Volumes2 & Issues1

The COVID-19 pandemic catalyzed an unprecedented expansion of research in healthcare artificial intelligence (AI) and machine learning (ML), accelerating advances in model architectures, representation learning, and multimodal systems. Despite this rapid methodological progress, the post-pandemic translation of AI/ML research into sustained, large-scale clinical deployment has remained limited. This review critically examines the widening gap between algorithmic innovation and the real-world implementation of healthcare AI/ML from 2019 to 2025. Across several medical domains, we identify recurring failure modes that undermine deployment readiness, including an overreliance on retrospective benchmarks, inadequate external and prospective validation, poor calibration, and vulnerability to dataset shift and distributional drift. We further analyze how emerging paradigms, such as continuously learning and foundation-scale models, challenge existing assumptions around model governance, traceability, and regulatory oversight. By contrasting domain-specific adoption patterns and examining case studies of both successful and stagnant deployments, we show that clinical impact depends less on algorithmic sophistication than on alignment with data stability, workflow integration, and regulatory feasibility. We argue that closing the research–deployment gap requires a shift toward robustness, calibration, and lifecycle-centric AI/ML design, supported by evaluation frameworks that reflect real clinical constraints rather than idealized experimental conditions.

Hamdaoui, R. (2026). Aligning ICT Ambitions with Reality: From Research Acceleration to Deployment Stagnation: A Critical Review of Post-COVID AI and Machine Learning in Healthcare, Journal of Shaqra University for Computing and Information Technology, 2(1), 1-13.

Q-Commerce is an example of the changing landscape of electronic commerce with the ultra-rapid on demand delivery of consumer products. Fulfilling the needs of the consumer for speed and convenience, it raises critical issues of customer trust and value on the data to be used for the right purpose. The purpose of this study is to examine how users perceive blockchain-enabled data transparency and its potential ability to build trust and sustainable value in the Q-Commerce operating environment. This study is based on socio-technical systems theory in combination with the Value-Based Adoption Model and develops a conceptual framework focusing on user perception in Q-Commerce platforms, where transparency acts as a driver of trust and sustainable value. This study used survey data collected from 160 Q-Commerce users in Saudi Arabia. Structural Equation Modeling (SEM) was employed to test the proposed model. The findings indicate that trust and perceived sustainable value are positively influenced by perceived blockchain-enabled transparency. Overall, trust, sustainable value, and transparency positively influence continued use. This study positions Q-Commerce transparency in a novel and unique manner from a theoretical perspective and provides a framework for practitioners and policymakers for developing trust in new digital-commerce systems.

Alaklabi, S. (2026). Enhancing Trust and Sustainable Value in Quick Commerce (Q-Commerce) via Blockchain-Enabled Data Transparency, Journal of Shaqra University for Computing and Information Technology, 2(1), 14-32.

Current reactive blocklisting mechanisms remain inadequate for detecting zero-day phishing URLs due to the rapid evolution of malicious patterns. Although machine learning models provide predictive capabilities, state-of-the-art models are associated with high computational costs, which restrain the applicability of these approaches in real-time and resource-constrained environments. In the current study, the trade-off between accuracy and computational costs of lightweight linear classifiers in phishing URL detection has been evaluated. The study uses a range of machine learning classifiers, including stochastic gradient descent (SGD), logistic regression, Linear SVC, passive-aggressive, ridge, and perceptron, on a dataset consisting of malicious and benign URLs obtained from multiple sources. Models are assessed in terms of classification performance, computational efficiency, and resource utilization. Experimental results show that linear models achieve over 99% accuracy on one dataset while maintaining significantly reduced training and inference time. However, the performance of the classifiers has been evaluated on another dataset, which reveals the performance variations of the classifiers, thereby highlighting the generalization challenges faced by the classifiers in different datasets. Moreover, the SHAP values provide a better understanding of the classifiers. These results show that lightweight linear models are an effective and scalable approach to detect phishing in real-time. The findings also stress the need for cross-dataset evaluation to obtain an accurate depiction of performance.

Albalawi, A. (2026). Technical Note: Performance Evaluation of Lightweight Linear Models for Phishing URL Detection on Commodity CPUs: Accuracy-Efficiency Trade-offs and Cross-Dataset Generalization, Journal of Shaqra University for Computing and Information Technology, 2(1), 33-42.

Drone-assisted mobile charging can extend electric-vehicle (EV) operability in areas where fixed charging infrastructure is sparse or congested, but it introduces new challenges in multi-party coordination, data integrity, and secure settlement. This paper presents a consortium-blockchain protocol for secure communication and coordination in drone-assisted EV charging. EVs and drones operate as lightweight clients, while charging providers and fleet stakeholders operate validator nodes using a low-latency Byzantine-fault-tolerant proof-of-stake style consensus. The protocol defines (i) role-based registration, (ii) signed charging requests and drone status updates, (iii) an assignment workflow with explicit confirmation timeouts, and (iv) an escrow-based payment contract that releases funds only after verifiable charge completion. To reduce on-chain overhead, highrate telemetry is kept off-chain and anchored on-chain via cryptographic hashes. In simulation (100 EVs in 50 km, 10 drones), the proposed approach achieves 8 s average request-to-assignment latency under normal load and 12 s under high load, outperforming centralized and nonblockchain baselines; it also reduces mean travel distance and lowers per-session energy usage (4.0 kWh vs. 4.2 kWh centralized). These results suggest that consortium blockchain can improve auditability and resilience without prohibitive latency for moderate-scale deployments.

Alangari, S. (2026). Blockchain-based Secure Communication and Coordination Protocol for Electric Vehicle Charging using Drone-assisted Mobil Charging Stations, Journal of Shaqra University for Computing and Information Technology, 2(1), 43-57.