Towards Robust Semantic Segmentation against Patch-Based Attack via Attention Refinement

doi:10.1007/s11263-024-02120-9

	Towards Robust Semantic Segmentation against Patch-Based Attack via Attention Refinement
	Yuan, Zheng 1,2,3; Zhang, Jie 1,2,3; Wang, Yude 1,2,3; Shan, Shiguang 1,2,3; Chen, Xilin 1,2,3
	2024-06-07
发表期刊	INTERNATIONAL JOURNAL OF COMPUTER VISION
ISSN	0920-5691
页码	23
摘要	The attention mechanism has been proven effective on various visual tasks in recent years. In the semantic segmentation task, the attention mechanism is applied in various methods, including the case of both convolution neural networks and vision transformer as backbones. However, we observe that the attention mechanism is vulnerable to patch-based adversarial attacks. Through the analysis of the effective receptive field, we attribute it to the fact that the wide receptive field brought by global attention may lead to the spread of the adversarial patch. To address this issue, in this paper, we propose a robust attention mechanism (RAM) to improve the robustness of the semantic segmentation model, which can notably relieve the vulnerability against patch-based attacks. Compared to the vallina attention mechanism, RAM introduces two novel modules called max attention suppression and random attention dropout, both of which aim to refine the attention matrix and limit the influence of a single adversarial patch on the semantic segmentation results of other positions. Extensive experiments demonstrate the effectiveness of our RAM to improve the robustness of semantic segmentation models against various patch-based attack methods under different attack settings.
关键词	Model robustness Attention mechanism Semantic segmentation Patch-based attack
DOI	10.1007/s11263-024-02120-9
收录类别	SCI
语种	英语
资助项目	National Key R &D Program of China ; Strategic Priority Research Program of the Chinese Academy of Sciences[XDB0680000] ; Beijing Nova Program[20230484368] ; National Natural Science Foundation of China[62176251] ; Youth Innovation Promotion Association CAS ; [2021YFC3310100]
WOS研究方向	Computer Science
WOS类目	Computer Science, Artificial Intelligence
WOS记录号	WOS:001240467000001
出版者	SPRINGER
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/40033
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Zhang, Jie
作者单位	1.Chinese Acad Sci, Key Lab Intelligent Informat Proc, Inst Comp Technol, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Beijing 100049, Peoples R China 3.Chinese Acad Sci, Key Lab AI Safety, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Yuan, Zheng,Zhang, Jie,Wang, Yude,et al. Towards Robust Semantic Segmentation against Patch-Based Attack via Attention Refinement[J]. INTERNATIONAL JOURNAL OF COMPUTER VISION,2024:23.
APA	Yuan, Zheng,Zhang, Jie,Wang, Yude,Shan, Shiguang,&Chen, Xilin.(2024).Towards Robust Semantic Segmentation against Patch-Based Attack via Attention Refinement.INTERNATIONAL JOURNAL OF COMPUTER VISION,23.
MLA	Yuan, Zheng,et al."Towards Robust Semantic Segmentation against Patch-Based Attack via Attention Refinement".INTERNATIONAL JOURNAL OF COMPUTER VISION (2024):23.