Facial biometrics is a technology where face recognition systems are quite important. Access controls, electronic devices, and secure systems like laptops and smart phones may be used for a variety of purposes. The safety of face recognition systems is at risk, however, since criminals are trying to get around biometric security measures by impersonating a legitimate user via photos, 3D masks, or replay videos. When it comes to hacking face recognition systems, video attacks are among the most common, inexpensive, and easy spoofing methods. By extracting video frames as well as classifying them employing the stacked Convolutional Neural Network (CNN), a deep learning algorithm, this research paper aims to regulate if the input biometric samples from a live face or a spoof attack. The focus is on face-liveness discovery in video attacks. This study used the WVU dataset's spoof videos as the data set. The findings demonstrated better accuracy in video liveness detection. Since processing time is modest due to the low number of frames, this result is promising. Regardless of the lighting, non-planar movement, or position changes in the video clips, the multimodal identification accuracy illustrates the superiority as well as resilience of the proposed technique. This is true even when certain modalities are absent.