Biometric Systems Anti-Spoofing using Innovated Machine Learning Techniques

Abstract

Biometric recognition has been increasingly used in recent applications that need authentication and verification. However, with the increase of technology, spoofing these biometric traits has become more easy. For example, an attacker can use a previously recorded photograph or a video replay of the face or eye region of a person to gain access to his/her smartphone or any application that needs authorization, this is known as Presentation Attack (PA). This has led to the increase of interest in developing reliable Presentation Attack Detection (PAD) algorithms. In this thesis, we survey current state-of-the-art solutions for PAD, highlight their problems, and propose several approaches to alleviate these problems and produce generalized presentation attack detection models. The thesis is divided into six chapters as listed below: Chapter 1: Introduction In this chapter we introduce the problem of biometric presentation attack detection, describes its broad applications and the challenges accompanied with its solutions. We then present an overview of our proposed approach together with a summary of the contributions introduced in this thesis.