In this paper, we explore a Convolutional Neural Network (CNN) binary classifier that leverages the Transfer Learning Technique, specifically Xception, integrated within the TensorFlow framework. Transfer Learning harnesses knowledge acquired from one task to enhance the performance of a related task. In this context, we exploit the pre-existing knowledge within the extensive dataset learned by Xception to enhance the binary classifier’s performance. A global average pooling layer is integrated into the Xception model architecture to further reduce the spatial dimensions, focusing on the most relevant and significant information from each feature map. A sigmoid activation function is employed to facilitate binary classification. Extensive experimentation on both the CASIA-FASD and LCC FASD datasets demonstrates that when classifying images as real or fake, the model achieves an F1-score of \(99.64\%\) . In addition, we conducted a comparative analysis with several state-of-the-art techniques. The results indicate that the proposed method outperforms these well-known techniques in face anti-spoofing.