Page 5 - IJEEE-2022-Vol18-ISSUE-1

Received: 24 September 2021 Revised: 30 October 2021 Accepted: 30 October 2021
DOI: 10.37917/ijeee.18.1.1
Vol. 18| Issue 1| June 2022
? Open Access

Iraqi Journal for Electrical and Electronic Engineering

Original Article

Face Recognition System Against Adversarial Attack
Using Convolutional Neural Network

Ansam Kadhim, Salah Al-Darraji
Department of Computer Science, College of Education for Pure Sciences, University of Basrah, Iraq

Correspondence
*Ansam Kadhim
Computer Science Department,
College of Education for Pure Sciences,
University of Basrah, Basrah, Iraq
Email: pgs2181@uobasrah.edu.iq

Abstract
Face recognition is the technology that verifies or recognizes faces from images, videos, or real-time streams. It can be used in
security or employee attendance systems. Face recognition systems may encounter some attacks that reduce their ability to
recognize faces properly. So, many noisy images mixed with original ones lead to confusion in the results. Various attacks that
exploit this weakness affect the face recognition systems such as Fast Gradient Sign Method (FGSM), Deep Fool, and
Projected Gradient Descent (PGD). This paper proposes a method to protect the face recognition system against these attacks
by distorting images through different attacks, then training the recognition deep network model, specifically Convolutional
Neural Network (CNN), using the original and distorted images. Diverse experiments have been conducted using combinations
of original and distorted images to test the effectiveness of the system. The system showed an accuracy of 93% using FGSM
attack, 97% using deep fool, and 95% using PGD.
KEYWORDS: Face Recognition, Convolutional Neural Network, Adversarial Attacks.

I. INTRODUCTION medicine, and security issues. FR technology is considered
one of the most important methods that deal with images of
Face Recognition (FR) is a technique that is used to faces for different people. This technology is compatible to
recognize faces in images and videos using various discover any noises in faces [6] using CNN algorithm.
algorithms. As the face is the most important identification Therefore, it decreases noises from most images of faces for
part in the body of a human, it is useful in many fields for the train and test dataset.
people's identification, such as in airports for security issues.
Therefore, face recognition is necessary for such It is not possible to tell the difference between a real
applications. Many factors affect the clarity of the face such face and an image of a face and cannot be easily recognized
as resolution, illumination, and facial expressions. Noise by machine learning algorithms. Therefore, biometric
also plays a negative role in faking faces. The technology of sensors can improve recognition accuracy. The advantage of
face recognition tries to remove noises from faces to get this technology is to enhance security and social
higher accuracy. Consequently, it discovers the original environments. It can be used in online banking and medical
images to enhance the results using suitable algorithms for records for Personal Identification. However, the CNN
this purpose. algorithm is used for this purpose in many areas. In some
applications, where the recognition accuracy is required to
Hence, it is reasonable that most companies use this be high, some factors may affect the recognition, such as
technology to get to know their staff and avoid strangers, intentionally added noise. When the noise was added, the
especially, if the number of employees is high. This classification of the input image was wrong, as explained by
technology is related to computer programming and gives Szegedy et al. [7]. To recognize faces, different systems can
full information about a person rapidly. The technology of be used, and these systems are required to eliminate noise in
face recognition was used in most popular regions, for faces.
example, 98 countries use this technology. So, the defense
techniques by using layers in the CNN algorithm were used This paper aims at identifying the problem of noisy faces
to increase adversarial training [1, 2, 3, 4, and 5] on dataset (unclear faces) and adversarial images. FR technique is a
training. This helps them to get results as soon as possible, difficult process, especially if the images are blurry or
especially, in the airport, passengers, traveling, working, unclear. Therefore, FR with a suitable algorithm can be used

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and
reproduction in any medium, provided the original work is properly cited.
© 2021 The Authors. Iraqi Journal for Electrical and Electronic Engineering by College of Engineering, University of Basrah.

https://doi.org/10.37917/ijeee.18.1.1 https://www.ijeee.edu.iq 1