Title
Panel E: What You See Is Not What You Know: Deepfake Image Manipulation
Faculty Mentor(s)
Dr. Bryson Payne, Dr. Chuck Robertson, Dr. Tamirat Abegaz, Dr. Royce Dansby-Sparks
Campus
Dahlonega
Proposal Type
Oral Presentation
Subject Area
Computer Science
Location
Nesbitt 3212
Start Date
25-3-2022 11:00 AM
End Date
25-3-2022 12:00 PM
Description/Abstract
Social media is becoming a large part of people's lives. What is posted online is quickly believed by those who view it. This is due to the many psychological components that cause people to accept information as truth. Because of this, deepfakes pose a potential online threat. Deepfakes are videos that have been altered from their original form by swapping faces, changing audio, or any other change made to the video that manipulates the meaning. Deepfakes can range from being lighthearted to being deceitful and containing misinformation, information that is not true. The deceitful videos are known to spread rapidly online and influence people's opinions and ideas. This study was designed to learn more about deepfakes by analyzing people's ability to determine if videos are deepfakes. The researcher created deepfake videos by using DeepFaceLab, a GitHub deepfake software. A survey was put together consisting of deepfake videos and original unedited videos. The participants had to view the videos and determine if the videos shown were deepfakes or originals. They also provided their confidence with their answer. The survey fluctuated the familiarity that the viewers had with the subjects of the videos, by occasionally providing photos of the subjects. Also, the number of videos shown at one time was manipulated. This survey helped to provide information on what creates more deceptive deepfakes, what could potentially counter a deepfake, and how well people discover deepfakes.
Media Format
flash_audio
Panel E: What You See Is Not What You Know: Deepfake Image Manipulation
Nesbitt 3212
Social media is becoming a large part of people's lives. What is posted online is quickly believed by those who view it. This is due to the many psychological components that cause people to accept information as truth. Because of this, deepfakes pose a potential online threat. Deepfakes are videos that have been altered from their original form by swapping faces, changing audio, or any other change made to the video that manipulates the meaning. Deepfakes can range from being lighthearted to being deceitful and containing misinformation, information that is not true. The deceitful videos are known to spread rapidly online and influence people's opinions and ideas. This study was designed to learn more about deepfakes by analyzing people's ability to determine if videos are deepfakes. The researcher created deepfake videos by using DeepFaceLab, a GitHub deepfake software. A survey was put together consisting of deepfake videos and original unedited videos. The participants had to view the videos and determine if the videos shown were deepfakes or originals. They also provided their confidence with their answer. The survey fluctuated the familiarity that the viewers had with the subjects of the videos, by occasionally providing photos of the subjects. Also, the number of videos shown at one time was manipulated. This survey helped to provide information on what creates more deceptive deepfakes, what could potentially counter a deepfake, and how well people discover deepfakes.