Detection of face spoofing attacks on biometric identification systems

Abstract

The object of the study is spoofing attacks on identification systems based on human face biometrics. The article provides an overview of the main types of spoofing attacks and features of their detection, as well as an overview of existing solutions for detecting attacks, and discusses their advantages and disadvantages. The subject of the research is systems for detecting and counteracting spoofing attacks in face biometric identification systems. The purpose of the work is to create an improved method for detecting and counteracting spoofing attacks based on convolutional neural networks using an image depth map that would provide a HTER of less than 1%. The proposed system allows detecting all available types of spoofing attacks (printed attack, replay attack and mask attack) and working with both single frame and video stream. Ref. 8, pic. 2, tabl. 1

Keywords

• anti-spoofing
• face identification
• convolutional neural networks
• key points and features
• biometric authentication and identification