Solving crimes, preventing fraud, biometric technology has enabled our society for years. Biometric facial recognition is now one of the most advanced and demanded identification technology for online use. These days, it comes to our daily lives and spreads prodigiously. Unlock your mobile at a glance, compile albums of people you hang out with, tag photos from your birthday party. It is convenient, easy and fast, no more password memorizing. But here comes the question. Do you know how it works? Have you wondered who memorizes your face and how these devices or applications identify you yourself? Technologies may vary, while here come four basic steps for you: Step 1. Face Detection An incoming image of your face will be captured and recognized from a camera, either you are alone or in a crowd. It’s better with the subject looking ahead straight into the camera. This would be much more secure than it captured from a static image whether it’s from a photo or a video. To match with the current public photo database, most facial recognition relies on 2D images currently. Step 2. Facial signature Facial recognition software reads and analyzes every distinguishable landmark and nodal point of your face. It includes the distance from forehead to chin and the shape of cheekbones.  Each of us would have 60-80 nodal points to distinguish from others. You can see these as your facial signature, which is one-of-a-kind. Step 3. Image to Data Once your signature is created, the analysis will turn it into a mathematical formula. It means that all detailed features of your face have become numbers in a code, which we call as a faceprint. Similar to the uniqueness of everyone’ thumbprint, the biometric data has to match the person who requests access to a mobile device, an application, or even a building. Step 4 Finding a match Thanks to the use of artificial intelligence (AI) and machine learning technologies, the current facial recognition system can process the data fast and accurately, and ensure the information security and reliability at once. Remarkably, with the integration of these algorithms and computing techniques, the process can be operated in real-time.