Exploratory Adversarial Attacks on Graph Neural Networks for Semi-Supervised Node Classification

doi:10.1016/j.patcog.2022.109042

	Exploratory Adversarial Attacks on Graph Neural Networks for Semi-Supervised Node Classification
	Lin, Xixun 1,6,7; Zhou, Chuan 2,6; Wu, Jia 3; Yang, Hong 4; Wang, Haibo 5; Cao, Yanan 1,6; Wang, Bin 8
刊名	PATTERN RECOGNITION
	2023
卷号	133 页码:12
关键词	Gradient -based attacks Maximal gradient Graph neural networks Semi-supervised node classification
ISSN号	0031-3203
DOI	10.1016/j.patcog.2022.109042
英文摘要	Graph neural networks (GNNs) have been successfully used to analyze non-Euclidean network data. Re-cently, there emerge a number of works to investigate the robustness of GNNs by adding adversarial noises into the graph topology, where the gradient-based attacks are widely studied due to their inherent efficiency and high effectiveness. However, the gradient-based attacks often lead to sub-optimal results due to the discrete structure of graph data. To address this issue, we propose a novel exploratory adver-sarial attack (termed as EpoAtk) to boost the gradient-based perturbations on graphs. The exploratory strategy in EpoAtk includes three phases, generation, evaluation and recombination, with the goal of sidestepping the possible misinformation that the maximal gradient provides. In particular, our evalu-ation phase introduces a self-training objective containing three effective evaluation functions to fully exploit the useful information of unlabeled nodes. EpoAtk is evaluated on multiple benchmark datasets for the task of semi-supervised node classification in different attack settings. Extensive experimental re-sults demonstrate that the proposed method achieves consistent and significant improvements over the state-of-the-art adversarial attacks with the same attack budgets.(c) 2022 Elsevier Ltd. All rights reserved.
资助项目	National Key Re- search and Development Program of China ; NSFC ; ARC DECRA ; CAS Project for Young Scientists in Basic Research ; [2021YFB310 060 0] ; [11688101] ; [61872360] ; [DE20 010 0964] ; [YSBR-0 08]
WOS研究方向	Computer Science ; Engineering
语种	英语
出版者	ELSEVIER SCI LTD
WOS记录号	WOS:000870987900006
内容类型	期刊论文
源URL	[http://ir.amss.ac.cn/handle/2S8OKBNM/60810]
专题	中国科学院数学与系统科学研究院
通讯作者	Zhou, Chuan; Yang, Hong
作者单位	1.Chinese Acad Sci, Inst Informat Engn, Beijing, Peoples R China 2.Chinese Acad Sci, Acad Math & Syst Sci, Beijing, Peoples R China 3.Macquarie Univ, Dept Comp, Sydney, Australia 4.Guangzhou Univ, Cyberspace Inst Adv Technol, Guangzhou, Peoples R China 5.Tsinghua Univ, Dept Comp Sci & Technol, Tsinghua, Peoples R China 6.Univ Chinese Acad Sci, Sch Cyber Secur, Beijing, Peoples R China 7.Baidu Inc, Beijing, Peoples R China 8.Xiaomi AI Lab, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Lin, Xixun,Zhou, Chuan,Wu, Jia,et al. Exploratory Adversarial Attacks on Graph Neural Networks for Semi-Supervised Node Classification[J]. PATTERN RECOGNITION,2023,133:12.
APA	Lin, Xixun.,Zhou, Chuan.,Wu, Jia.,Yang, Hong.,Wang, Haibo.,...&Wang, Bin.(2023).Exploratory Adversarial Attacks on Graph Neural Networks for Semi-Supervised Node Classification.PATTERN RECOGNITION,133,12.
MLA	Lin, Xixun,et al."Exploratory Adversarial Attacks on Graph Neural Networks for Semi-Supervised Node Classification".PATTERN RECOGNITION 133(2023):12.