Physical spoofing attacks, including printing, replay, and 3D mask attacks, pose significant security threats to face recognition systems. Advances in display resolution and 3D printing technology have rendered traditional motion and texture-based solutions less effective. Recent activity detection techniques, while promising, often rely on user cooperation or additional hardware, leading to increased costs and complexity. We propose a novel end-to-end face anti-spoofing network architecture that leverages remote photoplethysmography (rPPG) combined with 3D convolutional neural networks (3DCNNs) to extract weak physiological signals without the need for additional ROI selection. Our architecture is enhanced by a Deception Enhancement Module (DEM), improving its robustness against spoofing attacks. Our method demonstrates superior performance and efficiency, achieving high detection accuracy within short detection times (1 s) across various physical spoofing benchmarks.