Cornell computer science researchers have pioneered a novel technique to embed invisible secret codes in lighting—a digital watermark carried by light itself that enables forensic analysts to detect video manipulation. Unlike conventional watermarks, this coded lighting remains intact regardless of who records the footage or what device is used, reports Cornell Chronicle.
The system works by subtly modulating light—so imperceptible to the human eye—that a hidden signal is recorded in any video captured under it. If a video is altered (e.g., segments removed, objects added), forensic tools pick up mismatches: edited areas may appear black when decoding the watermark, and distinct code sequences fail to align across video segments.
The team has also successfully applied multiple independent codes simultaneously—such as overlaying three separate lighting patterns in one scene—making counterfeiting exponentially more difficult. Even if adversaries learn the technique, they must replicate each coded layer perfectly across all frames to avoid detection.
Crucially, the researchers tested the approach across real-world scenarios—including outdoor environments and subjects with varying skin tones—and confirmed robustness across lighting conditions and diversity.
While acknowledging that misinformation remains an advancing threat, the study offers a promising new tool in the arms race against deepfakes. As lead researcher, Davis notes, even if deceptive actors understand the scheme, producing consistent, multicode fake videos remains a daunting challenge.