Face anti-spoofing (FAS) is a critical component of biometric security systems, designed to detect and prevent attempts to deceive facial recognition systems through the use of fake faces, such as photos, videos, or masks. Applications of FAS include secure access control, financial transactions, and identity verification in various industries. While Near-Infrared (NIR) imaging has shown promise in enhancing FAS by capturing features invisible in the visible spectrum, it also presents challenges, such as the need for specialized hardware and the difficulty of capturing high-quality NIR images in real-time scenarios. Additionally, existing multi-modal models that combine RGB and NIR images often struggle with effective data fusion and suffer from increased computational complexity. In this research, we propose a novel framework that addresses these limitations by leveraging an RGB to NIR estimation module, which uses a GAN-based domain adaptation technique to generate NIR images directly from RGB inputs. This approach eliminates the need for dedicated NIR sensors, making it more practical for real-world applications. The generated NIR images are fused with the original RGB images and processed through a Vision Transformer-based classifier, which effectively distinguishes between live and spoofed faces. Experimental results demonstrate that our method improves the robustness and accuracy of FAS and reduces reliance on multi-modal sensors, offering a scalable and efficient solution for biometric security.