Generating Adversarial Examples By Makeup Attacks on Face Recognition

Zhu, Zheng-An; Lu, Yun-Zhong; Chiang, Chen-Kuo

doi:10.1109/icip.2019.8803269

Cited by 57 publications

(37 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As CNNs are increasingly applied in practice, many researchers have begun to focus on the security of CNNs and have studied AEs in areas such as autonomous driving [36] and face recognition [37], [38] and found that AEs reduce the robustness of CNNs. AEs exist not only in computer vision fields, such as image classification [21]- [23], [39]- [41], object detection [42], [43], and semantic segmentation [44], [45], but also in the natural language processing [46], [47] and speech recognition [48], [49].…”

Section: Related Workmentioning

confidence: 99%

Adversarial Examples for CNN-Based SAR Image Classification: An Experience Study

Huang

Chen

et al. 2021

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Section: Related Workmentioning

confidence: 99%

Adversarial Examples for CNN-Based SAR Image Classification: An Experience Study

Huang

Chen

et al. 2021

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

“…Zhu, Lu and Chiang (2019) aims to create negative examples by GAN to attack facial recognition models by applying makeup effects on facial images. In the experimental results, it is seen that the method created according to previous studies creates high quality facial makeup images [59]. Chen, Zhu and Wu (2019) used GAN to expand small-sized infrared datasets in their study.…”

Section: Work With Generative Adversarial Networkmentioning

confidence: 96%

Generative Networks and Royalty-Free Products

Özkan

Erdoğmuş

2020

Sakarya University Journal of Computer and Information Sciences

View full text Add to dashboard Cite

In recent years, with the increasing power of computers and Graphics Processing Units (GPUs), vast variety of deep neural networks architectures have been created and realized. One of the most interesting and generative type of the networks are Generative Adversarial Networks (GANs). GANs are used to create things such as music, images or a film scenerio. GANs consist of two networks working simultaneously. Generative network captures data distribution and discriminative network estimates the probability of the Generative Network output, coming from training data of discriminative network. The objective is to both maximizing the generative network products reality and minimize the discriminative network classification error. This procedure is a minimax two-player game. In this paper, it has been aimed to review the latest studies with GANs, to gather the recent studies in an article and to discuss the possible issues with royalty free products created by GANs. With this aim, from 2018 to today, the studies on GANs have been gathered to the citation numbers. As a result, the recent studies with GANs have been summarized and the potential issues related to GANs have been submitted.

show abstract

“…(ii) Untargeted attacks: The goal of an untargeted attack is to lead the neural network to misclassify the inputs. An attacker can simply employ similar approaches of wearing a mask, glasses [32], makeup [42] or have expressions [22] to impersonate another subject, typically an enrollee within the enrolment dataset.…”

Section: Threat Modelmentioning

confidence: 99%

Adversarial Attacks on Face Recognition Systems

Raja

Ramachandra

et al. 2022

Advances in Computer Vision and Pattern Recognition

View full text Add to dashboard Cite

Face recognition has been widely used for identity verification both in supervised and unsupervised access control applications. The advancement in deep neural networks has opened up the possibility of scaling it to multiple applications. Despite the improvement in performance, deep network-based Face Recognition Systems (FRS) are not well prepared against adversarial attacks at the deployment level. The output performance of such FRS can be drastically impacted simply by changing the trained parameters, for instance, by changing the number of layers, subnetworks, loss and activation functions. This chapter will first demonstrate the impact on biometric performance using a publicly available face dataset. Further to this, this chapter will also present some strategies to defend against such attacks by incorporating defense mechanisms at the training level to mitigate the performance degradation. With the empirical evaluation of the deep FRS with and without a defense mechanism, we demonstrate the impact on biometric performance for the completeness of the chapter.

show abstract

Generating Adversarial Examples By Makeup Attacks on Face Recognition

Cited by 57 publications

References 8 publications

Adversarial Examples for CNN-Based SAR Image Classification: An Experience Study

Adversarial Examples for CNN-Based SAR Image Classification: An Experience Study

Generative Networks and Royalty-Free Products

Adversarial Attacks on Face Recognition Systems

Contact Info

Product

Resources

About