Deepfake Detection Analysis Based on Video Face Analysis

The article examines the detection of manipulation of multimedia content, in particular in the form of videos and photos. The concept of deepfake is considered, the means and technologies involved in deepfake are analyzed, current and promising methods of detecting and preventing fake multimedia content are evaluated. Approaches to the methods of creating fakes on videos with elements of manipulation of the human face are defined. Faceswap and Deepfakes Web are found to be the most powerful solutions used to generate deepfakes. An analysis of tools for detecting deepfakes on the face was carried out, which made it possible to highlight the main technologies, libraries, and datasets used to search for fakes. The DeepFake Detection Challenge (DFDC) Model, Forensically, and others have been found to have significant potential and sample size to accurately detect human-faced fakes. This paper proposes a generalized approach to the detection of deep fakes on the human face using machine learning and computer vision technologies