Machine Learning for Deepfake Detection

reportActive / Technical Report | Accesssion Number: AD1180868 | Open PDF

Abstract:

We learned why we need detectors, but why do we need machine learning? Potential Dangers: >700k hours of video uploaded to YouTube daily; Deepfake apps can be run with push of a button; Deepfakes are generated with ML, logical then to think that we can detect them with ML; Castle defense. We need scalable detectors!

Author(s):
Gallagher, Shannon K.
Author Organization(s):
CARNEGIE-MELLON UNIV PITTSBURGH PA
Funding Organization(s):
AIR FORCE LIFE CYCLE MANAGEMENT CENTER HANSCOM AFB MA, HANSCOM AFB , MA
Document Type:
Technical Report/Briefing Charts
Publication Date:
2022 Sep 23
Pagination:
44

Security Markings

DOCUMENT & CONTEXTUAL SUMMARY

Distribution Code:
A - Approved For Public Release
Distribution Statement: Public Release

RECORD

Collection: TRECMS
Identifying Numbers
Contract Number(s):
FA8702-15-D-0002
Subject Terms
Modernization Areas:
AI and Machine Learning
Communities of Interest:
Autonomy
Descriptor(s):
detection, software development, deepfakes, detectors, learning, materials, universities, department of defense, engineering, machine learning, guarantees, images, governments, patents, trademarks
Keyword(s):
deepfake detection
Subject Categories:
Mathematical and Computer Sciences; Behavioral and Social Sciences
Creation Date:
2022 Sep 26
Update Date:
2022 Dec 21