Fake Face Detection using Local Binary Pattern and Ensemble Modeling

Abstract

Fake faces generated with Generative Adversarial Networks (GANs) are becoming more and more realistic and getting harder to be identified directly by human beings. However, CNN (Convolutional Neural network) based deep learning architecture can achieve almost perfect detection accuracy on such fake faces. In this paper we present a study of fake face detection with the exploration of the global texture features based on the empirical knowledge that the textures of fake faces are quite different from those of real faces. A new architecture, LBP (Local Binary Pattern)-Net, is designed to utilize binary representation image texture for the effective identification of fake images. Experimental results show that the proposed method is more robust than existing algorithms for detecting fake images edited by different image augmentation methods, such as blurring, cutout, brightness and color changing, equalization, etc. Ensemble models are also experimented to combine advantages of individual models. The most significant effect of ensemble models is the robustness for detecting edited fake images compared to single models. Experimental results show that our ensemble models outperform single models for detecting fake images.

An evolutionary- convolutional neural network for fake image detection

“…We employed 140 K [35] authentic and synthetic images in a dataset for our studies. This dataset contains 70 K authentic images taken from the flickr-faces-HQ (FFHQ) [18] data gathered by NVidia, as well as 70 K synthetic images selected from the one million synthetic images dataset (created by StyleGAN ); Figure 7 shows this network generates some images.…”

“…The comparisons are compared among proposed evolutionary CNN and some works. Note that Handcrafted-AMNET Guo et al [5] and LBP Wang et al [35] are developed for other forensics tasks, so We need them to be divergent for our forensics investigation. We use three datasets: the HFF, and 140 K. Table 2 reports the accuracy of detection on two datasets.…”

Detection of Deep Fake in Face Images Based Machine Learning

“…The second method ensemble model constructed from five models including LBP-Net, Gram-Net, Res-Net and two models utilizing Inception, ResnetV1 pre-trained on Casia -Webfaceare and vggface2. Results of detecting fake face images by several image augmentation such as downsample (66.32%), brightness (81.09% ), Solarize ( 75.04%), Contrast ( 85.42%) and color (91.06%) when using "140K Real and Fake Faces" [12].…”

“…Images of fake faces are often produced using Generative Adversarial Networks (GANs), which are high-quality, sophisticated images. As a result, recognition of this type of image is extremely difficult and is not directly possible with the naked eye [5]. The example of Deepfake images is shown in Fig.…”

Convolutional Neural Networks for Real and Fake Face Classification

“…Solution for the problem of detecting fake faces is presented in [9], where new architecture of CNN, Local Binary Pattern-Net is designed and detection is based on the texture features of fake faces as it is different than the texture of real faces. Few different CNN models are used in [10] and it is concluded that deep-learning algorithms and models are propriate for recognizing fake faces.…”