Evaluating the accuracy of algorithms of the blind detection for faked videos

Abstract

Forgery video detection methods can be divided into positive detection methods and blind detection methods, our research focuses on blind detection methods. We have evaluated a group of the most popular algorithms in terms of performance and speed, applying to a database of (22) fake videos, Most of them are produced by the researcher, some from accredited research centers, videos published on the Internet, and (6) real videos filmed by the researcher, and these videos have been forged in different ways and with varying degrees of forgery complexity, and we have found that most of these algorithms focus on detecting one type of forgery. And it is effective for it, while it is limited for other types of forgery, in addition to the proportionality between the effectiveness of the algorithm and the detection time, and this is a real obstacle, especially if the videos studied are long, or contain several types of forgery and in more than one place in the same frame.