Face Recognition Security Against Deepfakes by Using Multimodal Detection: A Survey

Technology has become an essential component of daily life, with millions of people placing their trust in it. Security is important, especially in maximizing the optimization of face recognition (FR) technology, since it has been an enormous concern. The process of recognizing people from images or video clips occurs by using algorithms that detect and extract distinctive facial features. This paper provides a comprehensive study of FR security threats, by reviewing multiple algorithms and Deep Learning (DL) approaches in the facial recognition field. The point of this paper is to discuss the challenge posed by facial spoofing attacks, using either 2D or 3D masks. In this paper we demonstrated different implementations of Convolutional Neural Networks (CNNs) and Artificial Neural Networks (ANNs) algorithms to address the complexity of deep fake detection, focusing on their varied methodologies for facial detection and recognition. The paper emphasizes on the importance of the CNNs and ANNs algorithms using the Inception V3 and ResNet-50 models that helps learn, verify, and recognize face images to enhance the results to be more accurate through using the precise models in common areas such as gates surveillance. Furthermore, different models will be discussed in the paper along with Inception V3 which obtained the maximum testing results based on high performance when interacting in various image environments and high reliability in handling complex data.