A New Face Swap Detection Technique for Digital Images

Abstract

In recent years, the rapid development of deep learning-based face image manipulation algorithms and applications became one of the challenges that are facing information forensics and information security systems. Using these applications, one can easily swap the face in a digital image with another face for different intentions where most of them are malicious intentions. Different face swap detection techniques have been presented in recent years to check the authenticity of the face in a digital image. Most of the available techniques are machine-learning or deep-learning based which makes them vulnerable to false detection results in addition to the time-consuming training process. In this paper, a new technique for face swap detection (FSD) is presented based on the image watermarking process. The proposed technique consists of two main algorithms called embedding and authentication algorithms. Several experiments have been conducted to evaluate the performance of the proposed technique and to prove its efficiency in detecting fake faces. The proposed technique outperforms various deep-learning-based techniques because no training is required and the detection accuracy is 100 %. The performance of the proposed FSD technique is promising therefore it is applicable in different practical applications.