Home // ICIMP 2021, The Sixteenth International Conference on Internet Monitoring and Protection // View article

Methods to Prevent Registration Using Fake Face Images

Authors:
Luis Carabe
Eduardo Cermeño

Keywords: Access control; biometrics; deep learning;FaceNet; face recognition; identification; morphing; security;spoofing attack

Abstract:
Face identification is increasingly being used to register and access specific applications and online services. This opens up new possibilities for malicious attacks, such as users registering multiple times with different images or impersonating other users. Morphing is often the preferred method for these attacks as it allows the physical features of a subject to be progressively modified to resemble another subject. Publications focus on impersonating this other person, usually, someone who is allowed access to a restricted area or software app. However, there is no such list of authorized people in many other applications, just a blacklist of people who cannot enter, log in, or register again. In such cases, the morphing target person is not relevant as the criminal’s main objective is to minimizethe probability of being detected. We present a comparison ofthe identification rate and behavior of 5 recognizers (Eigenfaces, Fisherfaces, Local Binary Patterns Histograms, Scale-invariant feature Transform, and FaceNet) against morphing attacks. We also show the performance that a morphing detector couldachieve. We prove that the use of FaceNet along with a morphing detector is an optimal resource to maintain a high-level ofsecurity, identification rate, and attack detection.

Pages: 21 to 30

Copyright: Copyright (c) IARIA, 2021

Publication date: May 30, 2021

Published in: conference

ISSN: 2308-3980

ISBN: 978-1-61208-862-4

Location: Valencia, Spain

Dates: from May 30, 2021 to June 3, 2021