With cyber-security incidents constantly in the news, small businesses are becoming increasingly aware of the risks of an online presence. However, a gap exists with small business remaining slow in adopting cyber-security practices. Their lack of confidence correlates with the social theories of Protection Motivation Theory and Theory of Planned Behaviour. Here we explore this awareness-action gap, focusing on small business cyber-security analysis. Drawing out the cybersecurity considerations using Australasian non-cyber-security case studies shows that an average small business needs to consider over 50 items for their cybersecurity analysis. We find that small businesses tend to focus on business tasks, data and external groups/tools rather than core cyber-security concepts such as operating systems or threat details. Our research will help guide the communication of cybersecurity principles to small business, incentivising adoption of cybersafe practices.
Cyber violence against women has become a critical issue in the digital age, which includes threats such as cyberstalking, online harassment, doxxing, and non-consensual image sharing. These incidents often result in psychological distress, reduced digital participation, and a loss of confidence in online safety. As women increasingly engage in digital spaces, they face gender-specific risks that demand targeted interventions. This study investigates the types and consequences of cyber violence experienced by women, their level of cybersecurity awareness, and the factors influencing their confidence in online protection. The research employs quantitative methodology for data collection and analysis. A quantitative survey was conducted with 30 women aged 18-35, analyzing their experiences, awareness, and behavioral responses. Descriptive analysis revealed that over half of the participants had encountered cyber violence, yet most did not seek legal or psychological support. Inferential analysis showed a significant relationship between cybersecurity familiarity and digital self-confidence, while demographic factors and victimization history were not significant predictors of cyber violence. Participants also expressed strong interest in learning to secure social media and detect phishing, with a preference for interactive learning formats. These findings highlight the need for practical, accessible digital literacy programs tailored to women's needs. The study calls for a need for a dedicated platform offering cybersecurity education, legal awareness, and community support to empower women and promote safe digital engagement, aligning with global gender equity goals.
Immersive Virtual Reality (VR) enables the creation of ecologically valid environments that can reliably elicit affective and behavioral responses, making it a promising medium for clinical interventions. Within the ARCADIA project, the Enhanced Food Exposure (EVE) protocol leverages VR cue exposure to target craving, emotional dysregulation, and context-specific triggers central to eating disorders. Building on traditional cue exposure therapy, EVE integrates real-time physiological monitoring, emotion-regulation strategies, and naturalistic handtracking interactions to enhance ecological fidelity and therapeutic engagement. This pilot study evaluated usability, realism, and tolerability of the VR tool through a dual-perspective approach involving a patient with bulimia nervosa and an eating disorder specialist. Both participants reported moderate- to-high presence and realism, minimal cybersickness, and excellent usability (SUS scores: 95.0 and 87.5). Qualitative accounts underscored the emotional salience of specific foods and contexts, confirming the tool's ability to evoke authentic craving responses. At the same time, both users identified refinements to improve ecological validity, such as enhanced food models, smoother object manipulation, and simplified rating scales. The findings support the feasibility and face validity of the EVE scenario as a clinically relevant addition. By demonstrating comfort, intuitiveness, and credibility of the exposure scenarios, this work establishes essential human-factor criteria to advance toward controlled trials. More broadly, it demonstrates how VR can combine ecological realism and therapeutic precision to strengthen exposure-based treatments for complex eating disorders.
The development of digital technologies enhances the studies about social interaction, e.g., cyberaggression and cybervictimization. Cyber-victimization (CV) is defined as a process in which an individual is victimized by negative actions in virtual space, e.g., cyberstalking, cyber fraud, and other types of cybercrime. Most of the previous studies were focused on the analysis of foundations of CV without systematic analysis of the psychological risk-factors. The aim of the current study was to indicate the role of psychological risk-factors of CV in the Russian context. In Study 1, 506 participants (248 female) completed the Online Victimization Scale (OVS) (e.g., individual's experience of general, sexual, individual national, and indirect national CV) and socio-demographic characteristics. The 4-factor model showed good fit indices (RMSEA =.062 [0.050; 0.075], TLI = 0.955, chi(2) (166) = 71, p <.001). The results of correlation analysis show the presence of a connection between the CV subscales and psychological variables, e.g., perceived stress, subjective well-being, and negative affect. The OVS shows invariance across gender, age, and education. In Study 2, 457 participants (243 female) took part in the online survey. Participants completed the measures of OVS, perceived stress, self-esteem, positive and negative affect. The results show that only negative affect positively contributes to the CV, whereas perceived stress - indirectly, through the negative affect. Self-esteem has no significant direct or indirect effects on the CV. The practical issues of CV detection are discussed.
Supporting adaptive functioning in individuals with neurodevelopmental conditions (NDCs) is a critical issue for society. People with NDCs, as well as their families, often experience unique challenges in adaptive functioning that can impact their quality of life, including socialisation, communication and self-care. These challenges necessitate personalised interventions that are tailored to an individual's distinct needs and abilities, however, information on these interventions is limited. Artificial intelligence (AI)-assistive technologies may offer a solution, as they are efficient, cost-effective, and capable of being highly personalised. To investigate this, The University of Sydney partnered with akin Technologies (R) to develop a support tool known as The Pixi Home-Hub ('Pixi'), to examine whether it can assist with families' adaptive functioning needs as they occur in real-time. Pixi was evaluated over a two-week period using experience sampling methodology (ESM) and semi-structured interviews. The current article presents results from this ongoing evaluation from 13 participants with children aged between 2.8-11.6 years. Overall signal response rate (AI administered vs responded) was 63.5%. Pixi users reported positive experiences using Pixi but provided recommendations on how it could be further tailored to their and their family's needs, such as wellbeing check-ins, automatic/adaptive goal generation and integrating a platform for their child's health information. Parents emphasised their appreciation of having the device ask them about their own wellbeing. This research emphasises that AI-assistive technology holds significant potential to support adaptive functioning for people living with NDCs in their everyday environments.
Deepfake technology poses a significant threat to digital security, enabling the creation of highly realistic fake images and videos that can be used for malicious purposes. This study proposes a deepfake detection system leveraging a Vision Transformer (ViT) model, demonstrating superior performance in image classification tasks. The system allows users to upload an image for classification, predicting whether the image is real or fake with an associated confidence score. The model was trained using the Kaggle DeepFake and Real Images dataset, employing preprocessing techniques such as normalisation and augmentation to enhance generalisation. Various data split ratios (60:40, 70:30, 80:20, and 90:10) were analysed to determine optimal model performance. Experimental results demonstrate that with the 90:10 split, the fine-tuned ViT model achieves a remarkable 99.28% accuracy. Comparative analysis against traditional CNN-based models further validates the effectiveness of the ViT model, demonstrating superior accuracy and robustness in deepfake detection, particularly in large-scale datasets. Additionally, the system incorporates a user-friendly single-image prediction functionality, allowing users to upload an image and receive a classification label (real or fake) along with a confidence score. The study underscores the viability of Vision Transformers for deepfake detection, addressing challenges such as computational efficiency and real-time implementation.
This research explores the impact of social engineering attacks, focusing on significant financial losses and psychological consequences faced by the victims. This study aims to provide an analysis of how social engineering attacks affect individuals psychologically and financially. The study aims to develop a conceptual framework for mitigating the impact of social engineering attacks. The research employs a mixed-methods approach, combining qualitative and quantitative analysis. Statistical analyses such as one-way ANOVA, Independent T-test, and Mann-Whitney U test are used to explore the correlation between demographics (age, education, and work position), cybersecurity awareness, and security behaviours associated with password management, privacy concerns, and security consciousness. The findings reveal that social engineering attacks can have a notable psychological impact on the victims, which manifests as trust issues, anxiety, and fear, and results in financial losses due to fake investment schemes, financial fraud, and identity theft, which mainly affect bank accounts and credit cards. Statistical analyses reveal that individuals with low cybersecurity awareness, particularly younger and elderly populations, are more susceptible to these attacks. In response, this study proposes a conceptual mitigation framework. The framework aims to enhance cybersecurity awareness, thereby reducing the risks associated with social engineering. This research highlights the need for targeted educational programs to promote adaptable security practices and safeguard vulnerable individuals.
This experimental study compares the diagnostic decisions made by radiologists and an artificially intelligent model (EfficientNetB0) in identifying breast cancer via digital mammography. The results demonstrate an overall accuracy of 95.06% for the AI model. Correlation analysis reveals significant differences in agreement levels between radiologists themselves versus their concordance with the AI model, particularly within crucial BIRADS categories (III-V). These findings highlight the need for further investigation into harmonizing expert opinions and AI-based approaches to reduce diagnostic uncertainty and minimize risks for patients at risk of breast cancer.
This systematic literature review examines the intersection of romance scams and cybersecurity awareness in Asia. The study investigates the prevalence and impact of romance scams in the region, analyzing the tactics employed by scammers and the vulnerabilities they exploit. It also explores the potential for empowering victims of romance scams to become advocates for cybersecurity awareness, thereby contributing to a safer online environment. Through a comprehensive search of relevant literature, this review synthesizes key findings, identifies research gaps, and proposes recommendations for targeted interventions and policy changes. The study emphasizes the importance of culturally sensitive awareness campaigns and collaborative efforts between stakeholders to effectively combat romance scams and enhance cybersecurity awareness in Asia. The insights gained from this review can inform the development of robust strategies to protect individuals from cyber threats and foster a resilient online community.
Body image is shaped by internal self-perceptions and external influences which can reinforce unrealistic beauty ideals and amplify self-discrepancies. This study employed MySelf, a desktop tool based on self-discrepancy theory, to calculate three indices: Body Image Social Attribution ( BI-SA), Body Image Social Desirability (BI-SD), and Body Image Discrepancy (BI-D). We investigated whether these indices predicted body image-related outcomes and if social media engagement shaped such self-perceptions. Multivariate multiple regressions showed that only BI-SA and BI-SD, indices reflecting perceived social evaluations, were significantly associated with body esteem and body appreciation, while BI-D was not. Additionally, social media variables were also linked to BI-SA and BI-SD. Specifically, engaging in social media activities involving friends' content was a positive predictor of interpersonal discrepancies (BI-SA and BI-SD), while Instagram appearance comparisons were negatively associated with them. These results may suggest that externalized and interpersonal self-discrepancies rooted in how individuals believe they are seen by othes, are more salient predictors of body image concerns than internal ideals. These findings support theories such as the Tripartite Influence Model, Objectification Theory, and the Looking-Glass Self Theory, highlighting the increasing social nature of bodily self-evaluation. Interventions aiming to improve body image should move beyond personal self-esteem enhancement to address social comparison dynamics and external validation processes. The MySelf tool offers a valuable means for assessing these multidimensional constructs and has potential applications in both clinical and preventive settings.
High-profile cyber-attacks continue to highlight the growing sophistication of malicious hacking. While much attention has been paid to human error on the defensive side, attackers are also subject to cognitive limitations. Recognising this opens a strategic opportunity: rather than viewing attackers solely as technical threats, we can exploit predictable human biases to reduce their effectiveness. One such bias is the representativeness heuristic, where individuals assess the likelihood of events based on how well they align with a mental prototype, often ignoring relevant statistics. This can lead attackers to make poor judgments under uncertainty. As part of a project funded by the Intelligence Advanced Research Projects Activity (IARPA), our multinational research team has developed and tested a range of interventions that expose skilled participants to cognitive biases during complex cyber tasks. Early findings suggest that certain patterns of behaviour-particularly under conditions of uncertainty-may predict susceptibility to bias and influence performance in realistic attack scenarios. These results indicate that even highly capable adversaries are vulnerable to subtle psychological manipulations. By incorporating these insights into cybersecurity systems-using ambiguous patterns, misleading cues, or strategically framed information-we can reduce attacker efficiency and disrupt decision-making. This psychological approach, when integrated with existing technical safeguards, offers a novel and scalable method for strengthening cyber-defence and protecting systems from emerging threats.
We investigated whether religiosity, political partisanship activism on the Ukraine-Russia conflict, decision making style, and time taken to complete an online Questionnaire predict willingness to forward the online Questionnaire to others. The Questionnaire study was conducted May - September 2024 on Qualtrics with an online sample of N=204. The online Questionnaire measured political support for Ukraine, Russia, or neither in the Ukraine-Russia War, decision making style on the Melbourne Decision Making Questionnaire (MDMQ), political activism, religiosity, and willingness to forward the Questionnaire. We found more support for Ukraine (73.5%) than for Russia (3.9%), than for neither side (22.5%). Religiosity was a factor in willingness to forward the Questionnaire. There was a tendency for religious participants to be more willing to forward the Questionnaire. A 2x2 MANOVA found that politically active participants were more willing to forward the Questionnaire and religious participants with a less vigilant decisional style were more willing to forward the Questionnaire. We measured time taken to complete the online Questionnaire. Participants neutral in the Ukraine-Russia conflict and who spent more time on the Questionnaire were more willing to forward the Questionnaire. The findings suggest people who are less vigilant, have stronger political viewpoints, and take more time completing Questionnaires, are more likely to spread information and accordingly might pose greater security risks in organisations.
Digital health technologies present a "comfort-growth paradox," a fundamental tension between the need for systems to be both predictable for user comfort and disruptive for therapeutic growth. This paper argues that cyberpsychology provides the essential conceptual tools to navigate this challenge. Technologies must offer psychological safety to build trust and reduce cognitive load, yet true behavioral and cognitive development requires engaging with challenges that stretch users' capabilities. The very comfort that fosters engagement can inadvertently constrain progress if it fails to introduce necessary, growth-oriented discomfort. To resolve this, the paper draws on the technological adaptation of psychological theories such as Flow Theory, the Transtheoretical Model of Change, Presence and Embodiment, and Self-Determination Theory. These models inform practical implementation strategies, including progressive scaffolding to gradually reduce support, enhancing user agency through customizable challenges, providing multiple modalities of support, and fostering metacognitive awareness of the comfort-growth dynamic. By integrating these approaches, designers can create adaptive systems that balance comfort and challenge in real-time. The ultimate goal is to develop technologies that serve not just as tools for managing illness but as partners in human flourishing.
The digital age has significantly reshaped beauty standards, particularly among young adults who are immersed in social media culture. This qualitative study explores how Malaysian undergraduates make sense of beauty in today's image-saturated online world. Semi-structured interviews were conducted with 15 purposively sampled undergraduates (aged 18-24) from a Malaysian university. Thematic analysis revealed four overarching themes: (1) Physical Perfection and Aesthetics, (2) Inner Qualities and Self-Worth, (3) Social Media Portrayals and Filters, and ( 4) Resistance and Inclusivity. Participants perceived beauty as complex while physical appearance remained prominent, inner attributes such as confidence and kindness were equally valued. Social media platforms, particularly influencers using highly edited images, were reported as powerful but often unrealistic sources of beauty ideals. Nevertheless, many students demonstrated critical awareness and coped through resisting unattainable standards, advocating for more inclusive representations. These findings underscore the importance of critical media literacy interventions aimed at empowering young adults to navigate digital beauty cultures and promote healthier self-perceptions. Implications include how universities and health educators should integrate critical-media-literacy modules that demystify editing tools and highlight algorithmic biases, while counselling services can incorporate self-compassion training to bolster intrinsic self-worth.
In today's digital era, the rapid integration of advanced technologies into everyday life across Southeast Asia coincides with a growing mental health crisis, marked by increasing rates of loneliness, depression, and suicidal ideation. In response, digital mental health platforms powered by artificial intelligence (AI) including chatbots, mood-prediction engines, and risk-detection models have emerged as scalable, stigma-free, and on-demand solutions, especially in the aftermath of the COVID-19 pandemic. These tools offer new opportunities for extending mental health care to underserved populations, particularly in lowresource settings. Yet, while AI-driven systems hold promise for enhancing diagnostic precision, enabling personalised interventions, and supporting continuous treatment monitoring, they also introduce acute ethical concerns. These include algorithmic bias, opaque "black-box" decision-making, deficits in accountability, and significant risks to privacy, challenges that are further complicated by Southeast Asia's diverse cultural norms, linguistic variation, and uneven levels of digital literacy. This paper critically examines the ethical integration of AI in digital mental health therapy within the Southeast Asian context. It outlines five core ethical challenges, accountability, transparency, fairness, privacy, and empowerment and maps the emerging policy landscape, including voluntary regional frameworks and national strategies from different countries. To advance these goals, the study proposes a practical five-pillar ethical framework to guide responsible AI development and deployment. By embedding these principles throughout the AI lifecycle, the paper aims to support the creation of mental health tools that are not only technically effective, but also socially responsible and culturally sensitive.
Although extensive literature addresses the technical components of scam disruption, there are significant gaps in analyzing scam-related cybercrime from both a technical and psychological perspective. Failing to consider these aspects equally has resulted in disjointed approaches to scam detection that are less effective than a holistic, transdisciplinary one. Aiming to fill this gap, this report conducts a bottom-up survey of research into scam disruption. In cross-analyzing the current state, challenges and emerging research in these fields, it became clear that most of the challenges in scam disruption stem from two over-arching issues. Firstly, practitioners require novel AI integrations to automate data collection and interpretation. Secondly, transdisciplinary collaborations are mandatory for effective scam disruption ontologies. In light of these findings, a novel scam disruption ontology is proposed. By merging recent innovations in AI and blockchain with expert cyber psychology knowledge, a novel architecture can be developed to assist with autonomous scam detection. Utilizing such technologies lowers the technical knowledge threshold for developing software and helps facilitate transdisciplinary approaches to combat online scams. This way, psychology practitioners can work more closely with software developers to create large-scale scam detection models based on targeted behavioral analysis. This research paves the way for more effective crime fighting ontologies through transdisciplinary development.
The integration of AI into decision support systems raises concerns about overreliance and distrust. To address this, we propose an experimental protocol combining Learning to Defer (LtD)-where AI delegates decisions to humans when appropriate-and Explainable AI (XAI), which provides users with decision rationales. Our study investigates how these approaches impact human decision-making, particularly in high-stakes contexts. Participants will classify noisy images from ImageNet under three between-subjects conditions: Defer (AI defers to user), Defer + XAI (AI provides an explanation), and Hidden Delegation (AI involvement is concealed). Each condition will be tested in neutral and high-stakes scenarios, the latter framed through narratives emphasizing the danger of misclassification. We will assess decision accuracy and reaction times, as well as psychological measures that explore the influence of individual differences (i.e., intolerance to uncertainty and cognitive styles), and emotions (e.g., emotion regulation, and AI-related anxiety). We hypothesize that Defer may prompt more analytical thinking, improving accuracy over Hidden Delegation, while Defer + XAI may further enhance performance. In contrast, Hidden Delegation could promote reliance on intuitive processing. We expect higher accuracy and longer response times in high-stakes conditions. Findings will inform the design of human-AI systems that optimize user engagement and reliability, particularly in domains like clinical decision-making.
Cognitive impairment is a common feature of schizophrenia, particularly affecting attention, memory, and executive functions. Early identification and remediation of cognitive impairments can enhance outcomes and well-being. This study investigates the efficacy and specificity of a novel training program based on a Virtual Reality environment aimed at remediating visuospatial memory, focused and sustained attention, working memory, and reasoning skills. Twenty-eight individuals diagnosed with schizophrenia were enrolled in a randomized controlled trial and allocated to either an experimental group and a control group. Pre- and post-intervention assessments measured cognitive performance. Results suggest that VR-based cognitive training offers a promising, immersive approach to cognitive rehabilitation.
This exploratory study investigates the effects of context, anxiety, and emotional valence/intensity on the sense of presence in young adults, using a contextual and multimodal anxiety assessment protocol. It also explores the contribution of physiological data to complement self-reported psychological measures. Thirty participants (aged 19-30) take part in four sessions: an initial assessment and three VR exposure sessions (soothing, neutral, anxiety-inducing), each spaced 15 days apart. Environments are individually selected based on participants' subjective experiences to ensure personal relevance. During each session, participants complete psychometric assessments directly in VR (BFI-10, STAI, HADS, ITC-SOPI) and rate their emotions (Likert scales) post-immersion. Cardiac activity is continuously recorded using a portable PPG device (EmbracePlus). Wilcoxon tests reveal significant differences in declared anxiety and perceived emotional valence across contexts, with higher anxiety in the anxious session and greater presence and engagement in the soothing session. Regression analyses show that emotional valence and intensity significantly predict dimensions of presence (spatial presence, natural, engagement), though with low coefficients. STAI is negatively correlated with engagement. No significant session effects correlations with physiological data emerge. However, at the individual level, presence, emotions, anxiety, and HRV show variable links, suggesting interindividual differences. Preliminary analyses also reveal associations between personality and physiology, with neuroticism negatively and extraversion positively related to HRV. The results suggest that the sense of presence is modulated by context and emotional state. These findings argue for a dynamic, user-centered model of presence that integrates emotional states, context, physiological measures and individual variability.
The evolving landscape of cyber threats necessitates innovative approaches to detecting and mitigating malware. This study presents a comprehensive machine learning-based framework for malware detection, integrating various feature analysis, selection, and modeling techniques to enhance the detection capabilities. By leveraging a combination of data from Kaggle and MalwareBazaar, the dataset used in this research provides a diverse representation of malware and legitimate files. Feature engineering and correlation analysis were performed using the K-Nearest Neighbors, Extra Trees, Bagging Classifier, and Hist Gradient Boosting, identifying the four most significant features for malware detection. The framework incorporates multiple machines learning algorithms, including Extra Trees, Bagging Classifier, Hist Gradient Boosting, and K-Nearest Neighbours, to evaluate their effectiveness in malware classification. The performance of each model is assessed based on metrics such as accuracy, precision, recall, F1-score, false positive rate, and ROC-AUC. The implementation of this framework also includes a user-friendly interface that allows for effective integration of threat intelligence from platforms such as MalwareBazaar and MITRE ATT&CK. This work aims to contribute to the development of a robust and adaptive malware detection system that can handle emerging threats and provide insights into their behavior. The results demonstrate the significant potential of integrating machine learning with threat intelligence, showcasing how these technologies can complement each other to improve cybersecurity defenses.
