Facial recognition technology is widely used for authentication and security purposes; however, its vulnerability to spoofing attacks presents significant risks. Spoofing techniques such as printed photographs, digital screen replays, and 3D masks can deceive recognition systems. This study addresses these vulnerabilities by employing Convolutional Neural Networks (CNNs) to differentiate between genuine and spoofed facial images. Using established datasets such as CelebA-Spoof and CASIA-FASD, along with widely adopted tools like OpenCV and TensorFlow, we develop and evaluate a model designed to enhance facial recognition security. The proposed model emphasizes robust liveness detection to ensure accurate biometric authentication and reduce the risk of identity fraud. Experimental results demonstrate high classification accuracy, underscoring the potential of CNN-based approaches to improve the reliability of facial recognition systems in real-world scenarios.