Detecting DeepFakes by Comparing Facial Action Units Eli Annoni, Scott Fitzpatrick, Lisa Pham, Cole Matsueda, Owen Walsh
May 11, 2021
A DeepFake is a digitally altered face intended to mimic another person. This technology is advancing rapidly, with the amount of DeepFakes increasing by 84% in 2019. Traditional detection methods that observe artifact and pixel inconsistencies fail to keep up. To combat this issue, our team developed a DeepFake detection software that: analyzes facial data derived from the motion capture software OpenFace, creates profiles to represent certain individuals, compares profiles to detect differences in facial Action Unit behavior.