One of the key part of the online identity verification process is Liveness detection.
PresentID Dynamic Face liveness detection can recognize the person in front of the video is real or fake. It can detect if the user is physically present and prevents masks, photos, or video deepfakes to fool the system. We use the convolutional neural network (CNN) deep learning algorithms to detect fraud.
In our solution, there are client and server-side processes. Each side consists of several machine learning algorithms.
The client-side processes are done with our client SDK which has been provided for iOS, Android, and Web. In the client SDK, we check that face should be in the camera all the time of the procedure and if we couldn’t find the face in all the frames then the process is stopped. Also, we check the gaze and head pose of the user in each frame.
The gaze model is used to check eye closeness(if the user is asleep or not). Head pose checking is used to prevent the diversity of data. That helps our Dynamic liveness model to train better and easier. This gaze and pose detection don’t complicate the user act, on the other side, it keeps the system safer. Besides that, in the whole procedure, a user should fit his face on an oval. We generate the random oval in the randomness position of the screens. Showing random oval in the random positions on the screen has two advantages. First, it prevents injection. Second, that movement makes bold some spoofing clues for attacks. Moreover, we collect 30 image frames from the client-side.
On the server-side, our Dynamic face anti-spoofing model which is the convolutional model has been trained based on our data. It uses 30 frames in the video and extracts a depth map that is able to detect the depth and is able to detect spoof attacks based on the depth detection. Providing Data for the training models was the main challenge. Our team collects a large dataset from our first attendance app. Also, we have crawled social networks to collect a large video dataset and then clean and create spoof data from them.
2-5 seconds in low-range smartphones for clients-side and less than 1 second on server-side with at least 8 cores & 2.4GH. CPU based server-side. Verify identity. Great with glasses, makeup & beards For all modern smart devices & webcams. Support iOS, Android devices. Web client SDK & API. Easy integration with your app
Use Cases Access Control
Car sharing & Taxi
SDKs are available for iOS, Android, Linux, or Windows
If you want to use our face liveness detection SDK just send us a request via this form.
Free Try Face Liveness Detection Web SDK
Dynamic Face Liveness
PresentID Dynamic Face liveness detection can recognize the person in front of the video is real or fake.