This study aims to address the insufficient model recognition accuracy and limitations of authentication techniques in current IoT authentication methods. The research presents a more accurate face video image authentication technique by using a new authentication method that combines convolutional neural networks (CNN) and remote Photoplethysmography (rPPG) volumetric tracing. This method comprehensively analyzes facial video images to achieve effective authentication of user identity. The results showed that the new method had higher recognition accuracy when the light was weak. The new method performed better in ablation experiments. The error rate was 1.12% lower than the separate CNN model and 1.73% lower than the rPPG model. The half-error rate was lower than the traditional face authentication recognition model, and the method had better performance effect. Meanwhile, the images with high similarity showed better recognition stability. It can be seen that the new method is able to solve problems such as the recognition accuracy in identity authentication, but the recognition effect under extreme conditions requires further research. The research provides a new technical solution for the authentication of Internet of Things devices, which helps to improve the security and accuracy of the authentication system. By combining the CNN model and rPPG, the research not only improves the recognition accuracy in complex environments, but also enhances the system's adaptability to environmental changes. The new method provides a new solution for the advancement of Internet of Things authentication technology.