Abstract:Recent advancements in machine learning and computer vision have led to the proliferation of Deepfakes. As technology democratizes over time, there is an increasing fear that novice users can create Deepfakes, to discredit others and undermine public discourse. In this paper, we conduct user studies to understand whether participants with advanced computer skills and varying levels of computer science expertise can create Deepfakes of a person saying a target statement using limited media files. We conduct two studies; in the first study (n = 39) participants try creating a target Deepfake in a constrained time frame using any tool they desire. In the second study (n = 29) participants use pre-specified deep learning-based tools to create the same Deepfake. We find that for the first study, 23.1% of the participants successfully created complete Deepfakes with audio and video, whereas, for the second user study, 58.6% of the participants were successful in stitching target speech to the target video. We further use Deepfake detection software tools as well as human examiner-based analysis, to classify the successfully generated Deepfake outputs as fake, suspicious, or real. The software detector classified 80% of the Deepfakes as fake, whereas the human examiners classified 100% of the videos as fake. We conclude that creating Deepfakes is a simple enough task for a novice user given adequate tools and time; however, the resulting Deepfakes are not sufficiently real-looking and are unable to completely fool detection software as well as human examiners

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve This paper aims to explore whether ordinary users can easily create Deepfakes (deepfake videos). With the development of machine learning and computer vision technologies, the generation of Deepfakes has become increasingly common. The authors are concerned that over time, the proliferation of technology may enable ordinary users to create Deepfakes for malicious purposes, such as defaming others or disrupting public opinion. To verify this hypothesis, the authors conducted two user studies: 1. **First Study** (n=39): Participants attempted to create a target Deepfake using any tools within a limited time. 2. **Second Study** (n=29): Participants attempted to create the same Deepfake using a pre-specified deep learning tool within the same time frame. ### Main Research Questions 1. **Ease of Deepfake Generation**: - Can ordinary users generate Deepfakes within a limited time? - How do users with different technical skill levels perform in generating Deepfakes? 2. **Realism of Generated Deepfakes**: - Are the generated Deepfakes realistic enough to deceive human reviewers? - Can the generated Deepfakes pass detection by software tools? ### Research Methods 1. **Participant Background**: - Participants with varying levels of computer skills were selected, including basic, intermediate, advanced computer skills, and digital media skills. - Participants were required to complete the task within 2 hours. 2. **Experimental Design**: - **Open-Ended Generation Study**: Participants were free to choose any tools to generate Deepfakes. - **Predefined Tool Generation Study**: Participants used pre-specified tools to generate Deepfakes. 3. **Evaluation Methods**: - The generated Deepfakes were classified by human reviewers and Deepfake detection software to determine their authenticity. ### Main Findings 1. **Success Rate of Deepfake Generation**: - In the open-ended generation study, 23.1% of participants successfully generated a complete Deepfake with both audio and video. - In the predefined tool generation study, 58.6% of participants successfully synthesized the target voice into the target video. 2. **Deepfake Detection Results**: - Human reviewers marked all generated Deepfake samples as fake. - Detection software marked 80% of the Deepfake samples as fake. ### Conclusion The study shows that although ordinary users can generate Deepfakes given the appropriate tools and time, the generated Deepfakes are not realistic enough to fully deceive detection software and human reviewers. This indicates that while Deepfake generation technology is becoming more accessible, its realism and deceptive capabilities are still limited.

Can deepfakes be created by novice users?

Unmasking the Deepfake Infocalypse: Debunking Manufactured Misinformation with a Prototype Model in the AI Era “Seeing and hearing, no longer believing.”

Deepfakes as a threat to a speaker and facial recognition: An overview of tools and attack vectors

Deepfakes generation and detection: state-of-the-art, open challenges, countermeasures, and way forward

DeepFakes: a New Threat to Face Recognition? Assessment and Detection

Why Do Facial Deepfake Detectors Fail?

Video and Audio Deepfake Datasets and Open Issues in Deepfake Technology: Being Ahead of the Curve

Warning: Humans Cannot Reliably Detect Speech Deepfakes

Human Detection of Political Speech Deepfakes across Transcripts, Audio, and Video

Deepfake Media Forensics: State of the Art and Challenges Ahead

Threat of Deepfake Enabled Attack

Are Deepfakes Concerning? Analyzing Conversations of Deepfakes on Reddit and Exploring Societal Implications

Unmasking Illusions: Understanding Human Perception of Audiovisual Deepfakes

Human Perception of Audio Deepfakes

DeepFakes: Detecting Forged and Synthetic Media Content Using Machine Learning

Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security

Protecting World Leaders Against Deep Fakes

Creation and Detection of German Voice Deepfakes

Detecting Deepfakes Without Seeing Any

Comparative Analysis of Deep-Fake Algorithms

Deepfake video detection: challenges and opportunities