MesoNet: a Compact Facial Video Forgery Detection Network

Darius Afchar (1) , Vincent Nozick (2, 3) , Junichi Yamagishi (4) , Isao Echizen (4)

1 ENPC - École des Ponts ParisTech
2 JFLI - Japanese French Laboratory for Informatics
3 LIGM - Laboratoire d'Informatique Gaspard-Monge
4 NII - National Institute of Informatics

This paper presents a method to automatically and efficiently detect face tampering in videos, and particularly focuses on two recent techniques used to generate hyper-realistic forged videos: Deepfake and Face2Face. Traditional image forensics techniques are usually not well suited to videos due to the compression that strongly degrades the data. Thus, this paper follows a deep learning approach and presents two networks, both with a low number of layers to focus on the mesoscopic properties of images. We evaluate those fast networks on both an existing dataset and a dataset we have constituted from online videos. The tests demonstrate a very successful detection rate with more than 98% for Deepfake and 95% for Face2Face.

Mots clés: Face; Training; Convolutional codes; Forgery; Deep learning; Decoding

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]; Traitement des images [eess.IV]

Darius Afchar, Vincent Nozick, Junichi Yamagishi, Isao Echizen. MesoNet: a Compact Facial Video Forgery Detection Network. WIFS 2018, Dec 2018, Hong Kong, China. pp.26.1-26.7, ⟨10.1109/WIFS.2018.8630761⟩. ⟨hal-01867298⟩