Deep learning-based deepfake detection and cybersecurity: A study on public perception

With the rapid advancement of technology, AI-based deepfake technologies are adding a new dimension to cybersecurity threats today. Deepfake technology poses serious risks targeting countries, prominent individuals, organizations, and even individuals through social engineering methods. This new technology causes harms such as financial losses, loss of reputation, and creating fear and anxiety in society among businesses and individuals. This study aims to examine in detail the working principles, production processes, and different types of deepfake technology. This thesis contributes to the literature by addressing deep learning used in the production and detection of deepfake technology, deepfake detection methods, research conducted in this field, methodologies, and classification types. Additionally, it aims to evaluate the positive and negative aspects of this technology by presenting examples of the use of deepfake content in social and societal domains. Within the scope of this study, how deepfake technology works, the processes through which it is produced and disseminated, the types used for different purposes, and common detection methods are examined in detail. Furthermore, by focusing on the effects and potential risks of deepfake technology on human life, the societal and ethical dimensions of this technology are scrutinized. Another important goal of the study is to evaluate society's perception of deepfake technology. In this context, through a conducted survey, individuals' abilities to perceive deepfake videos were measured, and participants' thoughts on deepfake as a cybersecurity threat were analyzed in depth. The findings obtained were evaluated to understand how society perceives this new and significant technology and to contribute to future cybersecurity strategies. This study aims to contribute not only to understanding the technical aspects of deepfake technology but also to understanding its social and individual impacts.