Saudi Cultural Missions Theses & Dissertations
Permanent URI for this communityhttps://drepo.sdl.edu.sa/handle/20.500.14154/10
Browse
147 results
Search Results
Item Restricted AI-Enabled Autonomous Knowledge Extraction from Large-Scale Textual Data(Saudi Digital Library, 2026) Alharbi, Abdulrahman; Obradovic, ZoranThe rapid growth of large-scale textual data across social media platforms, news media, and scientific repositories presents both unprecedented opportunities and significant challenges for extracting meaningful insights. During global events such as the COVID-19 pandemic, understanding public discourse requires analyzing vast amounts of noisy, heterogeneous, dynamic, and geographically distributed data. At the same time, the exponential increase in scientific publications has made traditional evidence synthesis methods increasingly labor-intensive, time-consuming, and difficult to scale. Existing approaches to textual knowledge extraction often operate in isolation, lack interpretability, fail to integrate heterogeneous data sources, and do not support scalable end-to-end automation. This dissertation addresses these limitations by proposing a unified framework for AI-enabled autonomous knowledge extraction from large-scale textual data. The research introduces a comprehensive pipeline that integrate sentiment analysis, topic modeling, semantic interpretation, spatiotemporal reasoning, and multi-agent automation for scalable, robust and reproducible text analysis across heterogeneous domains. First, this work introduces TriLex, a novel unsupervised sentiment analysis framework that combines multiple lexicon-based sentiment analysis methods through weighted aggregation, majority voting, and dynamic thresholding technique to improve robustness and accuracy for short and noisy textual data. Building on this foundation, a hierarchical spatiotemporal framework is developed to capture the evolution of public sentiment across global, national, and regional scales. The framework integrates over 7 million social media posts and thousands of news articles to analyze COVID-19 vaccine discourse across time, geographic regions, and platforms. To enhance topic interpretability, this research integrates BERTopic with large language models (LLMs), enabling automated generation of coherent and context-aware topic representations for large-scale textual discourse. A cross-platform analytical framework is further introduced to examine temporal relationships between social media and news media discourse, demonstrating a bidirectional relationship in which news coverage and public discourse influence each other over time. Extending beyond discourse analysis, this dissertation introduces an Agentic AI framework that automates the end-to-end process of large-scale multilingual knowledge extraction and evidence synthesis. The proposed multi-agent system coordinates specialized agents for query generation, multilingual retrieval, metadata harmonization, title and abstract screening, and full-text analysis. Evaluated on a multilingual corpus of over 52,000 scientific records, the framework achieves high screening performance while substantially reducing processing time from months to hours, demonstrating significant improvements in scalability, robustness, and reproducibility. Collectively, this dissertation bridges the gap between analytical understanding and autonomous knowledge extraction from large-scale textual data. By integrating robust sentiment analysis, interpretable topic modeling, spatiotemporal discourse analysis, and autonomous AI systems within a unified framework, this work establishes a scalable and extensible paradigm for transforming heterogeneous textual data into actionable knowledge. The proposed methodologies are validated using real-world datasets spanning social media, news media, and scientific literature across diverse application domains.13 0Item Restricted Harmonic Impact Assessment and Prediction in a Power System with Nonlinear Loads(Saudi Digital Library, 2025) Almogargesh, Yousef Yaqoob Y; Swamidoss, SathiakumarThis thesis presents a simulation-based and machine learning framework for predicting power quality (PQ) indices in residential power systems with nonlinear loads. Four MATLAB/Simulink models representing nonlinear loads, induction motor loads, thyristor converters, and photovoltaic grid-connected systems were developed to generate voltage and current waveforms under different operating conditions. Time-domain and frequency-domain features were extracted from the simulated signals and used to train artificial neural network (ANN) models for predicting six key PQ indices: Total Harmonic Distortion (THD), Crest Factor, Form Factor, Power Factor, RMS Voltage, and Peak Voltage. Both shallow and deep neural network architectures were implemented and evaluated using statistical performance metrics including Mean Squared Error (MSE), Mean Absolute Error (MAE), and Pearson correlation coefficient. The results demonstrate that the proposed approach accurately predicts future power quality conditions while reducing the need for repeated analytical calculations. This framework provides an efficient tool for power quality monitoring and supports the development of intelligent and proactive power system management strategies.17 0Item Restricted Personalized Course Recommendations Leveraging Machine & Transfer Learning Toward Improved Student Outcomes(Saudi Digital Library, 2026) Algarni, Shrooq; Frederick, SheldonAt matriculation, university advising typically operates under tight informational constraints, often with no access to post-enrolment interaction history. We propose a unified, leakage-controlled pipeline that (i) predicts early dropout risk and (ii) generates cold-start programme recommendations using only pre-enrolment signals, with an optional early-warning variant that additionally incorporates first-term academic aggregates. The pipeline instantiates lightweight multimodal components: a tabular RNN, a DistilBERT encoder for short profile sentences, and a cross-attention fusion module, trained and evaluated end-to-end on a public benchmark (UCI id 697; n = 3630 students across 17 programmes). For dropout prediction, fusing text with numeric features yields the strongest thresh olded performance (Hybrid RNN–DistilBERT: F1 ≈ 0.9161, MCC ≈ 0.7750), while simple ensembling modestly improves threshold-free discrimination (AUROC up to ≈ 0.9488 via Stacking Ensemble, compared to ≈ 0.9459 for Weighted Ensemble). A text-only branch performs substantially worse, indicating that numeric demographics and early curricular aggregates carry most of the predictive signal at this horizon. For programme recommendation, pre-enrolment demographics alone support actionable rankings (De mographic MLP: NDCG@10 ≈ 0.5793, Top-10 ≈ 0.9380), outperforming a popularity prior by roughly 25–27 percentage points in NDCG@10; adding text yields only marginal improvements in hit rate and does not improve NDCG on this cohort. Methodologically, we apply leakage guards, deterministic preprocessing, stratified splits, and comprehensive metric reporting to enable reproducibility on non-proprietary data. Practically, the pipeline supports orientation-time triage via high-recall early warning and shortlist generation for programme selection. Overall, the results cast matriculation-time advising as a joint prediction–recommendation problem solvable with carefully engineered pre-enrolment views and lightweight multimodal models, without relying on historical interactions.13 0Item Restricted Machine Learning Algorithms for Secure and Reliable Electric Grid Operations and Control(Saudi Digital Library, 2026-09) Bahwal, Obai; Sankar, LalithaThis dissertation develops Machine Learning (ML) algorithms for secure and reliable electric grid operations and control by addressing three related problems. The first part studies real-time event identification using synchrophasor measurements, physics-based modal decomposition, and interpretable classifiers to distinguish generation loss from load loss events. Targeted adversarial attacks are developed to evaluate robustness under both white box and gray box settings, showing that learned event identification models are susceptible to adversarial attacks and that simpler models such as logistic regression are generally more vulnerable than gradient boosting. The second part builds on this vulnerability analysis and focuses on enhancing the security of ML event identification models in a white box adversarial setting. Two mitigation strategies are developed: robust classification through iterative adversarial retraining, and a dual-classifier architecture that combines event classification with attack detection. Numerical results on the synthetic South Carolina 500-bus system show that while robust retraining provides only modest improvement, the dual classifier approach is highly effective, reducing successful undetected attacks to under 0.1%. The third part addresses reliable grid control through a forecast-integrated rolling-horizon Model Predictive Control (MPC) framework for net-demand balancing using Distributed Energy Resource Aggregators (DERAs). Each DERA is modeled as a generalized battery with state-of-charge, power, and ramping constraints, while Linear Regression (LR) and Long Short-TermMemory (LSTM) forecasting models are integrated with MPC to generate real-time allocation policies. Using high-resolution California Independent System Operator (CAISO) net-demand data, results show close tracking of net-demand and reveal clear tradeoffs among forecast horizon, update frequency, and control performance, with LSTM generally benefiting longer time-shifts and LR remaining competitive for shorter update intervals. These three parts show that effective ML for power systems must be accurate, physically grounded, cyber-resilient, and compatible with real-time operational constraints.16 0Item Restricted Algorithms for Growth Pattern Based Analysis of Lung Adenocarcinoma Histology(Saudi Digital Library, 2026) AlRubaian, Arwa; Rajpoot, Nasir; Raza, ShanLung Adenocarcinoma (LUAD) is a leading cause of cancer deaths and shows considerable variation in patient outcomes due to its complex histological, immune, and molecular diversity. Traditional computational pathology methods use global, pattern-agnostic representations that do not capture the structured organization of tumor tissue. This thesis develops new, pattern-aware computational approaches to address this gap. First, we introduce CellOMaps, a biologically informed image representation that encodes the cellular organization of LUAD tissue. By focusing on spatial structure rather than raw visual appearance, CellOMaps enables robust, scalable characterization of LUAD histological growth patterns. Using this representation, we propose a framework for dense, patch-level growth pattern classification across Whole Slide Images (WSIs), offering a reproducible and spatially resolved alternative to conventional predominant-pattern assessment and demonstrating improved robustness across institutions. Next, we investigate immune heterogeneity in LUAD by analyzing the spatial distribution of Tumor-Infiltrating Lymphocytes (TILs). We introduce GPS-TILs, a growth pattern–specific digital biomarker that quantifies the presence, density, abundance, and spatial dispersion of immune infiltration within distinct morphological patterns. This pattern-aware immune profiling improves prognostic stratification compared to global TILs assessment and manual grading, highlighting the importance of incorporating histological context into immune analysis. Finally, we address molecular inference from histopathology by predicting Tumor Mutational Burden (TMB) from WSIs. We introduce GRIL-GNN, a graph-based learning framework that aggregates information across spatially organized tissue regions while preserving detailed morphological features. This approach uses a novel unified learning objective that combines supervised learning with self-supervised regularization, enabling robust representation learning under weak supervision. By restricting analysis to localized tissue regions with shared histological patterns, this work examines the distribution of the TMB predictive signal within different morphological patterns. Overall, this work demonstrates that integrating spatial and pattern-aware analysis improves the interpretation of morphological, immune, and molecular signals in LUAD. The approaches proposed here provide a framework for more precise, clinically relevant computational pathology, with potential applications to other cancers characterized by complex tissue structure.14 0Item Restricted Saudi Dialect Sentiment Analysis: A Hybrid Approach and Evaluation on Real-World Consumer Reviews(Saudi Digital Library, 2026) Alsemaree, Ohud; Singh Gill, SukhpalThis thesis presents a hybrid sentiment analysis framework for the Saudi dialect, addressing the limited availability of linguistic resources and annotated datasets for Arabic sentiment analysis. The proposed approach combines lexicon-based methods with machine learning and deep learning techniques to improve sentiment classification performance. A Saudi dialect sentiment lexicon (LSAnArTe) and a large-scale annotated dataset were developed to support the research. Experimental results demonstrate that the hybrid framework outperforms baseline models, achieving high classification accuracy and providing valuable resources for future research in Arabic Natural Language Processing and sentiment analysis.11 0Item Restricted MALWARE CLASSIFICATION VIA BYTECODE VISUALIZATION AND MULTIMODAL DEEP LEARNING(Saudi Digital Library, 2026) مكاوي, صالح; Kenneth, Barner; Michael De. LuciaThe rapid proliferation of Android malware poses a critical threat to mobile security, driven by the open-source nature of the Android ecosystem, broad access to its Software Development Kit (SDK), and the availability of multiple app distribution channels. Traditional detection methods, including signature-based, static, and dynamic analysis, often fail against novel or obfuscated variants that use encryption, packing, and polymorphism to evade detection. This dissertation addresses these limitations through three progressive, interconnected contributions. Contribution I introduces a semantic bytecode-to-image encoding method based on Shannon entropy, computed over sliding windows of Dalvik Executable (DEX) bytecode and mapped to the red and blue channels of an RGB image. This encoding directly exposes obfuscation artifacts in encrypted, packed, and compressed code regions that blind color-mapping approaches cannot distinguish. Evaluated on 184,474 samples from AndroZoo for binary malware classification, the entropy encoder outperforms the MalNet and Classbyte baselines across five state-of-the-art CNN architectures, achieving up to 95.77% accuracy and 98.25% ROC-AUC. Contribution II extends this encoding into a full multiclass framework and dataset, MalVis, by incorporating byte-level N-gram frequency statistics into the green channel. By combining entropy and N-gram signals, MalVis images capture both low-level randomness and high-level structural code patterns. To support the research community, we release the MalVis dataset, the largest publicly available Android malware visualization resource, comprising over 1.3 million labeled RGB images spanning nine malware families and benign samples. To validate that CNN models rely on the proposed encodings rather than spurious image artifacts, we apply Grad-CAM and Grad-CAM++ attention mapping, confirming that model attention consistently aligns with the entropy- and N-gram-encoded channels across all malware classes. Contribution III introduces ViCoMal, a late-fusion multimodal framework that addresses the single point of failure (SPOF) inherent in unimodal approaches. ViCoMal pairs MalVis image-based malware representations with 65 DEX-based contextual features, including 19 novel engineered features consisting of normalized risk scores, binary capability indicators, and rule-based behavioral profiles extracted using static analysis. Five CNN architectures and seven classical machine learning models are trained independently per modality, and their class-probability outputs are combined through eight fusion strategies. To handle a severe 169:1 class imbalance, SMOTE oversampling and class-weighted training are applied jointly. Evaluated under five-fold stratified cross-validation on the full 1.3M ViCoMal dataset, the best ensemble, which averages the predictions of the top three models (Random Forest, ResNet50, and XGBoost), achieves 91.84% accuracy, outperforming the strongest image-only baseline by 3.43 percentage points and the strongest contextual-only baseline by 2.05 percentage points. Together, these contributions establish a scalable, interpretable, and high-performing pipeline for Android malware detection, advancing the state of the art in both malware visualization and multimodal security analysis.12 0Item Restricted Artificial Intelligence through Machine Learning techniques to enhance the application of 3D body scanning in apparel shape and sizing(Saudi Digital Library, 2026) Alhassawi, Ruqey Ali; Simeon, Gill; Steve, Hayes; Kristina, BrubacherSignificant challenges persist in realising the full potential of technology related to accurate and inclusive body dimension variation and garment sizing and fit. Traditional methods often fail to capture the complexity of human body morphology, highlighting the value of more detailed approaches to analysing body dimension variation. This doctoral research aims to support the visual analysis of anthropometric population data through the integration of artificial intelligence (AI) and machine learning (ML) techniques, addressing limitations in traditional anthropometric methods used for apparel sizing and body–to–pattern mapping. A mixed–methods approach was employed across five interconnected phases, leveraging 3D body scanning (3DBS) technology to analyse and compare real–world body dimensions, classical garment sizing classifications and garment patterns. The research involved: (1) a comprehensive analysis of 3DBS data to establish body dimension diversity, (2) a critical reassessment of the traditional 8–head figure ratio, (3) clustering algorithms (Hierarchical, self–organizing map (SOM), k–means) to classify body types, (4) application of support vector machine (SVM) and principal component analysis–SVM (PCA–SVM) models for accurate size prediction, and (5) enhanced regression analysis to develop a data–driven approach for garment pattern adjustment. A dataset of 677 female participants from a range of ethnic backgrounds was utilised. Significant dimensional variations within conventional size groups were identified, revealing limitations in traditional measurement-based sizing systems within the study sample. Key findings demonstrate frequent deviations from the classical 8–head figure proportion model, emphasising the need for a more comprehensive approach. Clustering algorithms successfully delineated distinct morphological categories, while SVM modelling exposed trade–offs between predictive accuracy and computational complexity. Regression analysis established quantitative relationships between body measurements and pattern block parameters, offering a means of examining how body dimension variation relates to patternmaking practice. This research makes several theoretical, methodological and practical contributions. Theoretically, it provides data-based evidence of body proportion variability within standard size categories, challenges the classical 8-head figure proportion model using measured data, and identifies distinct body shape clusters within the study sample. Methodologically, it applies an integrated analytical framework – combining 3D body scan data, statistical analysis, ML clustering and classification, and regression analysis – to examine body dimension variation and body-to-pattern relationships. Practically, it provides how data-driven analysis of anthropometric variation may inform patternmaking considerations, subject to further applied investigation. This research examines the integration of 3D body scanning and computational techniques within anthropometric analysis. The use of data-based derived visual tools provides a means of representing and exploring body variation within the study sample. The findings highlight the potential relevance of data-driven approaches to sizing and may inform further investigation into how body diversity is represented within garment sizing systems.17 0Item Restricted ENHANCING TRAFFIC SAFETY THROUGH AI-DRIVEN, PRIVACY-PRESERVING, AND SECURE IMPAIRED DRIVING DETECTION SYSTEMS(Saudi Digital Library, 2026) Alsulieman, Razan; Sherif, AhmedDrunk driving remains a major threat to road safety worldwide, contributing significantly to traffic injuries and fatalities each year. Traditional detection approaches are largely reactive and vehicle-centric, relying on in-vehicle sensors, breathalyzers, or post-incident enforcement. These methods often depend on driver cooperation, intrusive hardware installations, or limited monitoring environments, restricting their scalability and effectiveness in large transportation systems. At the same time, modern cities increasingly deploy roadside cameras, surveillance networks, and drone based monitoring systems, creating new opportunities for proactive intoxication detection at the infrastructure level. However, leveraging such external monitoring introduces challenges related to secure data collection, reliable AI-based analysis, privacy protection, and real-world deployment. This dissertation proposes a secure, privacy-preserving Artificial Intelligence framework for proactive drunk driving detection using out-of-vehicle surveillance data. The framework addresses three key aspects required for reliable infrastructure-level monitoring. First, a lightweight authentication scheme is developed to ensure secure data collection from distributed monitoring platforms such as drones and surveillance devices. The proposed design employs physically unclonable functions and symmetric cryptographic primitives to provide protection against impersonation, replay attacks, and device cloning while maintaining low computational overhead for resource-constrainedenvironments. Second, AI-based intoxication detection models are developed using Machine Learning and Deep Learning techniques to analyze facial imagery captured under real-world surveillance conditions. Extensive experiments evaluate multiple models under varying noise and disruption scenarios to ensure robustness across both low- and high-resource computational environments. The framework also incorporates explainable AI methods to improve transparency and verify that model decisions rely on meaningful facial features. Finally, the framework integrates privacy-preserving learning mechanisms through federated learning, enabling distributed model training without transferring sensitive facial images to centralized servers. This approach protects user privacy while maintaining strong detection performance across distributed monitoring nodes. These contributions establish a secure, scalable, and privacy-aware infrastructure-level system for proactive intoxication detection, supporting intelligent transportation systems aimed at improving traffic safety.7 0Item Embargo Understanding Ransomware and Enhancing Their Detection Using Machine Learning(Saudi Digital Library, 2026) Alzahrani, Saleh; Xiao, YangRansomware attacks have escalated significantly in recent years, causing substantial financial losses and operational disruptions to individuals, organizations, and critical infrastructure worldwide. According to The Chainalysis 2024 Crypto Crime Report, ransomware attacks have imposed increasing financial burdens on victims over recent years. The total value received by ransomware attackers reached $1.1 billion in 2023, representing a significant rise from $567 million in 2022. This trend highlights the evolving threat posed by ransomware as attackers continue to refine their methods. Compared to $220 million in 2019. Despite the proliferation of detection methods, contemporary ransomware continues to evade traditional security measures through increasingly sophisticated evasion techniques. This dissertation addresses critical gaps in ransomware detection research through a investigation that combines in-depth malware analysis, evolutionary tracking, systematic literature review, novel detection methodology, and dataset development. The research begins with a detailed examination of Conti ransomware, one of the most notorious Ransomware-as-a-Service operations that caused approximately $45 million in damages and significantly impacted healthcare systems. Through analysis of leaked source code and controlled environment testing, this study reveals advanced evasion mechanisms including API disguise techniques, anti-hook mechanisms, and multithreaded encryption for rapid file encryption. Building upon this foundation, the research tracks Conti's evolution from its beta version through multiple iterations, categorizing samples into seven distinct versions. This longitudinal analysis demonstrates that modern ransomware success stems from continuous development and delivery practices, with features such as API hashing and runtime API loading being progressively integrated over time. To contextualize these findings within the broader detection landscape, a survey of existing ransomware detection methods was conducted, examining both machine learning and non-machine learning approaches alongside available datasets. This survey identifies critical limitations in current research, specifically that non-machine learning methods fail to identify new samples from known variants, while machine learning approaches suffer from inadequate model design and the absence of comprehensive, standardized datasets. These deficiencies severely limit their effectiveness against emerging ransomware variants. Addressing these identified gaps, this dissertation introduces RansomFormer, a Transformer-based detection model that leverages cross-attention mechanisms to fuse Portable Executable byte data with Application Programming Interface information, including both static imports and dynamic sequence calls. Unlike existing single-feature approaches that ransomware developers can circumvent, RansomFormer's multi-modal architecture achieves exceptional accuracy of 99.25% on static datasets and 99.50% on combined static-dynamic datasets across more than 150 ransomware families. Furthermore, recognizing the fundamental need for comprehensive training data, this dissertation presents RanDS, a rigorously curated dataset comprising a large collection of ransomware samples spanning hundreds of families alongside a substantial set of benign samples, collected and verified over multiple years from an initial corpus of millions of malware files. RanDS includes several processed feature extraction datasets encompassing static raw strings, English strings, imported and exported APIs, demangled APIs, and dynamic behavioral activities, all made publicly available. This dissertation makes contributions to cybersecurity by providing deep insights into modern ransomware operations, demonstrating the importance of evolutionary analysis in understanding threat progression, and delivering both an detection methodology and a foundational dataset that addresses longstanding research limitations in the field.67 0