Real-world spoofing detection for human activity recognition applications

Abstract

Military, Police and law enforcement units have a vast amount of data to work with, but they lack the human resource to analyse and classify information. Biometric recognition applications, such as face recognition becomes more and more widespread for the authorities to apply the technology for increasing security efficiency. However, the pandemics influenced wearable objects covering the face, as wearing a face mask became a requirement by law in many countries. Studies revealed that the accuracy of face recognition significantly dropped because of the face masks. The objective is to present a novel field of research based on artificial intelligence and deep neural network teachings, namely, Human Activity Recognition (HAR) and the spoofing possibilities to bypass such automised systems.