|
|
Evaluation of Sources of Image Media for Deepfake Creation
Andrýsek, Kryštof ; Malinka, Kamil (referee) ; Firc, Anton (advisor)
In the recent years, artificial intelligence has been on the rise and it is starting to become a normal part of our lives. This thesis focuses on the dangers associated with image deepfakes. Nowadays, it turns out that it does not only concern high-ranking political leaders or Hollywood stars, but can affect lives of any of us. This is due to, besides other things, the increasing number of applications with the help of which even a person without expert knowledge and skills can create a deepfake. The primary goal of this thesis is to evaluate the sources that can be used to obtain image media suitable for creating facial deepfakes, and to propose methods that can be used to evaluate the quality of these sources in terms of their appropriateness for the creation of deepfakes. A further aim is to identify groups of people who are at higher risk in relation to image deepfakes, which may help to improve prevention and general knowledge. The paper offers an analysis of the sources used to create facial deepfakes, proposes methods for evaluating their quality, a detailed evaluation of the sources using these methods, and an analysis of the vulnerability of various groups of people. Finally, a questionnaire survey, focusing primarily on social network sources, examines how many and what kind of photos are most frequently shared by ordinary users, and uses this data to refine information about at-risk groups.
|
guest ::
