Story perspectives

New Light Technique Detects Deepfakes with Invisible Watermarks

8/14/2025

40 5

Summary

Story summary

Cornell University researchers developed noise-coded illumination to embed invisible watermarks in light for video authentication.
This technique, showcased at SIGGRAPH 2025, helps detect deepfake manipulations through encoded lighting patterns.
It remains effective post-editing, increasing trust in video evidence.
Challenges include requiring controlled lighting and addressing potential countermeasures.
Future efforts will focus on refining the technology for smartphone applications.

1 of 1

Story summary

Cornell University researchers developed noise-coded illumination to embed invisible watermarks in light for video authentication.
This technique, showcased at SIGGRAPH 2025, helps detect deepfake manipulations through encoded lighting patterns.
It remains effective post-editing, increasing trust in video evidence.
Challenges include requiring controlled lighting and addressing potential countermeasures.
Future efforts will focus on refining the technology for smartphone applications.