RollingEvidence: Autoregressive Video Evidence via Rolling Shutter Effect
summary
An active system embeds rolling shutter stripes by modulating an LED with a 16‑FSK LED probe and records them on commodity CMOS phones; an autoregressive encoder produces 4,096‑dim autoregressive tokens tied to device cryptographic keys and prior frames, packed into overlapping windows (optimal 30–150 frames). A multitask deep network extracts stripe intensity curves, decodes probes, and flags tampered frames with a 98% quantile test and exponential‑min implication; prototype (smartphone + Arduino Uno) shows zero FAR above −18.19 dB SNR and ≈99.5% detection accuracy.
AI-written summary · about
chapters (7)
- 0:00RollingEvidence system
- 1:15Rolling shutter probes
- 3:03Calibrating strip patterns
- 4:27Stochastic probe encoding
- 6:13Security against tampering
- 7:18Prototype validation
- 9:50Key contributions
related talks
40mVideo-Based Cryptanalysis: Recovering Cryptographic Keys from Device Using Video of Power LED
11mAIRCODE: Hidden Screen-Camera Communication on an Invisible and Inaudible Dual Channel
13mlo-fi spy: video, steganography, and you
24mBeyond the Frame: Using Computer Vision to Unlock the Potential of Video Data
30mInfrared: Old Threat Meets New Devices
10mInvestigating Stealthy Latency Attacks Against Camera-Based Object...
55mReliable and Secure Embedded Systems
23mMultimodal Video Face Liveness: A Better Alternative?