Evaluating methods of anti-spoofing of living entities and propose solutions

Abstract

In recent 10 years, object recognition technology has been widely developed with many solutions given with high practical applications. However, with the development of digital devices, the resulting images can be provided from a variety of sources: photos taken (or video captured) directly from living entities, photos printed on paper or photos (video) is displayed on the computer screen or mobile phone. Therefore, the image detection obtained from the camera system directly from the living entity or from images are printed/displayed to determine the correct object is the problem posed. The article focuses on the issue of anti-spoofing of human faces with the content outlining the measures applied in the banking system, the access monitoring system in accordance with the previous registration of these faces there. The purpose is to validate customer or user information such as: wink test results, response to random challenges, etc. The article evaluates the advantages and disadvantages of these methods, after that proposes an anti-spoofing solution by using 3D camera and combining with convolutional neural network to create a more accurate anti-spoofing system.