A research group led by Associate Professor Nozick (former visiting researcher at NII), Associate Professor Yamagishi, and Professor Echizen has developed a new deep learning technology to detect fake videos.

A research team consisting of Associate Professor Vincent Nozick (University of Paris-Est Marne-Vallée, France), former visiting researcher at the National Institute of Informatics (NII, Director General: Kiryugawa Masaru, Chiyoda-ku, Tokyo), Associate Professor Yamagishi Junichi of the Content Sciences Research Division, Professor Echizen Isao of the Information Society Interaction Research Division, and Darius Afka (former NII Special Joint Research Fellow/former graduate student at the National School of Civil Engineering, France), has developed a new deep learning technology that can automatically detect sophisticated fake videos generated by technology that automatically alters faces in videos.

This technology focuses on generating realistic fake videos like "DeepFake" and "Face2Face". In evaluation experiments, "DeepFake" achieved a detection accuracy of over 98% and "Face2Face" achieved an accuracy of over 95%.

The results of this research will be presented at the 2018 IEEE International Workshop on Information Forensics and Security (WIFS 2018), which will be held from December 11th.