ARGAN: Adversarially Robust Generative Adversarial Networks for Deep Neural Networks Against Adversarial Examples
An adversarial example, which is an input instance with small, intentional feature perturbations to machine learning models, represents a concrete problem in Artificial intelligence safety. As an emerging defense method to defend against adversarial examples, generative adversarial networks-based de...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2022-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9737142/ |
| _version_ | 1811328217127059456 |
|---|---|
| author | Seok-Hwan Choi Jin-Myeong Shin Peng Liu Yoon-Ho Choi |
| author_facet | Seok-Hwan Choi Jin-Myeong Shin Peng Liu Yoon-Ho Choi |
| author_sort | Seok-Hwan Choi |
| collection | DOAJ |
| description | An adversarial example, which is an input instance with small, intentional feature perturbations to machine learning models, represents a concrete problem in Artificial intelligence safety. As an emerging defense method to defend against adversarial examples, generative adversarial networks-based defense methods have recently been studied. However, the performance of the state-of-the-art generative adversarial networks-based defense methods is limited because the target deep neural network models with generative adversarial networks-based defense methods are robust against <italic>adversarial examples</italic> but make a false decision for <italic>legitimate input data</italic>. To solve the accuracy degradation of the generative adversarial networks-based defense methods for <italic>legitimate input data</italic>, we propose a new generative adversarial networks-based defense method, which is called Adversarially Robust Generative Adversarial Networks(ARGAN). While converting input data to machine learning models using the two-step transformation architecture, ARGAN learns the generator model to reflect the vulnerability of the target deep neural network model against adversarial examples and optimizes parameter values of the generator model for a joint loss function. From the experimental results under various datasets collected from diverse applications, we show that the accuracy of ARGAN for <italic>legitimate input data</italic> is good-enough while keeping the target deep neural network model robust against <italic>adversarial examples</italic>. We also show that the accuracy of ARGAN outperforms the accuracy of the state-of-the-art generative adversarial networks-based defense methods. |
| first_indexed | 2024-04-13T15:21:17Z |
| format | Article |
| id | doaj.art-8ab9dc9351d0496ca335ae7c651cb5d6 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-04-13T15:21:17Z |
| publishDate | 2022-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-8ab9dc9351d0496ca335ae7c651cb5d62022-12-22T02:41:39ZengIEEEIEEE Access2169-35362022-01-0110336023361510.1109/ACCESS.2022.31602839737142ARGAN: Adversarially Robust Generative Adversarial Networks for Deep Neural Networks Against Adversarial ExamplesSeok-Hwan Choi0https://orcid.org/0000-0003-3590-6024Jin-Myeong Shin1Peng Liu2https://orcid.org/0000-0002-5091-8464Yoon-Ho Choi3https://orcid.org/0000-0002-3556-5082School of Computer Science and Engineering, Pusan National University, Busan, Republic of KoreaSchool of Computer Science and Engineering, Pusan National University, Busan, Republic of KoreaCollege of Information Sciences and Technology, Pennsylvania State University, State College, PA, USASchool of Computer Science and Engineering, Pusan National University, Busan, Republic of KoreaAn adversarial example, which is an input instance with small, intentional feature perturbations to machine learning models, represents a concrete problem in Artificial intelligence safety. As an emerging defense method to defend against adversarial examples, generative adversarial networks-based defense methods have recently been studied. However, the performance of the state-of-the-art generative adversarial networks-based defense methods is limited because the target deep neural network models with generative adversarial networks-based defense methods are robust against <italic>adversarial examples</italic> but make a false decision for <italic>legitimate input data</italic>. To solve the accuracy degradation of the generative adversarial networks-based defense methods for <italic>legitimate input data</italic>, we propose a new generative adversarial networks-based defense method, which is called Adversarially Robust Generative Adversarial Networks(ARGAN). While converting input data to machine learning models using the two-step transformation architecture, ARGAN learns the generator model to reflect the vulnerability of the target deep neural network model against adversarial examples and optimizes parameter values of the generator model for a joint loss function. From the experimental results under various datasets collected from diverse applications, we show that the accuracy of ARGAN for <italic>legitimate input data</italic> is good-enough while keeping the target deep neural network model robust against <italic>adversarial examples</italic>. We also show that the accuracy of ARGAN outperforms the accuracy of the state-of-the-art generative adversarial networks-based defense methods.https://ieeexplore.ieee.org/document/9737142/Adversarial examplesadversarial perturbationdeep neural networks (DNNs)security |
| spellingShingle | Seok-Hwan Choi Jin-Myeong Shin Peng Liu Yoon-Ho Choi ARGAN: Adversarially Robust Generative Adversarial Networks for Deep Neural Networks Against Adversarial Examples IEEE Access Adversarial examples adversarial perturbation deep neural networks (DNNs) security |
| title | ARGAN: Adversarially Robust Generative Adversarial Networks for Deep Neural Networks Against Adversarial Examples |
| title_full | ARGAN: Adversarially Robust Generative Adversarial Networks for Deep Neural Networks Against Adversarial Examples |
| title_fullStr | ARGAN: Adversarially Robust Generative Adversarial Networks for Deep Neural Networks Against Adversarial Examples |
| title_full_unstemmed | ARGAN: Adversarially Robust Generative Adversarial Networks for Deep Neural Networks Against Adversarial Examples |
| title_short | ARGAN: Adversarially Robust Generative Adversarial Networks for Deep Neural Networks Against Adversarial Examples |
| title_sort | argan adversarially robust generative adversarial networks for deep neural networks against adversarial examples |
| topic | Adversarial examples adversarial perturbation deep neural networks (DNNs) security |
| url | https://ieeexplore.ieee.org/document/9737142/ |
| work_keys_str_mv | AT seokhwanchoi arganadversariallyrobustgenerativeadversarialnetworksfordeepneuralnetworksagainstadversarialexamples AT jinmyeongshin arganadversariallyrobustgenerativeadversarialnetworksfordeepneuralnetworksagainstadversarialexamples AT pengliu arganadversariallyrobustgenerativeadversarialnetworksfordeepneuralnetworksagainstadversarialexamples AT yoonhochoi arganadversariallyrobustgenerativeadversarialnetworksfordeepneuralnetworksagainstadversarialexamples |