A new face image manipulation reveal scheme based on face detection and image watermarking

Abstract

Face image manipulation (FIM) algorithms and applications are increasing and distributing rapidly. Nowadays, one can easily find an application to manipulate face images for different purposes. After the development and popularity of “DeepFakes”, the research community highlighted the necessity of implementing new detection techniques that can reveal FIM. The available FIM detection techniques have different limitations and most of the presented schemes require a prior knowledge of the FIM method that has been applied to generate the manipulated face image which makes them restricted. In this paper, we present a new face image manipulation reveal (FIMR) scheme based on face detection algorithm and image watermarking technique in the transform domain. The proposed scheme does not need a prior knowledge of the FIM method thus it can reveal different types of FIM. Experiments have been conducted to evaluate the performance of the proposed FIMR scheme for face images with different sizes. The results proved that the proposed scheme can successfully reveal different manipulation attacks such as face swap, expression swap, attribute attacks, retouching attacks, and morphing attacks.