A Forensic Method for DeepFake Image based on Face Recognition
Jian Wu, Kai Ping Feng, Xu Chang, Tongfeng Yang · 2020
DeepFake digital images have serious negative impacts on news integrity, legal forensics, and social security. In order to detect the DeepFake digital images more accurately, a method based on face recognition is proposed. Face image feature vectors are extracted by Facenet, and the Euclidean distances among the vectors of different face images are calculated as classification principle. Then, the machine learning algorithms is trained to perform binary classification of real and fake face images. The experimental results on the Celeb-DF data set show that the proposed method has better detection effect than the existing detection methods.