Adversarial Attack and Defense on Discrete Time Dynamic Graphs

被引:1
作者
Zhao, Ziwei [1 ]
Yang, Yu [2 ]
Yin, Zikai [1 ]
Xu, Tong [1 ]
Zhu, Xi [1 ]
Lin, Fake [1 ]
Li, Xueying [3 ]
Chen, Enhong [1 ]
机构
[1] Univ Sci & Technol China, State Key Lab Cognit Intelligence, Hefei 230026, Peoples R China
[2] City Univ Hong Kong, Sch Data Sci, Kowloon Tong, Hong Kong, Peoples R China
[3] Alibaba Grp, Hangzhou, Peoples R China
基金
中国国家自然科学基金;
关键词
Training; Robustness; Perturbation methods; Learning systems; Optimization; Topology; Task analysis; Adversarial attack; dynamic graph representation; graph learning; robust training; OPTIMIZATION; QUERIES;
D O I
10.1109/TKDE.2024.3438238
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Graph learning methods have achieved remarkable performance in various domains such as social recommendation, financial fraud detection, and so on. In real applications, the underlying graph is often dynamically evolving and thus, some recent studies focus on integrating the temporal topology information of graphs into the GNN for learning graph embedding. However, the robustness of training GNNs for dynamic graphs has not been discussed so far. The major reason is how to attack dynamic graph embedding still remains largely untouched, let alone how to defend against the attacks. To enable robust training of GNNs for dynamic graphs, in this paper, we investigate the problem of how to generate attacks and defend against attacks for dynamic graph embedding. Attacking dynamic graph embedding is more challenging than attacking static graph embedding as we need to understand the temporal dynamics of graphs as well as its impact on the embedding and the injected perturbations should be distinguished from the natural evolution. In addition, the defense is very challenging as the perturbations may be hidden within the natural evolution. To tackle these technical challenges, in this paper, we first develop a novel gradient-based attack method from an optimization perspective to generate perturbations to fool dynamic graph learning methods, where a key idea is to use gradient dynamics to attack the natural dynamics of the graph. Further, we borrow the idea of the attack method and integrate it with adversarial training to train a more robust dynamic graph learning method to defend against hand-crafted attacks. Finally, extensive experiments on two real-world datasets demonstrate the effectiveness of the proposed attack and defense method, where our defense method not only achieves comparable performance on clean graphs but also significantly increases the defense performance on attacked graphs.
引用
收藏
页码:7600 / 7611
页数:12
相关论文
共 77 条
[1]   On Sampling from Massive Graph Streams [J].
Ahmed, Nesreen K. ;
Duffield, Nick ;
Willke, Theodore L. ;
Rossi, Ryan A. .
PROCEEDINGS OF THE VLDB ENDOWMENT, 2017, 10 (11) :1430-1441
[2]   Network Sampling: From Static to Streaming Graphs [J].
Ahmed, Nesreen K. ;
Neville, Jennifer ;
Kompella, Ramana .
ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA, 2014, 8 (02)
[3]  
Akoglu L, 2010, LECT NOTES ARTIF INT, V6119, P410
[4]   Statistical mechanics of complex networks [J].
Albert, R ;
Barabási, AL .
REVIEWS OF MODERN PHYSICS, 2002, 74 (01) :47-97
[5]  
[Anonymous], 2012, KDD2012, DOI DOI 10.1145/2339530.2339628
[6]  
[Anonymous], 2013, P 1 ACM C ONL SOC NE
[7]  
Arnaboldi V, 2013, IEEE INFOCOM SER, P3459
[8]  
Bahdanau D, 2016, Arxiv, DOI [arXiv:1409.0473, 10.48550/arXiv.1409.0473, DOI 10.48550/ARXIV.1409.0473]
[9]  
Barrett TD, 2020, AAAI CONF ARTIF INTE, V34, P3251
[10]  
Blenn N, 2012, LECT NOTES COMPUT SC, V7289, P56, DOI 10.1007/978-3-642-30045-5_5