ResearchSpace@Auckland
University of Auckland Research Repository - ResearchSpace
ResearchSpace is an online archive for the University of Auckland and contains full text theses and other research outputs.
University Staff: Update your Discovery Profiles and deposit full text content using Te Waka Huia Rangahau | Research Outputs.
Communities in ResearchSpace
Select a community to browse its collections.
Recent Submissions
Understanding Situational Factors in Human-centred Phishing Susceptibility
(ResearchSpace@Auckland, 2025) Zhuo, Sijie ; Russello, Giovanni ; Lottridge, Danielle
Phishing is a type of social engineering attack that deceives individuals into disclosing sensitive information or downloading malware. It poses a significant threat to both individuals and organisations, causing substantial financial losses. This thesis aims to enhance the understanding of human-centred phishing susceptibility by investigating the factors influencing users' vulnerability to phishing attacks.
The thesis begins by categorising phishing susceptibility factors into long-term stable variables, situational variables, and in-the-moment variables. This categorisation leads to the development of a Phishing Susceptibility Model (PSM) and highlights the research gap in exploring situational factors that influence users' email processing behaviour and phishing susceptibility.
To address this gap, we conducted a large-scale simulated phishing campaign to explore the impact of device type and link presentation on users' tendency to click on phishing links. Our study reveals that while the device used (mobile vs. computer) does not significantly influence susceptibility, masking the phishing link with hypertext significantly increases the likelihood of user clicks.
Additionally, we identified a novel research method in phishing susceptibility, the Precision Email Interaction Study (PEIS), to closely observe users' interactions with phishing emails in a controlled yet realistic environment. PEIS allows precise measurement of user interactions with phishing emails under various conditions. To support the PEIS, we developed the Precision Email Simulator, which provides customisation for various research needs.
Using PEIS, we conducted two user studies to explore the impact of workload on phishing susceptibility. We found that high workload is associated with shorter email reading times and an increased tendency to interact with task-relevant phishing emails. Additionally, paying attention to the email sender can significantly reduce phishing susceptibility.
This thesis concludes by discussing the implications of these findings for designing anti-phishing interventions and training programmes. By providing actionable insights into the human factors of phishing susceptibility, this work contributes to developing more effective strategies to mitigate the risk of phishing attacks. We also discuss future research directions that could build on our findings and further enhance email security.
The Impact of Structured Intensive Training on the Preliminary Image Evaluation Performance of Medical Imaging Technologists in Taranaki, New Zealand
(ResearchSpace@Auckland, 2024) Lewis, Kim E.; Mdletshe, Sibusiso; Doubleday, Andrea; Pieterse, Tracey
Introduction
Preliminary Image Evaluation (PIE) is an abnormality detection system where medical imaging technologists (MITs) write a comment outlining acute abnormalities they have detected on X-ray examinations performed in the emergency department (ED), or that no abnormality was detected. Initially developed in the United Kingdom due to a shortage of radiologists and a subsequent delay in radiology reports, PIE has spread to other parts of the world, notably in Australia, where a similar shortage of radiologists exists. Studies have demonstrated that training MITs in PIE improves their sensitivity, specificity, and overall accuracy. There are, however, concerns about whether this improvement is maintained over time. One suggestion is to provide ongoing feedback to MITs when participating in a PIE system to maintain this performance improvement. This pre-post-intervention study aimed to investigate the impact intensive training in PIE, along with regular feedback on commonly missed abnormalities, had on MIT’s sensitivity, specificity, overall accuracy, and comment accuracy when performing PIE on extremity X-ray examinations in EDs in two provincial hospitals in NZ.
Methods
This study was a pre-post-intervention design using image evaluation tests (IETs) to assess participants’ performance when detecting and describing acute abnormalities on acute extremity X-ray examinations in ED before, immediately after, and six months following a two-day intensive, structured training course in PIE. Additionally, during the six months following the intensive training, regular feedback consisting of fortnightly e-mails and monthly revision sessions was provided to assist participants in maintaining their post-training PIE performance.
Results
The participants in this study demonstrated an improvement in PIE performance immediately following intensive training. When re-assessed six months later, the participants' sensitivity was the same as their pre-intervention sensitivity. However, the participants demonstrated increased specificity, comment accuracy, and overall accuracy. Additionally, the clinical evaluation demonstrated an improvement over six months.
Conclusion
As the participant's overall accuracy improved and was maintained for six months following intensive training, it is recommended that MITs undertake additional training to improve PIE performance before participating in a PIE system. Additionally, regular feedback on commonly missed abnormalities may be useful to ensure participants' improved performance when performing PIE is maintained over time.
Object-based Attention across the Meridian and Under Load: Evaluating the Boundary Conditions of Object Effects
(ResearchSpace@Auckland, 2024) Mutemeri, Kai ; Erb, Christopher
This study examines how working memory load (WML) impacts object-based attention by adapting Bleckley et al.'s (2015) hypothesis that the availability of working memory resources determines the ability to switch between object-based and space-based attention flexibly. Participants' gaze and reaction times (RT) were analysed across load and no-load conditions using the two-rectangle cueing task and eye-tracking measures. Results showed robust space-based effects, with faster RTs on valid trials. Participants were quicker on within-object trials than between-object trials, aligning with the known object-based attentional effects. However, object-based effects weakened under WML, with RTs slower overall and no significant difference between within- and between-object trials.
Notably, the position of the target relative to the cue influenced attentional shifts: In the no-load condition, strong object effects emerged on trials requiring horizontal attentional shifts across the meridian, while vertical shifts showed weaker object effects. Under load, object effects diminished in horizontal shift trials, while vertical shift trials displayed a small, inverse pattern with faster RTs on between-object trials. Gaze data reflected the same pattern of results, with faster fixations on trials requiring a horizontal scan relative to those requiring a vertical scan.
Critically, there was wide variation among individuals. Participants with more eye movement across the display had slower RTs, suggesting a reliance on visual scans rather than peripheral attention. Correlational analyses further indicated that some participants' object effects stemmed from slower responses on between-object trials rather than fast responses on within-object trials. These findings reinforce Bleckley et al.'s hypothesis while introducing new insights about the conditions under which object-based attention is deployed.
The Systematisation of Safety in Design Implementation in the New Zealand Construction Industry
(ResearchSpace@Auckland, 2024) Shi, Weifang; Chang-Richards, Alice; Ingham, Jason
Compared to other industries, the construction sector performs poorly in health and
safety, with a staggering number of injuries and fatalities. A significant percentage of
accidents were closely linked to a lack of consideration for ‘Safety in Design’ (SiD) and
its implementation. The gaps lie in the unrevealed SiD factors, dynamics interactions,
SiD maturity metrics, safe design knowledge shortage, and effective methods to educate
designers. This research aims to develop a comprehensive framework for effectively
and strategically implementing SiD. The research explored factors influencing SiD and
investigated strategies advancing its implementation and improving workers’ health and
safety through design activities. This study identifies five aspects of implementing SiD:
(1) Critical factors affecting SiD implementation.
(2) Motives within system dynamics interactions.
(3) Leading indicators measuring the SiD maturity.
(4) Significant design features mitigating hazards.
(5) A Conceptual framework including safe design practice training.
This research applied in-depth interviews with participants with SiD experience in New
Zealand and this research proceeded through:
(1) Grounded theory to reveal unearthing underpinned theories.
(2) System dynamics analysis to investigate inherent motives and interactions.
(3) Rigid programmatic triangulation method to decide leading indicators.
(4) Computational methods in finalising significant design features.
(5) Prototype integrating the elements of high technologies.
The results suggested that the potential factors were attributed to a theoretical
framework, including critical factors, challenges and opportunities. The identified
leading indicators are legislation change and designers’ knowledge, while significant
design features mitigating highly lethal hazards were investigated in the built
environment, such as accessibility and spatiality.
This research added significant value to theory and practice:
(1) An analytical framework that includes research trends and potentials guiding
future research.
(2) A theoretical framework initiated with unified terminology to guide decisionmaking.
(3) A system dynamics model revealing motives and interactions.
(4) A programmatic method determining leading indicators.
(5) The computational method pioneered the generation of significant design
features.
(6) A framework conceptualised in safe design training with explicit development
processes outlined in the prototype.
(7) A comprehensive framework encompassing enabling factors and strategies
was developed.
Evaluating the feasibility of Familial Hypercholesterolaemia Case Ascertainment Tool (FAMCAT) in NZ
(ResearchSpace@Auckland, 2024) Al-Samarrai, Zahra'a Taha; Harrison, Jeff
Aim/Background: Familial Hypercholesterolaemia (FH) is one of the most prevalent inherited autosomal dominant disorders, characterised by impaired lipoprotein metabolism which leads to elevated cholesterol levels. This elevation increases the risk of premature coronary heart disease (CHD) and mortality. Early identification and treatment are essential to increase life expectancy, however FH remains underdiagnosed globally. To overcome this hurdle, this project aims to explore the feasibility of applying the Familial Hypercholesterolaemia Case Ascertainment Tool (FAMCAT) to facilitate the identification of FH.
Methods : We first reviewed strategies for improving detection of FH both internationally and within New Zealand, along with the accuracy of these screening algorithms. Following this, we retrospectively applied FAMCAT1 and FAMCAT2 regression equations and the Dutch Lipid Network Criteria to a cohort of over two million individuals aged 16 years and older, using routinely collected laboratory data from 2007 to 2020.
Findings: The scoping review concluded any form of screening improved FH detection rates, though a number of issues such as inconsistent LDL-C thresholds applied, cohorts varying in their sources, and low participation made the studies difficult to generalise. Incomplete or incorrectly coded patient records were commonly cited issues. Algorithm accuracy results were mixed, with most studies concluding the FAMCAT algorithms were the most effective tools for screening electronic health records to identify high-risk patients. Applied to a New Zealand cohort of 126,747 of patients, assuming a FH population prevalence of 1 in 500, FAMCAT 1 classified 2.9% as likely FH (1 in 35 of cases screened). FAMCAT2 identified 2.7% (1 in 37 of cases screened) who warrant further investigation and potentially genetic testing.
Conclusion: We have demonstrated that applying FAMCAT to routinely collected health data is a feasible method for identifying individuals that may be suitable for triage and further screening. Further research is needed to validate FAMCAT2 in a New Zealand population, undertake a feasibility study of targeted screening (with cascade screening for positive cases) and a health economic evaluation.