Exploring the vulnerabilities and enhancing the adversarial robustness of deep neural networks
Deep learning, especially deep neural networks (DNNs), is at the heart of the current rise of artificial intelligence, and the major breakthroughs in the last few years have been made by DNNs. It has been demonstrated in recent works that DNNs are vulnerable to human-crafted adversarial examples, wh...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Doctor of Philosophy |
| Language: | English |
| Published: |
Nanyang Technological University
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/160963 |
| _version_ | 1826124921229541376 |
|---|---|
| author | Bai, Tao |
| author2 | Jun Zhao |
| author_facet | Jun Zhao Bai, Tao |
| author_sort | Bai, Tao |
| collection | NTU |
| description | Deep learning, especially deep neural networks (DNNs), is at the heart of the current rise of artificial intelligence, and the major breakthroughs in the last few years have been made by DNNs. It has been demonstrated in recent works that DNNs are vulnerable to human-crafted adversarial examples, which look normal in human eyes. Such adversarial instances can fool and mislead DNNs to misbehave as adversaries expected, causing serious consequences for various DNN-based applications in our daily life. To this end, this thesis dedicates to revealing the vulnerabilities of deep learning algorithms and developing defense strategies for combating adversaries effectively. We study current DNNs from the perspective of security with two sides: attack and defense. On the attack front, we explore the possibility of attacks against DNNs during test time with two types of adversarial examples: adversarial perturbations and adversarial patches. On the defense front, we develop solutions to defend against adversarial examples and investigate the robustness-preserving distillation techniques. |
| first_indexed | 2024-10-01T06:28:16Z |
| format | Thesis-Doctor of Philosophy |
| id | ntu-10356/160963 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T06:28:16Z |
| publishDate | 2022 |
| publisher | Nanyang Technological University |
| record_format | dspace |
| spelling | ntu-10356/1609632022-09-01T02:33:19Z Exploring the vulnerabilities and enhancing the adversarial robustness of deep neural networks Bai, Tao Jun Zhao Wen Bihan School of Computer Science and Engineering junzhao@ntu.edu.sg, bihan.wen@ntu.edu.sg Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision Deep learning, especially deep neural networks (DNNs), is at the heart of the current rise of artificial intelligence, and the major breakthroughs in the last few years have been made by DNNs. It has been demonstrated in recent works that DNNs are vulnerable to human-crafted adversarial examples, which look normal in human eyes. Such adversarial instances can fool and mislead DNNs to misbehave as adversaries expected, causing serious consequences for various DNN-based applications in our daily life. To this end, this thesis dedicates to revealing the vulnerabilities of deep learning algorithms and developing defense strategies for combating adversaries effectively. We study current DNNs from the perspective of security with two sides: attack and defense. On the attack front, we explore the possibility of attacks against DNNs during test time with two types of adversarial examples: adversarial perturbations and adversarial patches. On the defense front, we develop solutions to defend against adversarial examples and investigate the robustness-preserving distillation techniques. Doctor of Philosophy 2022-08-11T01:54:38Z 2022-08-11T01:54:38Z 2022 Thesis-Doctor of Philosophy Bai, T. (2022). Exploring the vulnerabilities and enhancing the adversarial robustness of deep neural networks. Doctoral thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/160963 https://hdl.handle.net/10356/160963 10.32657/10356/160963 en This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0). application/pdf Nanyang Technological University |
| spellingShingle | Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision Bai, Tao Exploring the vulnerabilities and enhancing the adversarial robustness of deep neural networks |
| title | Exploring the vulnerabilities and enhancing the adversarial robustness of deep neural networks |
| title_full | Exploring the vulnerabilities and enhancing the adversarial robustness of deep neural networks |
| title_fullStr | Exploring the vulnerabilities and enhancing the adversarial robustness of deep neural networks |
| title_full_unstemmed | Exploring the vulnerabilities and enhancing the adversarial robustness of deep neural networks |
| title_short | Exploring the vulnerabilities and enhancing the adversarial robustness of deep neural networks |
| title_sort | exploring the vulnerabilities and enhancing the adversarial robustness of deep neural networks |
| topic | Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision |
| url | https://hdl.handle.net/10356/160963 |
| work_keys_str_mv | AT baitao exploringthevulnerabilitiesandenhancingtheadversarialrobustnessofdeepneuralnetworks |