Article Review 1
Introduction
This review examines Bray, Johnson, and Kleinberg’s (2023) study, Testing human ability to detect ‘deepfake’ images of human faces, published in the Journal of Cybersecurity. The study explores whether people can distinguish AI-generated deepfake images from real human faces and whether simple interventions can improve detection accuracy. This topic is crucial to cybersecurity because deepfakes represent a rapidly growing digital threat that exploits human psychological weaknesses and social trust.
Relation to Social Science Principles
The study connects to several social science principles:
- Human perception – It explores how individuals process facial cues and make judgments under uncertainty, reflecting psychological principles of perception and decision-making.
- Behavioral sociology – It highlights how people adapt to new technological threats, influencing trust and interaction in online environments.
- Criminology and cybercrime – The misuse of deepfakes for fraud, scams, and misinformation aligns with crime opportunity and victimization theories in criminology.
Research Questions, Hypotheses, IV & DV
The authors addressed three primary research questions:
- RQ1: Can participants detect deepfake images above chance levels?
- RQ2: Do interventions (familiarization, advice, reminders) improve detection accuracy?
- RQ3: Does participant confidence correlate with accuracy?
Hypotheses: Interventions would improve accuracy, and participant confidence would correspond with correctness.
Independent Variables (IVs): Experimental condition (control, familiarization, one-time advice, advice with reminders).
Dependent Variables (DVs): Accuracy in labeling images (real vs. deepfake), reported confidence, and reasoning strategies.
Research Methods
The study used an experimental online survey with 280 participants recruited via the Prolific platform. Participants were randomly assigned to one of four groups: a control group or one of three intervention groups. Each participant was shown 20 images (from 50 real and 50 AI-generated human faces) and asked to identify each image as ‘real’ or ‘AI-generated,’ along with rating their confidence on a 10-point scale.
Data and Analysis
Data collected included binary image classifications, confidence ratings, and qualitative explanations. Statistical tests such as t-tests and ANOVAs were used to analyze accuracy, while qualitative data were coded using NVivo to identify reasoning strategies.
Findings: Participants averaged about 62% accuracy, only slightly better than chance. Interventions did not significantly improve results, and confidence remained high despite low accuracy. These results suggest that human intuition is unreliable for detecting AI-generated images.
Relation to Class Concepts
This research connects to multiple class concepts:
- Trust in digital environments – It reflects the challenge of verifying authenticity in digital contexts.
- Human factors in cybersecurity – Deepfake detection mirrors issues of user vulnerability, similar to phishing.
- Critical thinking and decision biases – The study highlights overconfidence bias, where individuals trust their judgment more than evidence supports.
- Policy implications – It supports the idea that defending against digital deception requires education and behavioral change alongside technical tools.
Marginalized Groups and Social Impact
Deepfake technology disproportionately impacts marginalized populations. Women and minority groups face higher risks of non-consensual sexual deepfakes, harassment, and reputational damage. Such misuse exacerbates social inequalities, emphasizing the need for stronger ethical, legal, and technological protections for vulnerable communities.
Contributions to Society
This study contributes to society by:
- Demonstrating the limitations of human perception in recognizing AI-generated content.
- Showing how overconfidence can amplify misinformation risks.
- Highlighting the need for interdisciplinary solutions that combine technical detection tools with social awareness and education.
Conclusion
Bray et al. (2023) reveal that humans perform only slightly above chance when identifying deepfakes and that confidence does not correspond with accuracy. The study underscores an urgent cybersecurity and social challenge: human perception cannot keep pace with rapidly advancing AI manipulation. Addressing deepfakes requires not only technological innovation but also societal adaptation to strengthen digital literacy and trust.
References
Bray, S. D., Johnson, S. D., & Kleinberg, B. (2023). Testing human ability to detect ‘deepfake’ images of human faces. Journal of Cybersecurity, 9(1), 1–18. https://doi.org/10.1093/cybsec/tyad011
Article Review 2
Introduction / BLUF
This article, Controlling Cyber Crime through Information Security Compliance Behavior, examines how organizational culture, cybersecurity awareness, employee engagement, and trust in management shape employees’ compliance with security policies. BLUF: The study finds that organizational and psychological factors strongly influence compliance, and trust in management plays a key mediating role in how culture and awareness shape secure behavior.
Relation to Social Science Principles
This study reflects several social science principles including human behavior, organizational culture, and social influence. Human behavior is central because compliance is driven by employees’ psychological perceptions, motivation, and awareness. Organizational culture represents a major social science principle concerning shared norms, values, and social expectations within a group. Social influence and social learning theories help explain how employees model behaviors from leadership and peer norms. Together, these principles show that cybersecurity is as much a social and behavioral issue as it is technical.
Research Question / Hypotheses / Independent & Dependent Variables
Research Questions:
- How does organizational culture impact information security compliance?
- To what extent does cybersecurity awareness influence compliance behavior?
- Does employee engagement moderate these relationships?
- Does trust in top management mediate these effects?
Hypotheses:
The study proposes six hypotheses predicting that organizational culture, cybersecurity awareness, employee engagement, and trust in management each impact compliance directly or indirectly.
Independent Variables:
• Organizational culture
• Cybersecurity awareness
Moderators / Mediators:
• Employee engagement (moderator)
• Trust in top management (mediator)
Dependent Variable:
• Information security compliance behavior
Types of Research Methods Used
The study uses a quantitative research design. Data was collected from 261 employees across various production companies using validated questionnaires. The researchers applied Structural Equation Modeling (SEM) to test relationships among the variables and evaluate the overall model.
Types of Data and Analysis Used
Data consisted of Likert-scale survey responses measuring culture, awareness, engagement, trust, and compliance behavior. Following data collection, the authors performed Confirmatory Factor Analysis (CFA) to test validity and reliability, followed by SEM path analysis to test hypotheses. Reliability tests (Cronbach’s Alpha), model fit indices (CFI, RMSEA, SRMR), and correlation tables strengthened the statistical rigor.
Connections to Course Concepts
This study directly aligns with course concepts such as the social science view of cybersecurity, organizational behavior, trust theory, and human factors. It reflects how non-technical factors influence risk, echoing class discussions about behavior being the weakest link in cybersecurity. The findings also connect to the Theory of Planned Behavior, covered in lectures, by showing that attitudes, norms, and perceived control shape compliance. Leadership’s role in shaping norms ties to modules on power, influence, and culture.
Connections to Concerns or Contributions of Marginalized Groups
While the study does not directly analyze marginalized groups, its themes relate to their experiences in organizational environments. Marginalized employees may experience lower trust in management or feel excluded from security decision-making, influencing compliance behavior. Additionally, unequal access to training or psychological safety can affect awareness and engagement levels. Understanding these dynamics helps ensure cybersecurity programs are equitable and inclusive.
Overall Societal Contributions / Conclusion
This study contributes to society by demonstrating that cybersecurity compliance is not just technical—it is fundamentally social and behavioral. By showing that trust, culture, and engagement significantly influence compliance, the article provides guidance for organizations seeking to strengthen security through people-focused strategies. Developing supportive cultures and trustworthy leadership improves employee adherence to security protocols, reducing cyber risk and enhancing organizational resilience.
Reference
Ghaleb, M. M. S., & Pardaev, J. (2025). Controlling cyber crime through information security compliance behavior: Role of cybersecurity awareness, organizational culture, and trust in management. International Journal of Cyber Criminology, 19(1), 1–26.