SACM - Australia
Permanent URI for this collectionhttps://drepo.sdl.edu.sa/handle/20.500.14154/9648
Browse
2011 results
Search Results
Item Unknown Robust Machine Learning Under Distribution Shift: Out-of-Distribution Detection, Adversarial Robustness, and Fairness(Saudi Digital Library, 2026) Alotaibi, Dalha Alhumaidi N; Jianlong, Zhou; Fang, ChenMachine learning systems have demonstrated remarkable success across a wide range of real-world applications; however, their reliability remains highly sensitive to changes in data distributions. In practical deployment settings, the assumption that training and test data are identically distributed rarely holds. Instead, data may evolve over time due to temporal shifts in population behaviour, feedback effects induced by model deployment, the emergence of previously unseen out-of-distribution (OOD) inputs, or deliberate manipulation by adversarial attackers. These challenges arise from a common underlying cause, namely the breakdown of the assumption that training and deployment data follow the same distribution, highlighting that robustness is not a single property but a multi-faceted problem requiring coordinated solutions. Consequently, ensuring reliable and fair model performance under such dynamic conditions has become a critical research challenge. This thesis addresses robustness under distributional shift from a unified perspective, focusing on three principal axes: concept drift, unseen inputs, and adversarial manipulation. First, to address the impact of temporal and concept drift on fairness, this thesis investigates how existing fairness metrics fail to capture evolving disparities over time. To overcome this limitation, it introduces a novel drift-aware fairness metric that monitors and quantifies multi-attribute disparity as data distributions change. This approach enables more reliable detection of fairness degradation, particularly in dynamic environments where traditional static metrics are insufficient. Second, to address the challenge of unseen samples, this thesis investigates how previously unseen inputs can be identified using a simple and generalisable mechanism. To this end, it proposes a model scoring function that leverages intrinsic model behaviour to enable binary detection without requiring retraining or architectural modifications. This method provides a practical and scalable solution for real-world deployment scenarios where model flexibility and efficiency are essential. Third, in the context of adversarial shift, this thesis addresses the challenge of detecting and defending against intentionally manipulated inputs. It designs specialised adversarial detection mechanisms that exploit structural and representational inconsistencies introduced by perturbations. In addition, it proposes second-order, multi-layer defence strategies that leverage curvature-based information to reshape feature representations and enhance robustness against modern, stronger attack scenarios. Collectively, the contributions of this thesis establish a unified framework for monitoring, detecting, and defending against diverse forms of distributional shift. This work advances the development of machine learning systems that remain reliable, fair, and robust in dynamic, high-stakes environments.17 0Item Unknown Designing a Cybersecurity Culture Index – Measuring Human Risk in the Workplace(Saudi Digital Library, 2026) AlMardef, Fawaz Hamad; Priyanka, SinghThe goal of this study is to create a Cybersecurity Culture Index (CCI), which is a standardised framework for measuring and quantifying the organisation's cybersecurity culture and human-oriented cybersecurity risk. Many current cybersecurity strategies are technical, involving firewalls, encryption and intrusion detection, and have overlooked the importance of behavioural and cultural aspects that greatly affect the security results. Human error has been consistently identified as a top reason for cybersecurity incidents, as found in previous research, and organisations must take a systematic approach to understanding and managing their security culture. There is no unified organisational-level composite measurement model in the current tools and governance frameworks, such as the Human Aspects of Information Security Questionnaire (HAIS-Q), the NIST Cybersecurity Framework 2.0 or the ENISA Cybersecurity Culture Guidelines. The study employs a constructive research approach using systematic secondary data analysis and presents a seven-dimensional Cybersecurity Culture Index: Leadership Commitment, Behavioural Compliance, Policy Awareness, Security Training, Peer Influence, Risk Perception, and Incident Response. The framework comprises a 45-item Likert scale questionnaire and an aggregated model with weighting that delivers a composite score from a scale of 0 to 100. The study also theoretically validates the study by aligning the study constructs with the existing validated frameworks and instruments. The results indicate that the proposed CCI is designed to complement current cybersecurity culture measurement models by incorporating elements of organisational culture theory, psychometrics, and governance to form a measurable model. The CCI offers organisations a tangible mechanism for benchmarking, tracking progress over time, and pinpointing cultural areas where human-factor cybersecurity risks exist. The study ends with the observation that the CCI has the potential to serve as a very solid theoretical basis for testing and application in various organisational settings in the future.14 0Item Unknown Three Essays on Subjective Wellbeing, Efficient Energy Use and Vaccination Uptake: Evidence from the European Social Survey(Saudi Digital Library, 2026) Alsharari, Abdulaziz Zaid S; Onur, Ilke; Zhu, RongThis thesis examines why migration status matters for social, environmental and public health outcomes in contemporary Europe. Although the migration economics literature has traditionally focused on labour-market integration, earnings and fiscal outcomes, less is known about how migration is associated with non-market dimensions of integration and wellbeing. Existing studies also tend to examine subjective wellbeing, environmental behaviour and health behaviour separately, leaving limited cross-country evidence on whether immigrants differ from native-born populations across these broader outcomes, and on the mechanisms and heterogeneity underlying these differences. Addressing this gap is important because European countries are simultaneously managing migration, social integration, energy transition and public health challenges. The thesis consists of three empirical studies using individual-level data from the European Social Survey (ESS). The analyses of subjective wellbeing and energy-related behaviour draw on repeated cross-sectional ESS data from 2002 to 2023, while the analysis of coronavirus disease (COVID-19) vaccination uptake uses the relevant ESS wave collected between 2020 and 2022. Across the three studies, the thesis applies several econometric approaches, including fixed effects regressions, entropy balancing, instrumental variable estimation, heterogeneity analysis and non-linear modelling, to provide robust and policy-relevant evidence. The first study examines the relationship between migration status and subjective wellbeing, measured through life satisfaction, happiness, self-reported health and depression. The findings show that immigrants generally report lower levels of subjective wellbeing than native-born respondents, although the magnitude of this relationship varies across gender, regional groups and the distribution of wellbeing. The results suggest that migration-related differences in wellbeing are not uniform and should be understood in relation to social and demographic context. The second study investigates whether migration status is associated with household energy-related behaviour in European countries. The findings indicate that immigrant status is positively associated with self-reported energy-saving behaviour, with immigrants reporting more frequent energy-saving actions than native-born v respondents.. This highlights the importance of considering demographic change, household characteristics and migrant integration when designing energy-efficiency and environmental policies. The third study examines the relationship between migration status, institutional trust and COVID-19 vaccination uptake. The findings indicate that immigrants are less likely to be vaccinated than native-born respondents, and that higher levels of trust in institutions and scientists are positively associated with vaccine uptake. However, trust does not fully explain the vaccination gap between immigrants and native-born respondents, suggesting that other barriers may also shape public health behaviour. Overall, the thesis contributes to migration economics by extending the analysis of migration beyond employment and income to include wellbeing, energy behaviour and vaccination uptake. The findings provide evidence that migration status is linked to broader social outcomes and offer implications for policies on social integration, energy efficiency and public health communication in European countries.13 0Item Unknown Reconfiguring Teacher Professionalism through Standards: A Bourdieusian Exploration of Early Childhood Teachers in Saudi Arabia(Saudi Digital Library, 2026) Alhussain, Nouf; I-Fang Lee, Joanne AilwoodThis thesis examines the impact of the national professional standards (NPS) on early childhood teachers (ECTs) in Saudi Arabia. The NPS was introduced as part of Saudi Vision 2030 to enhance teacher quality through structured competencies, licensing requirements, and ranked career pathways. While the standards were intended to improve educational outcomes and align with international benchmarks, these reforms also intensified accountability measures and introduced performative pressures, particularly in the early childhood sector, in which relational and context-sensitive practices are central to effective teaching. Through this qualitative study, grounded in Pierre Bourdieu’s theoretical concepts of field,habitus, and capital, I integrated document analysis with data from semi-structured interviews with ECTs at various stages of their careers. I employed a critical analysis of the policy to identify the constructions of professionalism and the expectations placed on teachers. The interview data added an exploration of teachers’ experiences of the NPS and the tensions they faced between policy frameworks and professional values. My findings indicated that the NPS promotes a managerial model of professionalism, focused on compliance, performance metrics, and alignment with national development goals. Study participants’ engagement with the NPS varied based on their career stage,access to support, and institutional culture. While novice teachers tended to view the NPS as a pathway to recognition and advancement, more experienced teachers raised concerns regarding diminished autonomy, increased pressure, and the marginalisation of care-based and collaborative practices. Professional identity emerged as a contested space shaped by social recognition, policy discourse, and the availability of resources and networks. This study contributes to the understanding of teacher professionalism in non-Western contexts by shedding light on the lived experiences of Saudi ECTs. My findings challenged the universal application of standardised reforms and highlighted the need for policies that are sensitive to local values, professional agency, and the relational dimensions of teaching.11 0Item Unknown The L(j1,j2)-labelling and L(j1,j2,j3)-labelling problems for planar and outerplanar graphs(Saudi Digital Library, 2022) Alotaibi, Njud; Sanming, ZhouThe distance-labelling problem is an NP-complete problem thatmathematicians and com- puter scientists have studied extensively. In particular, for several graph classes, the L(i1,i2,...,in)−labelling problem has been widely researched. This thesis studies the distance-labelling problem for planar and outerplanar graphs. ThisresearchprojectfocusesonusingseveraltechniquestosolvetheL(i1,i2,...,in)−labelling problem. A genetic algorithm has been used in many NP-hard problems; however, there has been no study on L(i1,i2,i3)−labelling with a genetic algorithm introduced as a useful technique to provide results for hard optimisation problems. The aim of this project is to discuss the results on the distance-labelling problem and toproducetheexperimentalresultsforthegeneticalgorithm, andtwootheralgorithmsfor trees, identifying features of trees that impact the algorithms’ performance and comparing all algorithms.8 0Item Unknown Exploring Talent Management Practices and Their Role in Shaping Organisational Commitment and Employee Turnover Intention in the Saudi Arabian Retail Sector(Saudi Digital Library, 2026) Alharbi, Abdullah Ahmad; Campbell, Nadine; Klikauer, ThomasTalent management has grown in significance for organisations that aim to boost commitment and to reduce turnover, especially in labour-intensive industries such as retail. While previous studies have examined the effects of talent management practices, there is a lack of qualitative insight into how employees perceive these practices in particular settings, such as the Saudi Arabian private sector. This study investigates how talent management practices influence organisational commitment and turnover intention among frontline staff and managers in the Saudi private retail industry. The study employs a qualitative, interpretivist approach to explore participants’ lived experiences. Data were gathered via open-ended, face-to-face interviews with frontline and line managers in private retail firms. Thematic analysis, using Braun and Clarke’s framework, was used to analyse the interview data, providing a detailed understanding of how talent management practices are perceived and implemented in daily work settings. The findings indicate that talent management practices vary across organisational levels and are influenced by factors such as organisational size, job role and daily interactions with managers. Organisational commitment is mostly relational and localised, especially among frontline staff whose loyalty tends to be directed more towards immediate supervisors and work teams than to the organisation. Turnover intentions tend to develop gradually when talent management strategies are seen as inconsistent, unclear or lacking sufficient support at the local level. The study underscores the critical role of localised organisational support in shaping how talent management affect organisational commitment and employees’ decisions to stay or leave. This study provides context-specific insights into how talent management, organisational Commitment and turnover intentions are related within the Saudi retail sector. In addition, it offers practical recommendations for managers by highlighting the importance of consistent local support and tailored talent management practices in enhancing employee commitment and retention.25 0Item Unknown Evaluating the Implementation of the Vision 2030 Healthcare Strategy in Improving Breast Cancer Screening in Rural Saudi Arabia(Saudi Digital Library, 2026) Almalki, Shehab; McGrail, Matthew; Chater, Bruce; Bushra, NasirAbstract Introduction: Breast cancer is a major global health problem, and the timeliness of diagnosis by screening is important in providing a better prognosis. In Saudi Arabia, the provision of screening services has been increased through national reform led by the National Health Sector Transformation Program (NHSTP) under the health agenda of Vision 2030. These reforms aim to strengthen preventive healthcare services, improve early detection of disease, and enhance the overall efficiency and accessibility of the healthcare system. This thesis is designed to assess the implementation of the National Health Sector Transformation Program (NHSTP) in relation to breast cancer screening service access, delivery, and patient experience was in the rural governorate of Bisha, Saudi Arabia, which was part of the Vision 2030 reform agenda. It is especially significant to study these factors in a rural setting, where healthcare access, availability of resources, and community awareness levels might be different than in urban settings, which could influence the success of national screening programs. Considering the assessment of the NHSTP reforms implementation at the local level, the thesis would contribute some evidence that could be used in the future to improve breast cancer screening services and to facilitate further change in the provision of healthcare in Saudi Arabia. Methods: A multi-method design was employed in that, quantitatively, (Objective 1) routine data from the Ministry of Health dashboards was analysed to track MOC screening utilisation and diagnostic outcomes, and (Objective 3) a structured patient-reported satisfaction survey was used to capture domains of experience (access, communication, discomfort, environment) of women using MOC clinics. Qualitatively (Objective 2), semi-structured interviews with healthcare providers explored NHSTP/MOC implementation, pathway bottlenecks and context-specific constraints. Quantitative indicators included screening rates, diagnostic resolution, and service capacity for 2017-2023; qualitative information provided operational perspectives on outreach, referral timeliness, privacy arrangements, staffing, and cultural acceptability. Results: From these quantitative results, a significant improvement in breast cancer screening rates in Bisha over the period of 2017–2023 was observed. The number of women screened increased from 810 in 2017 to 7,101 in 2023, while diagnosed cases increased from 29 to 88 during the same period. Screening coverage among women aged 40–69 years increased from 4% in 2022 to 16% in 2023, and Bisha achieved 88.6% of its annual screening target in 2023. Bisha's national service coverage rate, however, stood at 53.8% against the national average of 80%, thus suggesting that regional disparities still existed after the service's expansion. These improvements occurred during the NHSTP implementation period but should not be attributed solely to the programme, as infrastructure expansion, staffing changes, dual shifts, and awareness campaigns may also have contributed. Qualitative findings indicated that healthcare providers perceived improvements in screening access, mobile services, public awareness, and service capacity, while also identifying persistent cultural, logistical, workforce, and follow-up barriers. Patient-reported findings showed generally favourable satisfaction with screening services, particularly in relation to access, privacy, communication, and overall experience, although concerns remained regarding physical discomfort, psychological anxiety, waiting for results, and confidence in technical aspects of care. Conclusion: The NHSTP has achieved significant improvements in Bisha in terms of access, capacity and utilization of breast cancer screening. But coverage is still not up to national standards and the structural change is not necessarily reflected in equitable and enduring participation, as the barriers to participation are experiential (procedural discomfort, psychological distress, cultural norms, and unequal digital access). To ensure equity in rural areas, such as Bisha, following national reform, targeted investments must be made in training providers, gender responsive staffing, culturally responsive community outreach, preprocedure psychological preparation, and offline access to appointments.15 0Item Restricted Multilingual Model Enhancement Framework for Spam Detection Using a Human-Centred Approach(Saudi Digital Library, 2026) Alqhtani, Saleh; Priyadarsi, Nanda; Qiang, WuPhishing attacks continue to exploit human behaviour across digital communication platforms, yet most defence solutions are designed for English speakers. Arabic-speaking users face increased risk due to limited awareness and a lack of high-quality phishing datasets. This thesis introduces a human-centred and culturally-aware cybersecurity framework that enhances phishing understanding and detection for Arabic users. The work begins with three empirical studies focused on phishing perception, emo- tional triggers, and behavioural responses among Arabic-speaking university students. Results reveal that urgency, authority appeals, and religious or familial influence sig- nificantly increase susceptibility. These insights expose cultural and emotional factors that are not represented within current English-centric phishing defences. To address dataset scarcity, the thesis develops a multi-stage English-to-Arabic translation pipeline that produces high-fidelity security text while preserving core de- ception signals such as URLs, numeric codes, and persuasion techniques. The pipeline combines context-aware prompting with multilingual neural translation and human validation to ensure linguistic accuracy and security relevance. A post-translation en- richment process integrates the user study insights by adding structured labels such as phishing type, emotional manipulation, cultural exploitation, behavioural red flags, audience targeting, and recommended defence actions. Three corpora are produced for email, SMS, and instant messaging, each containing 1,000 validated and enriched sam- ples. These represent the first Arabic phishing datasets that embed human behavioural intelligence within a unified annotation standard. The thesis then evaluates machine learning approaches for Arabic phishing detec- tion using the newly developed datasets. In the cross-channel evaluation, the fine- tuned mT5 model achieved accuracies of 87.2% on SMS, 91.5% on email, and 83.5% on messaging (87.4% average), with corresponding F1-scores of 86.9%, 91.5%, and 82.9%. In separate baseline-to-enhanced mT5 experiments, overall accuracy increased by 1.37 percentage points on SMS (96.84% to 98.21%) and by 7.46 points on Enron email (74.68% to 82.14%); Enron ham accuracy increased by 22.17 points (56.75% to 78.92%). The system also generates interpretable outputs that explain prediction rationale and reinforce user awareness, supporting educational and defensive goals. Overall, this thesis advances cybersecurity for underserved language communities by linking behavioural research with cross-lingual artificial intelligence. The frame- work improves phishing resilience among Arabic-speaking users and establishes essen- tial linguistic and cultural resources for future research on secure digital transforma- tion in the region.27 0Item Restricted Responsive nanoprobe for redox biomolecules quantification(Saudi Digital Library, 2026) Alanazi, Mazen; Run, ZhangRedox molecules have attracted scientific interest due to their crucial roles in regulating cellular function and maintaining redox balance. Among them, reactive oxygen species (ROS) and reactive sulfur species (RSS) are considered as highly important groups of signaling mediators. Behind their association with oxidative stress, these species critically participate in regulated redox reactions that impact diverse biological processes, including metabolism, inflammation, and cellular proliferation. Fluorescent-based nanoprobes were engineered to facilitate the monitoring of OH and H2S generation during critical biological processes, including macrophage polarization, tumor progression, and antioxidant interventions. These nanoprobes hold substantial potential to advance future research aimed at elucidating the redox dynamics underlying inflammatory responses and oncogenic transformations, thereby contributing to the development of more effective diagnostic and therapeutic strategies for related diseases.20 0Item Restricted Numerical solutions of higher-order flow of curves(Saudi Digital Library, 2026) Alhawiti, Noura; Bishnu, Lamichhane; James, McCoyThe purpose of this thesis is to develop efficient numerical tools for dealing with complex fourth-order flow evolution of curves, employing a step-by-step approach that starts from the heat equation in a finite interval. Our work establishes a novel framework utilising biorthogonal systems in the context of fourth-order PDE equations. The thesis develops numerical approximate solution methods for higher-order partial differential equations (PDEs) for geometric flows. In particular, a mixed finite element (MFE) method with a biorthogonal basis is developed after expressing a fourth-order PDE as a system of two second-order PDEs, then imposing constraints using Lagrange multipliers. Subsequently, we explore the parabolic evolution of curve diffusion type, and elastic flows with a nonlinear operator involving curvature, considering various boundary conditions (such as generalised Dirichlet (DBC), Neumann (NBC), clamped (CBC), and mixed (MBC) boundary conditions with initial conditions). We employ finite element (FE) techniques, initially for curves in arc length and graph parameterisation evolving by fourth-order nonlinear flows. We present applications of this method, using different initial curves. The error using the $L^{\infty}$-norm, $L^2$-error using the $L^2$-norm and $H^1$-error using the $H^1$-norm are calculated for the approach. By applying our mixed finite element method with a biorthogonal system to a specific class of flows, we achieve greater accuracy and efficiency compared to other methods. This application of a biorthogonal system to higher-order diffusive curvature flow is new. Before this project it needed to be verified if this approach improves accuracy as well as efficiency for evolution PDEs of order four or higher. It is shown that, as well as improving efficiency, it improves accuracy a little, compared to standard FEM with the same number of elements. A sensible progression is made through a sequence of practically interesting target fourth-order PDEs: fourth-order diffusion, surface diffusion and elastic gradient flow from the Willmore elastic energy functional.9 0