Saudi Cultural Missions Theses & Dissertations
Permanent URI for this communityhttps://drepo.sdl.edu.sa/handle/20.500.14154/10
Browse
98 results
Search Results
Item Restricted Evaluating Static, Contextual, and End-to-End Embedding Techniques for Malware Detection on Dynamic API Call Data(Saudi Digital Library, 2026) Basfar, Mohammed Raed; Joey, LamThe rate of malware development continues to challenge cybersecurity, with traditional signature- and heuristic-based techniques overwhelmed by polymorphic and zero-day attacks. Natural language processing (NLP) offers a promising direction by modeling dynamic API call sequences as semantic linguistic data, enabling sophisticated embedding and sequence-learning methods to be used for malware detection. This dissertation contrasts and analyzes three typical embedding methods static, contextual, and end-to-end task-learned representations—under a shared experimental framework. Specifically, it employs Word2Vec embeddings with a Convolutional Neural Network (CNN), contextual BERT embeddings with a CNN, and a Bidirectional Long Short-Term Memory (BiLSTM) network with a trainable embedding layer and weighted loss function to address class imbalance. The experiments were conducted on a dynamic API call dataset of around 44,000 malware and 1,000 benign samples, summarized by the first 100 API calls executed under sandboxed conditions. Results indicate that the Word2Vec + CNN pipeline had the highest overall accuracy and malware detection precision but the lowest benign recall. The BERT + CNN model provided more balanced class performance, but at the expense of added computational overhead. The BiLSTM had the highest benign recall, as it was able to easily distinguish from non-malicious activity, but the lowest precision and hugely added resource use. The findings point out the competing trade-offs among detection accuracy, benign recall, and processing efficiency, highlighting the issue of aligning model selection with actual security contexts' resource constraints and priorities. The study contributes by reporting a comparative systematic review of the embedding approaches for malware detection and offering informative insights into performance vs. efficiency trade-offs. Apart from its scientific significance, it proves the larger potential of NLP-based approaches to supporting malware detection systems and to informing the design of responsive, resource-aware cybersecurity systems.7 0Item Restricted The Impact of Digital Transformation on Enhancing the Medical Supply Chain(Saudi Digital Library, 2025) Abdulaziz, Albaraa; Theo, FotisThe current thesis aims to investigate how digital transformation (DT) technologies, artificial intelligence (AI), the Internet of Things (IoT), and blockchain can be used to improve the medical supply chain in Saudi Arabia through the Vision 2030 metric. A systematic review based on PRISMA was used to identify and screen the studies (n=6); narrative and thematic synthesis were performed as part of the study, and three general themes were created: operational excellence, economic viability, and implementation readiness. According to the findings, DT enhances visibility, predictability, responsiveness, and resilience, but the strength of evidence was not consistent, and some of the studies have indicated a high implementation cost and organisational barriers. It is important to note that the synthesis has revealed the convergence as well as tensions among studies that technical interoperability and workforce readiness are important in the effective adoption of DT. Some of these recommendations are gradual execution of technology, capacity building initiatives and sponsorship of national policy to enable equal adoption in all the facilities. Its implications go beyond Saudi Arabia, and the fact that DT serves as an agent of predictive and agile systems and the global discussions about how digital health transformation strategies are to be implemented.6 0Item Restricted The Impact of Artificial Intelligence on Supply Chain Optimization in Saudi Arabia(Saudi Digital Library, 2025) AlQahtani, Abdullah Saeed; Khobzi, HamidThis dissertation examines the impact of Artificial Intelligence (AI) on supply chain optimization in Saudi Arabia, with particular emphasis on its alignment with the Kingdom’s Vision 2030 objectives. AI technologies such as machine learning, predictive analytics, robotic process automation, and the Internet of Things are increasingly recognized for their potential to enhance efficiency, resilience, and sustainability within supply chain operations. However, despite growing national interest, empirical research focusing on AI adoption in the Saudi supply chain context remains limited. The study adopts a qualitative, interpretivist approach based on multiple secondary case studies drawn from peer-reviewed literature published between 2020 and 2025. The analysis is guided by the Technology–Organization–Environment (TOE) framework, supported by the Supply Chain Operations Reference (SCOR) model, to examine both adoption drivers and process-level applications across key sectors, including telecommunications, healthcare, manufacturing, and national mega-projects. Findings indicate that AI adoption in Saudi supply chains is most advanced in planning, forecasting, and logistics delivery, while challenges persist in system integration, data quality, workforce readiness, and organizational resistance to change. Environmental factors such as Vision 2030 initiatives and government support act as strong enablers, although adoption remains concentrated among large organizations and flagship projects. The study concludes that while AI has significant potential to transform Saudi supply chains, its full benefits depend on improved digital integration, skills development, and supportive policy frameworks.45 0Item Restricted Unblocking Operational Excellence: The Role of Digital Transformation in Optimising the Oil and Gas Supply Chains(Saudi Digital Library, 2025) AlObaidi, Danah; Jappie, AGDigital transformation is increasingly recognised as a systemic enabler of operational excellence in oil and gas supply chains, where volatility, and regulatory demands exceed traditional efficiency tools. This study investigates how four technologies -the Internet of Things (IoT), Artificial Intelligence (AI), blockchain, and digital twins- contribute through alignment. Using a qualitative, interpretivist approach, a systematic literature review and thematic analysis of 36 peer-reviewed studies identified five themes: predictive and AI-enabled performance, real-time visibility, integration bottlenecks, blockchain for compliance, and organisational fitness. Results show that integration maturity consistently outperforms novelty in driving value, while tensions such as explainability versus latency and governance versus agility shape adoption outcomes. The dissertation advances a conceptual framework linking digital enablers to Operational excellence through a “stacked capability” logic and offers practical recommendations for oil and gas leaders. Ultimately, digital transformation emerges not as a linear solution but as a discipline of orchestration that enables reliable, scalable, and trusted operational performance.17 0Item Embargo Comparative Study of the Performance of an Artificial Intelligence Platform in Detecting Periapical Radiolucencies Across Different Imaging Modalities(Saudi Digital Library, 2025) Allihaibi, Marwa; Koller, Garrit; Mannocci, FrancescoAim: This thesis aimed to evaluate the diagnostic accuracy of a commercial artificial intelligence (AI) platform in detecting periapical radiolucencies (PARLs) across different imaging modalities. The evaluation included preoperative assessment of teeth requiring primary endodontic treatment with comparison against dental professionals, radiographic healing assessment at follow-up, and assessment of teeth referred for apical microsurgery. Methods: Five retrospective diagnostic accuracy studies were conducted to evaluate the commercial AI platform Diagnocat (versions 1.0 and 2.0) for PARL detection across multiple imaging modalities. The studies utilised radiographic data from patients treated at Guy's and St Thomas' NHS Foundation Trust between 2012-2023. The study sample included: (1) 339 teeth indicated for primary root canal treatment, assessed on periapical radiographs (PARs) and compared with two experienced endodontists; (2) 376 teeth assessed at minimum one-year follow-up on PARs for radiographic healing outcomes, compared with two endodontists; (3) 134 molars evaluated on cone-beam computed tomography (CBCT) for preoperative and postoperative assessment; (4) 177 posterior teeth requiring primary endodontic treatment, assessed on PARs and compared with eleven general dental practitioners (GDPs); and (5) 116 anterior teeth referred for apical microsurgery, evaluated on both PARs and CBCT. Reference standards varied by study design: CBCT for PAR validation, expert consensus for CBCT assessment, and histopathology for cases referred for apical microsurgery. Statistical analyses included calculation of sensitivity, specificity and accuracy with 95% confidence intervals. McNemar's test assessed diagnostic performance differences. Subgroup analyses examined performance across anatomical variables. Results: Across five retrospective studies, Diagnocat demonstrated significant performance variability dependent on imaging modality, anatomical location, and treatment status. On PARs, for non-root-filled teeth requiring primary root canal treatment, sensitivity was 47.9% and specificity 95.4%, indicating reliable exclusion of disease but missing over half of actual lesions. In root-filled teeth assessed at one-year follow-up, sensitivity increased to 67.3% while specificity decreased to 82.3%, suggesting altered diagnostic thresholds based on treatment status. Performance on CBCT scans of molars showed marked improvement, achieving 93.9% sensitivity and 65.2% specificity in preoperative cases, and 88.6% sensitivity and 63.3% specificity in follow-up cases. While three-dimensional (3D) imaging substantially enhanced sensitivity for posterior teeth, it was accompanied by reduced specificity, indicating potential for overdiagnosis. Anatomical analysis revealed consistent underperformance in maxillary teeth and specific roots on PARs, limitations that were largely resolved on CBCT for posterior teeth. In contrast, anterior teeth demonstrated persistently poor performance regardless of imaging modality, achieving only 63.8% sensitivity on PARs and 57.5% on CBCT despite histopathological confirmation of periapical pathology. Cross-modality consistency was poor, with only 43.8% of lesions detected on both imaging modalities. Compared to clinicians, Diagnocat showed lower sensitivity (47.9% vs 65.3%) but comparable specificity (95.4% vs 97.7%) when assessed against endodontists in non-root-filled teeth. In root-filled teeth, this pattern reversed, with the AI achieving higher sensitivity (67.3% vs 49.3%) but lower specificity (82.3% vs 92.5%). When compared with GDPs, Diagnocat demonstrated lower sensitivity (44.9% vs 80.8%) but markedly superior specificity (94.3% vs 47.5%). Re-evaluation with version 2.0 showed no improvement in PARL detection across 1,308 PARs and 268 CBCT scans. Conclusion: This thesis demonstrated that multiple factors critically determine AI diagnostic accuracy for PARL detection, including imaging modality, anatomical location, and treatment status, thus highlighting fundamental limitations in training data representation and model development. AI platforms require comprehensive training on datasets with balanced anatomical representation and the incorporation of three-dimensional imaging before being considered for reliable implementation in endodontic diagnosis.15 0Item Restricted AI-Based Analysis of Magnetic Nanoparticle Relaxometry Curves for Structure-Specific Cancer Detection and Classification(Saudi Digital Library, 2025) AlHumam, Malack; Hovorka, OndrejCancer remains one of the world’s leading causes of death, and the key to successful treatment relies heavily on early and accurate diagnosis. This thesis explores a minimally invasive diagnostic method by combining magnetorelaxometry (MRX) with artificial intelligence (AI). Magnetorelaxometry measures how magnetic nanoparticles relax after being excited by an external magnetic field, producing relaxation curves that depend on anisotropy orientation and variation, particle number, structure geometry. Among magnetic nanoparticles, superparamagnetic iron oxide nanoparticles (SPIONs) are particularly suited for biomedical applications due to their biocompatibility and tunable relaxation properties. However, these curves often overlap and appear indistinguishable to the human eye, making traditional analysis challenging. The central research question of this thesis is whether AI can classify nanoparticle ensembles by structure and particle number from their relaxation curves, using them as unique markers for cancer detection and classification. To address this, five simulated datasets were generated, each incorporating multiple structures with different particle numbers under varying anisotropy conditions. After preprocessing, the data were analyzed with supervised, semi-supervised, and unsupervised models, supported by dimensionality reduction visualizations (PCA, t-SNE, UMAP). Supervised models achieved the strongest performance, with multiclass logistic regression reaching an accuracy of 0.89 in the dataset with aligned anisotropy and no variation. ZChains consistently emerged as the most distinguishable ensembles, relaxing roughly twice as long as YChains and providing clearer separability in both geometry and particle number, as confirmed by PCA scatter plots. In contrast, YChains frequently collapsed under z-axis anisotropy alignment, while Triangles and Rings were distinguishable only under controlled anisotropy variation. Arkus structures degraded rapidly when anisotropy variation increased. Semi-supervised pseudo-labeling maintained comparable accuracy of 0.817 under limited labeling, while unsupervised KMeans clustering, although non-predictive, provided insights into ensemble overlap and natural similarity groupings. The main contribution of this work is the demonstration that AI can classify nanoparticle ensembles through relaxation curve morphology rather than biomarker binding assays. This represents a shift from proof of detection toward structure-based classification, bridging magnetic physics with biomedical AI applications. Future directions include aligning anisotropy axes experimentally, exploring relaxation saturation for cancer staging, and translating AI pipelines to real biological magnetorelaxometry data.6 0Item Restricted Artificial Intelligence in Armed Conflict: Responsibility and Legal Reforms to Ensure Compliance with International Humanitarian Law(Saudi Digital Library, 2025) AlJerais, Najd; Master’s, degreeThe growing integration of artificial intelligence (AI) into military operations presents significant and unique challenges for international humanitarian law (IHL). This dissertation aims to answer the question of the extent to which AI-enabled weapons challenge compliance with IHL, and how effective potential reforms are in addressing these challenges. It first explores current and emerging military AI technologies and the application of IHL, analysing the risks posed to the core principles of distinction, proportionality, and precaution. It then identifies the main compliance challenges introduced by the deployment of AI-enabled weapons in the conduct of hostilities. These include the weakening of human judgment, responsibility gaps in both state and individual responsibility within the chain of command, and the fragmentation caused by the absence of international consensus. Finally, the dissertation considers potential responses, focusing on strengthening Article 36 weapons review procedures under Additional Protocol I, implementing human rights and ethical oversight, and developing enforceable international standards to ensure meaningful human control. It argues that while IHL remains the central legal framework governing armed conflict, it requires urgent strengthening through binding legal, ethical, and governance reforms. Ultimately, the discussion contributes to ongoing debates on the future of warfare by clarifying the limits of the IHL framework, evaluating the effectiveness of proposed reforms, and highlighting the need for a coherent regulatory approach capable of responding effectively to rapid technological change.12 0Item Restricted A Facial Expression-Aware Edge AI System For Driver Safety Monitoring(Saudi Digital Library, 2025) Almodhwahi, Maram; Wang, BinThis dissertation presents a driver monitoring system (DMS) that integrates emotion recognition to address critical issues in road safety. Road safety has become a global concern due to the significant increase in vehicle numbers and the rapid growth of transportation infrastructure. The number one cause of road accidents is human error, with a 90% ratio, with common contributing factors like distraction, drowsiness, panic, and fatigue. Traditional DMS approaches often fall short in identifying these emotional and cognitive states, limiting their effectiveness in accident prevention. To address these limitations, this research proposes a robust, deep-learning-based DMS framework designed to identify and respond to driver emotions and behaviors that may compromise safety. The proposed system utilizes advanced convolutional neural networks (CNN), specifically the inception module and Caffe-based ResNet-10 with a single-shot detector (SSD), to perform efficient facial detection and classification. These chosen model structures helped balance computational efficiency and accuracy. The DMS is trained on an extensive, diverse dataset comprising approximately 198,000 images and 1,600 videos sourced from multiple public and private datasets, ensuring the system’s robustness across a range of emotions and real-world driving scenarios. Emotions of interest include high-risk states such as drowsiness, distraction, and fear, alongside neutral conditions, and the model can perform well in different conditions, including low-light and foggy/blurry environments. Methodologically, the system incorporates essential data preprocessing techniques such as resizing, brightness normalization, pixel scaling, and noise reduction to optimize the model’s performance. On top of that, data augmentation and grayscale conversion improves the dataset’s variability, allowing the decrease of computational costs without sacrificing accuracy. This approach enabled the model to achieve high performance metrics, with an overall accuracy of 98.6% , an F1-score of 0.979, precision of 0.980, and recall of 0.979 across the four primary emotional states. This research contributes to the field by offering a less invasive, real-time solution for monitoring high-risk driver behaviors and providing insights for further advancements in automated driver assistance technologies. Future directions include optimizing the system for microcontrollers with low power consumption and implementing alerts for high-risk states to further mitigate accident risks, as well as including a multi-modal fusion of data from different sources (Infrared Camera, and a Microphone) to increase emotion recognition accuracy, which leads to taking better control and initiating more efficient proactive interventions.19 0Item Restricted MAPPING ELIXIR COMMUNITIES: EVALUATING CONTENT-BASED AND NETWORK-BASED APPROACHES FOR SYSTEMATIC EXPERT DISCOVERY IN UK BIOMEDICAL RESEARCH(Saudi Digital Library, 2025) Shafi, Suha; Goble, CaroleELIXIR, Europe’s distributed life sciences infrastructure coordinating 18 communities, currently lacks systematic methods for identifying UK researchers whose work aligns with its community goals. This dissertation develops and evaluates two computational pipelines to address this challenge: a content-based approach using semantic matching of publication content, and a network-based approach using co-authorship expansion from verified UK ELIXIR members. The content-based pipeline processed 80,849 UK-affiliated publications retrieved through structured PubMed queries derived from ELIXIR communities’ descriptions, generating BioBERT embeddings for semantic similarity search. The network-based pipeline expanded from 86 verified ELIXIR UK authors through co-authorship networks, processing 370,282 publications and assigning themes using a hybrid keyword-embedding approach. Both systems were integrated into Retrieval-Augmented Generation (RAG) architectures enabling complex expert discovery queries. Evaluation employed four complementary frameworks: coverage validation of the content-based approach achieving 100% success in identifying known UK ELIXIR authors; systematic overlap analysis revealing 24.4% author overlap but only 3.6% publication overlap between systems; literature-based expert evaluation using evidence-based bibliometrics criteria; and parameter sensitivity testing across more than 600 configurations confirming robustness of the system across different thresholds and expert scoring methods. When rigorous evaluation criteria (ten or more publications, multi-institutional collaboration, recent activity since 2020) were applied, the content-based system identified 26,111 experts with 100% confidence, while the network-based system identified 28,567 experts with 99.97% confidence. Critically, only 924 experts (5.6%) were validated by both methods using the literature-based expert evaluation, demonstrating that the approaches identify fundamentally different expert populations. Network-based discovery excelled at finding collaborative, early-career researchers (70.9% versus 49.6%) in established computational domains like Galaxy workflows. Content-based discovery excelled in finding focused specialists (99.1% single-theme) and mid-career researchers in emerging interdisciplinary areas like Rare Diseases. The 924 overlapping experts proved to be cross-domain bridges, appearing in different ELIXIR communities 93% of the time and discovered through entirely different evidence by each method. The investigation demonstrates that content-based and network-based approaches access different dimensions of expertise: intellectual contribution versus social integration, validating the need for combined deployment. Together, the systems identify over 17,333 unique UK experts across more than 54,500 researcher-community mappings, providing ELIXIR with comprehensive, quality-assessed mappings for strategic community development and engagement.21 0Item Restricted Generating biodegradable molecular composites with MolGPT : A transformer based approach(Saudi Digital Library, 2025) AlJeldah, Futoon M; Hosni, ZiedThis work presents the development of biodegradable polymer composites using the MolGPT generative transformer model. MolGPT was trained on the GuacaMol dataset and fine-tuned on the COCONUT datasets to produce valid, unique, and novel molecules. The model achieved 98.7%, 96.4%, and 94.1% in validity, uniqueness, and novelty, respectively. confirming its capability to generate chemically diverse structures. A Random Forest classifier trained on a QSAR biodegradation dataset was used to classify candidates as readily or non-readily biodegradable. Readily biodegradable molecules were selected for further evaluation and validation. AutoDock Vina was employed to dock these candidates onto a polyethylene (PE) fragment, with the lowest-energy mode subjected to DFT calculations at the B3LYP/6-31G(d) level. The docked PE–biodegradable complex exhibited HOMO–LUMO gaps of 2.1 eV, together with a binding energy of –17.6 kcal/mol. These results demonstrate that MolGPT can generate novel biodegradable candidates and that their interactions with polyethylene enhance electronic reactivity, providing a foundation for understanding how biodegradable molecules can promote polymer degradation and a basis for future laboratory validation and material design.11 0