Dynamic facial presentation attack detection for automated border control systems

Abstract

Millions of passengers travel every day, border crossing being one of their most common activities. At these points it is extremely important that security is completely guaranteed. However, the maintaining adequate security levels is a very demanding issue. This has promoted the development of systems capable of providing support to border authorities by automatising some of their tasks. Thus, Automated Border Control (ABC) systems have become a key tool. These systems increase the flow of travellers as they can achieve fast evaluations of individuals through their machine-readable travel documents. However, this has motivated the appearance of attacks that try to avoid the identity detection of individuals by these systems. Presentation Attack Detection (PAD) algorithms have emerged to mitigate such a problem. This paper presents the On-the-Fly Presentation Attack Detection (FlyPAD) framework that implements a set of dynamic PAD techniques. It allows multiple types of attacks to be detected as the traveller approaches the ABC system, rather than being static in front of the cameras. Several experiments have been carried out, both in the laboratory and in real environments, obtaining promising results.