Demystifying Deepfakes: An In-Depth Analysis of Technical Possibilities and Societal Implications
Schenk, Robert Ville (2024)
2024
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-202402273498
https://urn.fi/URN:NBN:fi:amk-202402273498
Tiivistelmä
This research-based thesis investigates the up-and-coming phenomenon of “deepfakes”, a term used to describe fake videos made using deep learning algorithms and face-swapping techniques.
Several academic papers and books serve as the basis for the research. Prominent cases of deepfakes are also described to explain the use cases and possibilities of deepfakes.
The research paper takes a qualitative approach, covering a wide range of topics, including ethics, legalities, and the algorithms behind deepfake software.
The thesis identifies the positive and negative possibilities of deepfakes and investigate technical aspects, such as specific algorithms used to make deepfakes and the technical requirements for generating credible deepfakes.
The motivations behind the people who create deepfakes differ significantly. As part of this study, five such content creators gave their input to answer questions about the risks, possibilities, and technical difficulty of generating deepfakes.
This research involves a practical part where an artifact was produced in the form of a deepfake featuring two prominent celebrities whose faces were swapped using the deepfake software DeepFaceLab. The artifact works as a proof-of-concept, and the process of creating the artifact can be followed in the thesis. This artifact was also evaluated by deepfake content creators on its quality and credibility.
Research questions are formed at the beginning of the study, and they guide the direction of the research. Conclusions to the research questions and the outcome of the study's findings are formulated in the end.
Several academic papers and books serve as the basis for the research. Prominent cases of deepfakes are also described to explain the use cases and possibilities of deepfakes.
The research paper takes a qualitative approach, covering a wide range of topics, including ethics, legalities, and the algorithms behind deepfake software.
The thesis identifies the positive and negative possibilities of deepfakes and investigate technical aspects, such as specific algorithms used to make deepfakes and the technical requirements for generating credible deepfakes.
The motivations behind the people who create deepfakes differ significantly. As part of this study, five such content creators gave their input to answer questions about the risks, possibilities, and technical difficulty of generating deepfakes.
This research involves a practical part where an artifact was produced in the form of a deepfake featuring two prominent celebrities whose faces were swapped using the deepfake software DeepFaceLab. The artifact works as a proof-of-concept, and the process of creating the artifact can be followed in the thesis. This artifact was also evaluated by deepfake content creators on its quality and credibility.
Research questions are formed at the beginning of the study, and they guide the direction of the research. Conclusions to the research questions and the outcome of the study's findings are formulated in the end.