Light‑based system detects deepfake videos with 99.86% accuracy
Summary
- The technique projects patterned light onto video frames to expose inconsistencies caused by AI manipulation.
- In laboratory tests it identified manipulated videos with 99.86% accuracy and authentic videos with 95.72% accuracy.
- The system can analyse up to 15 video clips simultaneously, delivering results within a few seconds.
- It requires only standard camera equipment and can be integrated into existing media‑verification workflows.
- Potential applications include newsrooms, law‑enforcement agencies, and social‑media platforms to curb misinformation.
Key Facts
- · Detection accuracy for deepfakes: 99.86%
- · Accuracy for genuine videos: 95.72%
- · Processing capacity: 15 videos at a time
- · Analysis time: a few seconds per batch
- · Method relies on light‑pattern analysis of pixel-level inconsistencies
- · Tested on a diverse dataset of AI‑generated and real videos
- · Requires only conventional cameras and illumination sources
- · Designed for real‑time deployment in media and security contexts
Why it matters
As deepfake technology becomes more sophisticated, a rapid and highly accurate detection tool strengthens media credibility, supports law‑enforcement investigations, and protects national security against misinformation campaigns.
#deepfake#AI#media verification#light‑based detection#cybersecurity#technology