Publication:
A defense and detection against adversarial attack using De-noising auto-encoder and super resolution GAN
A defense and detection against adversarial attack using De-noising auto-encoder and super resolution GAN
| dc.contributor.author | Md Maruf Hassan | |
| dc.contributor.author | Subroto Karmokar | |
| dc.date.accessioned | 2025-07-16T04:17:10Z | |
| dc.date.available | 2025-07-16T04:17:10Z | |
| dc.date.issued | 2023 | |
| dc.description.abstract | Neural networks have flourished in heterogeneous industries to automate tasks that evince it being an utmost priority for the adopters. The adversarial attack poses a threat for Deep Neural Networks and their variants. This attack is designed such that it adds adversarial noise to an image. Several such techniques can be found in contemporary research capable of corrupting neural networks leading to misclassification. Various defense mechanisms have been purported and built with Deep Neural Networks to defend and increase the robustness of the primary classifier neural network model. However, models accommodating high-resolution image data and pre-trained neural network classifiers are sparse. This research develops a model that can be integrated with any existing trained neural network, establishing a generic line of defense against adversarial attacks. The proposed model detects highly distorted images, repudiating to avoid misclassification. Additionally, this work is intended to work with high-resolution adversarial image samples. The ADDA-Adv. tool restores the adversarial samples and provides better accuracy as compared to recent works. | |
| dc.identifier.doi | 10.1063/5.0127810 | |
| dc.identifier.uri | https://pubs.aip.org/aip/acp/article-abstract/2608/1/020028/2895761/A-defense-and-detection-against-adversarial-attack?redirectedFrom=fulltext | |
| dc.identifier.uri | https://pubs.aip.org/aip | |
| dc.identifier.uri | https://hdl.handle.net/20.500.14170/14113 | |
| dc.language.iso | en | |
| dc.publisher | AIP Publishing | |
| dc.relation.ispartof | AIP Conference Proceedings | |
| dc.relation.ispartofseries | 2nd International Recent Trends in Engineering Advance Computing and Technology Conference (RETREAT2021) | |
| dc.relation.issn | 0094-243X | |
| dc.subject | Neural networks | |
| dc.title | A defense and detection against adversarial attack using De-noising auto-encoder and super resolution GAN | |
| dc.type | Resource Types::text::conference output::conference proceedings | |
| dspace.entity.type | Publication | |
| oaire.citation.issue | 1 | |
| oaire.citation.volume | 2608 | |
| oairecerif.author.affiliation | Universiti Malaysia Perlis | |
| oairecerif.author.affiliation | Daffodil International University, Dhaka, Bangladesh |
Files
Original bundle
1 - 1 of 1
- Name:
- A defense and detection against adversarial attack using De-noising auto-encoder and super resolution GAN.pdf
- Size:
- 96.67 KB
- Format:
- Adobe Portable Document Format
- Description: