Face anti-spoofing is currently a critical aspect in face biometric systems to enhance security by distinguishing between real and spoof images. Convolution Neural Networks (CNNs), in particular, are deep learning models that have shown impressive results recently in enhancing the precision and resilience of facial anti-spoofing mechanisms. However, many existing methods are highly sensitive to variations in input data, which reduces their effectiveness in detecting spoofed face images. This research addresses this gap by proposing a novel Convolutional Neural Network model integrated with normalized features for improved face spoofing detection. To acquire normalized features, the CNN model is incorporated with batch normalization process which standardizes layer activations across mini-batches, mitigates internal covariate shift and promotes robust feature learning. This approach reduces sensitivity to input distribution variations, facilitating smoother convergence and faster training. Additionally, training with CelebA-Spoof dataset, which has an extensive coverage of spoofing attacks, ensures that the model trained with the dataset is exposed to a diverse array of spoofing techniques, enhancing their ability to generalize to unseen attacks. A key contribution is the use of batch normalization in the CNN, which improves model stability, generalization, and accuracy in face spoofing detection. The developed model is evaluated using key performance metrics, including Confusion Matrix, Equal Error Rate (EER), Half Total Error Rate (HTER), Precision, Recall, Area under the Receiver Operating Characteristic Curve (AUC), and False Acceptance Rate (FAR). The batch-normalized CNN model achieves an accuracy of 97.71%, demonstrating its superiority over the non-batch normalized version in detecting spoofed faces.