He emphasizes that no purely technical solution exists, advocating for a multi-layered approach centered on human analysis, monitoring, and methodology. The first layer involves automated forensic detection of synthetic content, though its effectiveness against real-world deepfakes is limited, with performance dropping by 45-50% in videos, audio, and images. This is due to an arms race where detection constantly lags behind generation.
The second, more promising layer, focuses on certifying authentic content through standards like C2PA and Content Credentials, which embed cryptographically signed manifests into files. While adopted by major players like Adobe, Microsoft, and the BBC, its widespread adoption by device manufacturers and platforms is crucial for its full potential. The core of the defense, however, lies in monitoring and Open Source Intelligence OSINT.
Monitoring acts as an early warning system, involving continuous, automated surveillance of platforms, filtering, anomaly detection, and real-time alerts to human analysts. Human qualification is irreplaceable for determining if a signal is noise, legitimate organic activity, or an operation requiring investigation. The goal is to detect operations before they go viral, as the window for effective action is often hours or minutes.
OSINT then takes over for verification and attribut
