Scoop: Mitigation of Recapture Attacks on Provenance-Based...
summary
Provenance stamps alone are vulnerable to recapture attacks that display fakes on screens or printouts and then rephotograph them; Scoop detects these by comparing Time-of-Flight depth sensors (device depth provenance) with learning-based monocular depth estimates to find disagreement. Implemented on iPhone 14 Pro and Samsung Galaxy S20 Plus with desktop tools, prototypes flag recaptures with ≈95% accuracy on iPhone and 74% on Samsung across a first-of-its-kind dataset of 122 captures including 78 recapture scenarios; false positives are low but current prototypes incur nontrivial, unoptimized runtime and storage overhead.
AI-written summary · about
chapters (11)
- 0:00Provenance-based media auth
- 1:14Why provenance matters
- 3:01Recapture attack threat
- 5:54Existing countermeasure limits
- 7:19Human detection failures
- 8:02Depth-enabled provenance
- 10:35Scoop detects recapture
- 13:26Scoop prototypes & apps
- 14:12Evaluating Scoop dataset
- 15:59Performance overhead
- 16:42Summary & resources
related talks
30mCapture, Store, Verify with Authenticity by Design
14mBreaking Infrared Recapture Detection: Optical-Synthesis Attacks and Depth-Aware...
08mAdvancements in facial recognition
40mNew Face, Who Dis? Protecting Privacy in a World of Surveillance
26mVirtual U: Defeating Face Liveness Detection by Building Virtual Models...
28mFace spoofing detection: theory and practice
16mHands-On with Media Verification for Blockchain Provenance & Integrity Proof
45mYour passport to the 7th continent