Dan Oneata

Fake media consistently generates lower-magnitude representations, allowing simple statistical methods to rival complex detection systems.

A lightweight audio deepfake detection tool that outperforms existing models while preserving user privacy by operating entirely on-device.

Achieving spoken word segmentation without any text supervision, this method outperforms traditional neural approaches while enhancing interpretability.

Achieving 99.76% accuracy in TTS source attribution, this framework redefines the standards for audio forensics in the era of deepfake audio.