi-agent reinforcement learning using Proximal Policy Optimization (PPO), multi-objective optimization via NSGA-II, runtime verification, and digital twin-based validation. A quantitative, desk-based methodology utilizing synthetic and benchmark datasets was employed to model CPS environments without physical hardware. NSGA-II generated Pareto-optimal policy configurations, while PPO enabled adaptive runtime control supported by simplex switching for resilience. Monte Carlo simulations across normal, attack, and failure scenarios showed significant improvements, with Governance Efficiency (0.961), Resilience Index (0.993), and Downtime Reduction (0.988), all statistically outperforming baselines (p < 0.0001). PRO demonstrates a scalable, reliable solution for trustworthy smart manufacturing systems.

This study investigates how permission inheritance and machine-speed execution influence cybersecurity risk escalation in agentic artificial intelligence–driven cloud environments. A quantitative empirical design was adopted using a network-based privilege propagation model and survival analysis. Identity and access relationships were represented as a directed authorization graph G=(V,E)G=(V,E)G=(V,E), while privilege expansion was evaluated using the Privilege Amplification Factor and Privilege Propagation Score derived from simulated cloud IAM datasets. Machine-speed escalation dynamics were examined through time-to-event modeling using survival probability and hazard ratio estimation. The findings reveal substantial privilege amplification, with administrative agents demonstrating an average Privilege Amplification Factor of 6.07 and access to 71% of reachable cloud resources. In addition, increasing execution speed from 5 to 200 operations per second reduced mean escalation time from 842 seconds to 68 seconds while raising the hazard ratio to 6.15. Based on these findings, the study proposes the Agentic Cloud Security Governance Model and recommends agent-specific identity isolation, dynamic least-privilege enforcement, real-time behavioral monitoring, and policy frameworks for governing autonomous agents in cloud infrastructures.