DeepFakes and Beyond: A Survey of Face Manipulation and Fake Detection

The free access to large-scale public databases, together with the fast progress of deep learning techniques, in particular Generative Adversarial Networks, have led to the generation of very realistic fake contents with its corresponding implications towards society in this era of fake news. This survey provides a thorough review of techniques for manipulating face images including DeepFake methods, and methods to detect such manipulations. In particular, four types of facial manipulation are reviewed: i) entire face synthesis, ii) face identity swap (DeepFakes), iii) facial attributes manipulation, and iv) facial expression manipulation. For each manipulation type, we provide details regarding manipulation techniques, existing public databases, and key benchmarks for technology evaluation of fake detection methods, including a summary of results from those evaluations. Among the different databases available and discussed in the survey, FaceForensics++ is for example one of the most widely used for detecting both face identity swap and facial expression manipulations, with results in the literature in the range of 90-100% of manipulation detection accuracy. In addition to the survey information, we also discuss trends and provide an outlook of the ongoing work in this field, e.g., the recently announced DeepFake Detection Challenge (DFDC).