Towards the transferable audio adversarial attack via ensemble methods
Abstract In recent years, deep learning (DL) models have achieved significant progress in many domains, such as autonomous driving, facial recognition, and speech recognition. However, the vulnerability of deep learning models to adversarial attacks has raised serious concerns in the community becau...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2023-12-01
|
| Series: | Cybersecurity |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s42400-023-00175-8 |
| _version_ | 1797397905618763776 |
|---|---|
| author | Feng Guo Zheng Sun Yuxuan Chen Lei Ju |
| author_facet | Feng Guo Zheng Sun Yuxuan Chen Lei Ju |
| author_sort | Feng Guo |
| collection | DOAJ |
| description | Abstract In recent years, deep learning (DL) models have achieved significant progress in many domains, such as autonomous driving, facial recognition, and speech recognition. However, the vulnerability of deep learning models to adversarial attacks has raised serious concerns in the community because of their insufficient robustness and generalization. Also, transferable attacks have become a prominent method for black-box attacks. In this work, we explore the potential factors that impact adversarial examples (AEs) transferability in DL-based speech recognition. We also discuss the vulnerability of different DL systems and the irregular nature of decision boundaries. Our results show a remarkable difference in the transferability of AEs between speech and images, with the data relevance being low in images but opposite in speech recognition. Motivated by dropout-based ensemble approaches, we propose random gradient ensembles and dynamic gradient-weighted ensembles, and we evaluate the impact of ensembles on the transferability of AEs. The results show that the AEs created by both approaches are valid for transfer to the black box API. |
| first_indexed | 2024-03-09T01:17:59Z |
| format | Article |
| id | doaj.art-cd602e5db44145d3b7fe3ea829ebbbd1 |
| institution | Directory Open Access Journal |
| issn | 2523-3246 |
| language | English |
| last_indexed | 2024-03-09T01:17:59Z |
| publishDate | 2023-12-01 |
| publisher | SpringerOpen |
| record_format | Article |
| series | Cybersecurity |
| spelling | doaj.art-cd602e5db44145d3b7fe3ea829ebbbd12023-12-10T12:22:51ZengSpringerOpenCybersecurity2523-32462023-12-016111710.1186/s42400-023-00175-8Towards the transferable audio adversarial attack via ensemble methodsFeng Guo0Zheng Sun1Yuxuan Chen2Lei Ju3School of Cyber Science and Technology, Shandong UniversitySchool of Cyber Science and Technology, Shandong UniversitySchool of Cyber Science and Technology, Shandong UniversitySchool of Cyber Science and Technology, Shandong UniversityAbstract In recent years, deep learning (DL) models have achieved significant progress in many domains, such as autonomous driving, facial recognition, and speech recognition. However, the vulnerability of deep learning models to adversarial attacks has raised serious concerns in the community because of their insufficient robustness and generalization. Also, transferable attacks have become a prominent method for black-box attacks. In this work, we explore the potential factors that impact adversarial examples (AEs) transferability in DL-based speech recognition. We also discuss the vulnerability of different DL systems and the irregular nature of decision boundaries. Our results show a remarkable difference in the transferability of AEs between speech and images, with the data relevance being low in images but opposite in speech recognition. Motivated by dropout-based ensemble approaches, we propose random gradient ensembles and dynamic gradient-weighted ensembles, and we evaluate the impact of ensembles on the transferability of AEs. The results show that the AEs created by both approaches are valid for transfer to the black box API.https://doi.org/10.1186/s42400-023-00175-8Adversarial attacksDynamic gradient weightingTransferabilityEnsemble methods |
| spellingShingle | Feng Guo Zheng Sun Yuxuan Chen Lei Ju Towards the transferable audio adversarial attack via ensemble methods Cybersecurity Adversarial attacks Dynamic gradient weighting Transferability Ensemble methods |
| title | Towards the transferable audio adversarial attack via ensemble methods |
| title_full | Towards the transferable audio adversarial attack via ensemble methods |
| title_fullStr | Towards the transferable audio adversarial attack via ensemble methods |
| title_full_unstemmed | Towards the transferable audio adversarial attack via ensemble methods |
| title_short | Towards the transferable audio adversarial attack via ensemble methods |
| title_sort | towards the transferable audio adversarial attack via ensemble methods |
| topic | Adversarial attacks Dynamic gradient weighting Transferability Ensemble methods |
| url | https://doi.org/10.1186/s42400-023-00175-8 |
| work_keys_str_mv | AT fengguo towardsthetransferableaudioadversarialattackviaensemblemethods AT zhengsun towardsthetransferableaudioadversarialattackviaensemblemethods AT yuxuanchen towardsthetransferableaudioadversarialattackviaensemblemethods AT leiju towardsthetransferableaudioadversarialattackviaensemblemethods |