Is now the time to solve the deepfake threat?
summary
Claim: an explosion of deepfakes is imminent and must be addressed now. Evidence: a convincing demo assembled from Keanu Reeves footage using Facebook open source software on cloud credits for $0 and about 30 minutes. Experiments: training ResNet classifiers on generated fakes, splicing and blur/color artifact detection, and finding post‑processing and transcoding defeat detectors. Proposal: per‑frame hashing with a modulus to produce invariant time windows, store signed references (blockchain or similar) and validate edited files against source hashes.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 0:53Why the sense of urgency
- 1:20What are deepfakes
- 2:12How I made this video
- 3:58What are the implications
- 4:58Salzburg example
- 5:19Disinformation
- 8:34Critical videos
- 9:36What can we do
- 11:00AI on deepfake
- 12:18Detecting splicing
- 13:45Automated deepfakes
- 14:34Systemic challenges
- 16:12Trust layer
- 16:56Existing approaches
- 17:26Video
- 18:11Video files
- 19:08Merkle trees
- 19:40Video hashes
- 21:26Hash of hashes
- 21:55Hash per frame
- 22:33Invariant time windows
- 22:45In practice
- 23:14Jim Acosta
- 26:02The future of video
related talks
30mReality Under Attack: Just How Serious Is the Deep Fake Threat?
16mThe Threat of Deepfakes to Remote Identity Verification Systems
30mAI-Generated Synthetic Media, aka Deepfakes and its impact
08mAdvancements in facial recognition
41mThe Deepfake Supply Chain: Weaponizing Synthetic Media into Modern Cybercrime Operation
27mTruth or Fiction The Art of Unveiling DeepFakes in Text, Audio, Image
28mDefending our reality in the era of deepfakes
41mDeepfake Threats in Business: What You Need to Know