Toward a Universal Deepfake Audio Detector: Robust Deepfake Detection of Speech and Non-Speech in Real-World Conditions

Period: July 2025 – March 2027
Grant type: Japan Society for the Promotion of Science Grant-in-Aid for Scientific Research (Research Activity Start-up Support)
Project number: 25K24398
URL: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-25K24398/

This project will analyze artifact differences between speech and nonspeech deepfakes, identify detector vulnerabilities, and develop a universal model robust to speech, music, noise, and environmental sounds, enhancing reliability under real-world conditions.