Academic Literature On Deepfakes And Related Topıcs In Taiwan: A Structural Topic Modeling Analysis Of Emerging Research Themes

 Abstract

Since its emergence in 2017, deepfake technology has evolved from a niche innovation into a global concern with significant implications for politics, security, ethics, and privacy. Its ability to generate synthetic yet hyper-realistic content—including video, audio, text, and images—has made it a powerful tool for both creative applications and malicious activities such as disinformation, fraud, and sexual exploitation. Taiwan, which has been repeatedly targeted by deepfake-driven disinformation campaigns and non-consensual content, presents a particularly critical case for understanding how academia engages with the challenges posed by this technology. This study conducts one of the first systematic analyses of academic literature on deepfakes in Taiwan, examining the characteristics, evolution, and thematic focus of the research. Using Structural Topic Modeling (STM) and web-scraping techniques, 143 academic studies—including journal articles, master’s theses, doctoral dissertations, book chapters, and institutional reports—were analyzed to identify dominant research trends and their development over time. The results indicate that academic interest in deepfakes in Taiwan has grown rapidly since 2019, with the majority of publications written in Chinese. Ten major thematic clusters were identified, primarily focusing on detection algorithms, machine learning applications, and legal frameworks for regulating deepfakes. However, the analysis also revealed a relative lack of interdisciplinary studies addressing psychological and sociopolitical aspects—areas more prevalent in global deepfake research. Comparatively, Taiwanese scholarship demonstrates a strong emphasis on technological and legal countermeasures rather than on societal impacts or public perception. Overall, the study highlights Taiwan’s increasing but technically focused academic engagement with deepfakes and emphasizes the need for expanded cross-disciplinary collaboration. Strengthening policy-oriented, ethical, and sociotechnical research will be essential for developing comprehensive national strategies to mitigate the multifaceted risks posed by deepfake technology.

LINK: https://dergipark.org.tr/tr/pub/dasad/article/1797836

AI-driven fraud as an emerging cyber risk: Evidence from a global incident-based analysis

The rapid proliferation of artificial intelligence (AI) and deepfake technologies has introduced new and complex risks to individuals, companies, financial systems, and digital trust. While existing research has primarily examined deepfakes in sexual or political contexts, systematic analyses of AI-enabled fraud remain limited. This study addresses this gap by conducting an incident-based analysis of 167 documented cases of AI- and deepfake-enabled fraud worldwide between 2019 and 2025. Drawing on Cyber-Routine Activities Theory (C-RAT), the study examines temporal trends, victim targeting patterns, financial losses, and cross-national variations to assess how emerging AI technologies reshape opportunity structures for cybercrime. The findings reveal a sharp increase in AI-assisted fraud after 2022, coinciding with the public availability of generative AI tools. Victimization patterns shifted from companies toward individuals, while financial losses initially concentrated among companies before increasingly affecting individual victims. Country-level analysis highlights substantial variation, including evidence that targeted regulatory interventions can reduce exposure to AI-enabled fraud, as demonstrated in the People’s Republic of China. Overall, the results support C-RAT’s core assumptions regarding motivated offenders, suitable targets, and capable guardianship, while extending the theory to account for AI-driven cyber threats and systemic forms of guardianship. The study emphasizes that AI-enabled fraud represents a structural social risk inherent in modern digital infrastructures. Effective mitigation requires multi-layered strategies integrating technical controls, organizational investment in cybersecurity, and adaptive regulatory governance.

LINK: https://www.tandfonline.com/doi/full/10.1080/07366981.2026.2631066

SHIFTING NARRATIVES: A COMPARATIVE ANALYSIS OF ISLAMIC STATE OF IRAQ AND SYRIA (ISIS) SUPPORTERS' TWITTER HASHTAG USAGE IN TÜRKIYE BEFORE AND AFTER ITS TERRITORIAL DEFEAT

This study examines the evolution of hashtags used by Turkish supporters of the Islamic State of Iraq and Syria (ISIS) on Twitter, comparing the predefeat period (2014–2018) with the postdefeat period (2019–2022). The analysis encompasses 312,197 tweets to identify thematic shifts and network structures. In the predefeat phase, hashtags were primarily centered on geopolitical conflicts and political discourse, with prominent hashtags like #islamdevleti (Islamic State) and #incirlikkapatilsin (Incirlik Base must be closed). The network analysis revealed a main cluster of interconnected hashtags alongside smaller, niche groups. Postdefeat, the focus shifted towards humanitarian aid, financial support, and religious themes, with hashtags such as #sadaka (charity) and #infak (alms) becoming more prominent. This period exhibited multiple smaller, independent hashtag clusters, reflecting a more fragmented thematic focus. Despite these changes, core hashtags like #islamdevleti (Islamic State) and #hilafettakip (follow the caliphate) remained consistent, indicating enduring support for ISIS ideologies. The study underscores the adaptability of terrorist propaganda strategies in response to evolving circumstances and highlights the dynamic nature of online extremist activities. The findings contribute to a deeper understanding of how online platforms are leveraged by terrorist supporters to sustain and evolve their narratives over time.

Keywords: Twitter hashtags, terrorist propaganda, social media extremism, ISIS Turkish supporters, hashtag network analysis 

LINK: https://dergipark.org.tr/en/pub/ijshs/article/1853490

Deepfake Survey

I have prepared a survey as part of my postdoctoral research project at the Taiwan Foundation for Democracy. I would be very grateful for your participation. All participant information will remain strictly confidential and will not be disclosed under any circumstances. The results will not be used in any academic publication without prior ethical approval and will be used solely for the Taiwan Foundation for Democracy’s final report.

LINK: https://docs.google.com/forms/d/e/1FAIpQLSfrpkGSoYSb9lc68nKh5iHEJ8bkHabyKPQXvKvwegMhAtZRZQ/viewform?usp=dialog

Evaluating web-based deepfake detection tools: Risks, limits, and practical solutions for information security

ABSTRACT

The rapid advancement of artificial intelligence (AI) has facilitated the rise of deepfakes—synthetic audio, image, text, and video content generated by deep learning algorithms that can convincingly fabricate events, statements, or identities. While deepfake technology has positive applications in fields such as education and entertainment, it has predominantly been exploited for malicious purposes, including sexual exploitation, political manipulation, fraud, and disinformation campaigns. In response, web-based deepfake detection tools have emerged as accessible alternatives to algorithmic models, enabling non-experts such as researchers, journalists, and policymakers to verify suspicious content. However, systematic evaluations of these tools remain scarce in the academic literature. To address this gap, the present study examines 37 web-based deepfake detection tools. Employing descriptive and statistical analyses—including accuracy analysis, weighted score, and margin of error calculations—the study evaluates tool performance and categorizes them into high, moderate, and low accuracy groups. Results indicate that photo detection tools are the most reliable, with several achieving weighted scores above 80 percent, while video detection tools demonstrate only moderate effectiveness. Audio detection remains underdeveloped, showing significant weaknesses in both accuracy and tool availability. Hive Moderation stands out as the most versatile platform, though none of the tools offer comprehensive, multi-modal, or real-time detection. The findings highlight critical gaps in current detection capabilities, particularly in audio and in large, long-form media. This underscores the importance of scalable, multi-modal, real-time solutions to safeguard democratic institutions, organizations, companies, individual security, and public trust.

LINK: https://www.tandfonline.com/doi/full/10.1080/07366981.2025.2564783

Evolution and Focus of Terrorism Studies in Türkiye: A Comprehensive Analysis of Academic Literature (1992–2024)

The aim of this study is to analyze the evolution and characteristics of the academic literature on terrorism in Türkiye from 1992 to 2024. The study addresses two main research questions: the distinctive features of terrorism literature in Türkiye and the types of terrorist organizations that have been the focus of this literature, along with temporal changes in these focuses. In this direction, the study used descriptive and dictionary-based analyses on a dataset of 1066 academic publications identified through "Publish or Perish" software, focusing on studies with relevant keywords in Turkish. The descriptive analysis examined publication trends, number of citations, number of authors, and study types, while the dictionary-based analysis identified the most studied terrorist organizations and tracked changes over time. The findings reveal that most studies in this literature, which experienced significant growth after 9/11 and peaked in 2019, received fewer than ten citations. The literature predominantly focuses on organizations like the PKK, with shifting attention to groups like ISIS and FETÖ in recent years. This study contributes to a deeper understanding of how terrorism studies in Türkiye have developed and evolved over the last three decades.

LINK: https://dergipark.org.tr/tr/pub/ijshs/issue/93549/1736233

Understanding the Fear of Terrorism in Türkiye and Taiwan

Abstract 

This study investigates the heightened fear of terrorism in Türkiye and Taiwan, two countries with distinct histories of terrorism yet similar levels of fear, despite differing exposure to terrorist incidents. Utilizing data from the seventh wave of the World Values Survey, the research employs Ordinal Logistic Regression (OLR) to examine the effects of various media sources -television, newspapers, and the internet- on fear of terrorism in both countries. The findings indicate that while internet usage increases fear of terrorism in Türkiye, it has no significant effect in Taiwan. Conversely, exposure to television elevates fear of terrorism in Taiwan but not in Türkiye. Furthermore, the study underscores the influence of demographic and sociocultural factors, such as gender and the importance of religion, in shaping perceptions of terrorism. Women in both countries report higher levels of fear, and participants with stronger religious beliefs demonstrate increased fear. This research contributes to the existing literature by providing a comparative analysis of two geographically and politically distinct nations, highlighting the role of media and contextual factors in understanding public fear of terrorism. The findings offer valuable insights for policymakers in developing targeted interventions to address these fears.

Keywords: Taiwan, Türkiye, Fear of Terrorism, Ordinal Logistic Regression.

LINK: https://dergipark.org.tr/tr/download/article-file/4326872