Brown Digital Repository | Item

Full Metadata

Overview

Year:: 2020
Contributor:: Tursman, Eleanor (Author); George, Marilyn (Author); Kamara, Seny (Author); Tompkin, James (Author)
Subject:: Deepfakes; Facial geometry; Social Video Verification
DOI: https://doi.org/10.26300/kyyg-w765

Files

Views

- Opens Archive Contents-View in Iframe
- Opens Archive Contents-View in Current Window
Archive Contents
- Opens Archive Default-View in Iframe
- Opens Archive Default-View in Current Window
Archive Default

Description

Abstract:: Deepfakes can spread misinformation, defamation, and propaganda by faking videos of public speakers. We assume that future deepfakes will be visually indistinguishable from real video, and will also fool current deepfake detection methods. As such, we posit a social verification system that instead validates the truth of an event via a set of videos. To confirm which, if any, videos are being faked at any point in time, we check for consistent facial geometry across videos. We demonstrate that by comparing mouth movement across views using a combination of PCA and hierarchical clustering, we can detect a deepfake with subtle mouth manipulations out of a set of six videos at high accuracy. Using our new multi-view dataset of 25 speakers, we show that our performance gracefully decays as we increase the number of identically faked videos from different input views.

Content

Access Conditions

Use and Reproduction: This work is licensed under a GNU GPL3 License

Citation

Tursman, Eleanor, George, Marilyn, Kamara, Seny, et al., "Processed Landmarks" (2020). Brown University Open Data Collection, Dataset and codebase for Towards Untrusted Social Video Verification to Combat Deepfakes via Face Geometry Consistency. Brown Digital Repository. Brown University Library. https://doi.org/10.26300/kyyg-w765

Relations

Collections:

Brown University Open Data Collection

This collection contains open and publicly-funded data sets created by Brown University faculty and student researchers. Increasingly, publishers, and funders are requiring that protocols, data sets, metadata, and code underlying published research be retained and preserved, their locations cited within …
...
Dataset and codebase for Towards Untrusted Social Video Verification to Combat Deepfakes via Face Geometry Consistency

Raw video and the post-processed landmark matrices for E. Tursman, M. George, S. Kamara and J. Tompkin, "Towards Untrusted Social Video Verification to Combat Deepfakes via Face Geometry Consistency," 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), …
...

Processed Landmarks