Abstarct: The use of biometric authentication methods, especially face recognition, is a good solution for identification due to its easy access. The existence of various commercial identification systems is a testament to the significant progress made in this field. Despite these achievements, face recognition is still an active topic in the field of computer vision research. In this context, researchers' attention is drawn to methods of preventing fraud by a non-genuine person. Fraud in face recognition is done in various ways such as face mask, replay of face image from screen and face image printed on paper. So far, methods have been introduced to detect fraud that are sometimes intrusive, that is, forcing the user to perform a movement such as shaking the head to lead to the detection. There are differences between real face images and images containing fraud such as face texture and edges. For example, in image reproduction fraud, there is periodic noise in the texture, or in mask fraud, the cut edges of the image are important for detecting fraud. In this study, to differentiate the difference between fake and real images, a local binary pattern was used to highlight the edges and texture of the image. By examining the obtained results, higher accuracy is obtained by using the local binary model. This is due to the dual effect of this pattern on the edges and texture in one place. Deep neural network with three laxyers were used to extract the feature of the image. One of the advantages of deep neural network is automatic feature extraction. The CASIA dataset and the prepared dataset CASIA+, which includes face images covered with sunglasses, were used to check the accuracy. Finally, the proposed method of fraud detection with 98% accuracy and 97% detection were obtained in these datasets, respectively.