Link-Backdoor: Backdoor Attack on Link Prediction via Node Injection

被引:4
|
作者
Zheng, Haibin [1 ,2 ]
Xiong, Haiyang [2 ]
Ma, Haonan [2 ]
Huang, Guohan [2 ]
Chen, Jinyin [1 ,2 ]
机构
[1] Zhejiang Univ Technol, Inst Cyberspace Secur, Hangzhou 310023, Peoples R China
[2] Zhejiang Univ Technol, Coll Informat Engn, Hangzhou 310023, Peoples R China
关键词
Backdoor attack; defense; gradient information; link prediction; node injection;
D O I
10.1109/TCSS.2023.3260833
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
Link prediction, inferring the undiscovered or potential links of the graph, is widely applied in the real-world. By facilitating labeled links of the graph as the training data, numerous deep learning based link prediction methods have been studied, which have dominant prediction accuracy compared with non-deep methods. However,the threats of maliciously crafted training graph will leave a specific backdoor in the deep model, thus when some specific examples are fed into the model, it will make wrong prediction, defined as backdoor attack. It is an important aspect that has been overlooked in the current literature. In this paper, we prompt the concept of backdoor attack on link prediction, and propose Link-Backdoor to reveal the training vulnerability of the existing link prediction methods. Specifically, the Link-Backdoor combines the fake nodes with the nodes of the target link to form a trigger. Moreover, it optimizes the trigger by the gradient information from the target model. Consequently, the link prediction model trained on the backdoored dataset will predict the link with trigger to the target state. Extensive experiments on five benchmark datasets and five well-performing link prediction models demonstrate that the Link-Backdoor achieves the state-of-the-art attack success rate under both white-box (i.e., available of the target model parameter)and black-box (i.e., unavailable of the target model parameter) scenarios. Additionally, we testify the attack under defensive circumstance, and the results indicate that the Link-Backdoor still can construct successful attack on the well-performing link prediction methods. The code and data are available at https://github.com/Seaocn/Link-Backdoor.
引用
收藏
页码:1816 / 1831
页数:16
相关论文
共 50 条
  • [1] Dyn-Backdoor: Backdoor Attack on Dynamic Link Prediction
    Chen, Jinyin
    Xiong, Haiyang
    Zheng, Haibin
    Zhang, Jian
    Liu, Yi
    IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING, 2024, 11 (01): : 525 - 542
  • [2] Backdoor attack detection via prediction trustworthiness assessment
    Zhong, Nan
    Qian, Zhenxing
    Zhang, Xinpeng
    INFORMATION SCIENCES, 2024, 662
  • [3] Node Injection Link Stealing Attack
    Zari, Oualid
    Parra-Arnau, Javier
    Unsal, Ayse
    Onen, Melek
    PRIVACY IN STATISTICAL DATABASES, PSD 2024, 2024, 14915 : 358 - 373
  • [4] Motif-Backdoor: Rethinking the Backdoor Attack on Graph Neural Networks via Motifs
    Zheng, Haibin
    Xiong, Haiyang
    Chen, Jinyin
    Ma, Haonan
    Huang, Guohan
    IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS, 2024, 11 (02): : 2479 - 2493
  • [5] Textual Backdoor Attack via Keyword Positioning
    Chen, Depeng
    Mao, Fangfang
    Jin, Hulin
    Cui, Jie
    ADVANCED INTELLIGENT COMPUTING TECHNOLOGY AND APPLICATIONS, PT X, ICIC 2024, 2024, 14871 : 55 - 66
  • [6] Conditional Backdoor Attack via JPEG Compression
    Duan, Qiuyu
    Hua, Zhongyun
    Liao, Qing
    Zhang, Yushu
    Zhang, Leo Yu
    THIRTY-EIGHTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 38 NO 18, 2024, : 19823 - 19831
  • [7] Inaudible Backdoor Attack via Stealthy Frequency Trigger Injection in Audio Spectrogram
    Zhang, Tianfang
    Huy Phan
    Tang, Zijie
    Shi, Cong
    Wang, Yan
    Yuan, Bo
    Chen, Yingying
    PROCEEDINGS OF THE THIRTIETH INTERNATIONAL CONFERENCE ON MOBILE COMPUTING AND NETWORKING, ACM MOBICOM 2024, 2024, : 31 - 45
  • [8] LinkBreaker: Breaking the Backdoor-Trigger Link in DNNs via Neurons Consistency Check
    Chen, Zhenzhu
    Wang, Shang
    Fu, Anmin
    Gao, Yansong
    Yu, Shui
    Deng, Robert H.
    IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY, 2022, 17 : 2000 - 2014
  • [9] TridentShell: a Covert and Scalable Backdoor Injection Attack on Web Applications
    Yu, Xiaobo
    Meng, Weizhi
    Zhao, Lei
    Liu, Yining
    Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2021, 13118 LNCS : 177 - 194
  • [10] TridentShell: A Covert and Scalable Backdoor Injection Attack on Web Applications
    Yu, Xiaobo
    Meng, Weizhi
    Zhao, Lei
    Liu, Yining
    INFORMATION SECURITY (ISC 2021), 2021, 13118 : 177 - 194