Future Trends in Digital Face Manipulation and Detection

Abstract

Recently, digital face manipulation and its detection have sparked large interest in industry and academia around the world. Numerous approaches have been proposed in the literature to create realistic face manipulations, such as DeepFakes and face morphs. To the human eye manipulated images and videos can be almost indistinguishable from real content. Although impressive progress has been reported in the automatic detection of such face manipulations, this research field is often considered to be a cat and mouse game. This chapter briefly discusses the state of the art of digital face manipulation and detection. Issues and challenges that need to be tackled by the research community are summarized, along with future trends in the field.

A novel blind tamper detection and localization scheme for multiple faces in digital images

“…For future work, the authors of [10] suggested the use of unsupervised or semi-supervised learning to improve the performance in terms of time and manual annotation processes. In [48], the authors highlighted the challenges and limitations that should be taken in consideration in the future researches. They described the work in this field as "a cat and mouse game"; whereas the detection techniques are improved, the manipulation methods are improved.…”

“…They described the work in this field as "a cat and mouse game"; whereas the detection techniques are improved, the manipulation methods are improved. Many researches have been introduced tackling the manipulation detection issue; however, to the current date, there is no global reliable face manipulation detection technique, which means this field is still nascent [48][49][50]. One of the limitations that have been mentioned in [48] is the low detection accuracy in the GAN-based methods [51][52][53][54] when bad-quality input images are tested such as images in bad lighting conditions, noisy, blurry, and other low-resolution images.…”

“…Many researches have been introduced tackling the manipulation detection issue; however, to the current date, there is no global reliable face manipulation detection technique, which means this field is still nascent [48][49][50]. One of the limitations that have been mentioned in [48] is the low detection accuracy in the GAN-based methods [51][52][53][54] when bad-quality input images are tested such as images in bad lighting conditions, noisy, blurry, and other low-resolution images. The second limitation mentioned in [48] is the high computational complexity of the deep-learning based techniques which increases the processing time.…”

“…One of the limitations that have been mentioned in [48] is the low detection accuracy in the GAN-based methods [51][52][53][54] when bad-quality input images are tested such as images in bad lighting conditions, noisy, blurry, and other low-resolution images. The second limitation mentioned in [48] is the high computational complexity of the deep-learning based techniques which increases the processing time. In addition, the authors highlighted the following challenges: generalizability, emerging face manipulation techniques, interpretability, vulnerability, and the difficulty of characterizing the type of manipulations, which can restrict the practical application of a large number of the available detection techniques.…”

IRL-Net: Inpainted Region Localization Network via Spatial Attention

“…Even recent advances in inpainting methods show that they can fill large missing areas with meaningful structures and objects that do not exist anywhere else in the image [9]. Such advancements make the manipulation detection a very challenging process [13], especially when the aim is not only to discriminate manipulated images from the authentic ones, but also to pinpoint tampered regions at the pixel level [14]. Notably, different categories of GAN-based inpainting methods [15] are trained using various sizes of masks which enable them to predict small or large masked regions, leading to, as shown in Fig.…”