Nowadays, smart home technology has become one of the leading IoT-based projects and as a result of that there are lots of new IoT-based products available in society that allow people to live more convenient and secure lives at home. Many people are aware of this technology, and the smart home application helps people to manage their schedules, home lighting, electricity bills, grocery lists, and also their home security. Today, face recognition is a well-established and popular process to keep homes safe. Here facial recognition helps detect and identify faces that we want to allow into our home. In this aims research, we propose an intelligent door system using an Arduino microcontroller with face recognition. It has two steps for extra security and a smart door system using face recognition. The first step is to find and predict whether the image is fraudulent or real. The fraud image is a face image taken from a screen smartphone, tablet, TV monitor, or computer laptop. If the system predicts a fraud image, it will lock the door, otherwise, if it predicts a real image, then the system will run the second step, the step of identifying the owner by face recognition. In the second step, we created two models for comparing the accuracy rate, Model 1 is a new model from the Convolutional Neural Network (CNN), and Model 2 uses a transfer learning model from the renet18 model. In both models, we get an accuracy rate on the training set of 95.83% for model 1 and 100.00% for model 2. In this second step, if the system determines the owner, the door will be unlocked in 30 seconds, and if not, it will be locked.