Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore a groundbreaking conference presentation that introduces RollingEvidence, an innovative system designed to combat video manipulation and ensure the authenticity of digital evidence. Learn how this active defense mechanism leverages the rolling shutter effect inherent in CMOS cameras to embed real-time physical-layer probes during video recording, creating distinctive stripe patterns that enhance the integrity of both the probes and the recorded content. Discover the system's autoregressive encoding scheme that generates compact, high-dimensional probes for subsequent frames by incorporating data from previous frames and device-specific cryptographic keys. Understand how deep neural networks are employed during the verification process to decode embedded probes and identify tampered frames using exponential-min implication techniques. Examine the theoretical foundations, prototype implementation, and comprehensive experimental results that demonstrate RollingEvidence's effectiveness in producing and verifying authentic video evidence, addressing critical challenges posed by advanced video manipulation techniques and the vulnerabilities of existing defensive measures including data-driven models and digital watermarks.
