Techniques and Challenges for Generation and Detection Face Morphing Attacks: A Survey

Abstract

      Face recognition system is the most widely used application in the field of security and especially in border control. This system may be exposed to direct or indirect attacks through the use of face morphing attacks (FMAs). Face morphing attacks is the process of producing a passport photo resulting from a mixture of two images, one of which is for an ordinary person and the other is a judicially required. In this case, a face recognition system may allow travel of  persons not permitted to travel through face morphing image in a Machine-Readable Electronic Travel Document (eMRTD) or electronic passport at Automatic Border Control (ABC) gates. In creating an electronic passport, most countries rely on applicant to submit images in a form of a document or via the Internet, and this allows applicants to manipulate the images to produce morphing images. These photos allow both beneficial and harmful partners to cross borders using the same passport. This is considered a major threat to the security systems that allow them to travel without revealing their true identity. This paper aims to provide a comprehensive overview of face morphing attacks and the development taking place in this specialty. This paper describes the techniques for generating metamorphic images and challenges they face, in addition to the advantages and disadvantages of these techniques. It also dealt with types of techniques used in detecting and determining the attack of mutant faces in the field of deep learning or machine learning, in addition to the laws and criteria for measuring the efficiency of the algorithms used. It provides a general summary of the work that has been produced in this field.