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
  • [21] Backdoor Attack on Deep Neural Networks Triggered by Fault Injection Attack on Image Sensor Interface
    Oyama, Tatsuya
    Okura, Shunsuke
    Yoshida, Kota
    Fujino, Takeshi
    SENSORS, 2023, 23 (10)
  • [22] BABE: Backdoor attack with bokeh effects via latent separation suppression
    Li, Junjian
    Chen, Honglong
    Gao, Yudong
    Guo, Shaozhong
    Lin, Kai
    Liu, Yuping
    Sun, Peng
    ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, 2024, 138
  • [23] FIBA: Frequency-Injection based Backdoor Attack in Medical Image Analysis
    Feng, Yu
    Ma, Benteng
    Zhang, Jing
    Zhao, Shanshan
    Xia, Yong
    Tao, Dacheng
    2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022), 2022, : 20844 - 20853
  • [24] Adversarial Backdoor Attack by Naturalistic Data Poisoning on Trajectory Prediction in Autonomous Driving
    Pourkeshavarz, Mozhgan
    Sabokrou, Mohammad
    Rasouli, Amir
    2024 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2024, : 14885 - 14894
  • [25] Fake the Real: Backdoor Attack on Deep Speech Classification via Voice Conversion
    Ye, Zhe
    Mao, Terui
    Dong, Li
    Yan, Diqun
    INTERSPEECH 2023, 2023, : 4923 - 4927
  • [26] Hidden Trigger Backdoor Attack on NLP Models via Linguistic Style Manipulation
    Pan, Xudong
    Zhang, Mi
    Sheng, Beina
    Zhu, Jiaming
    Yang, Min
    PROCEEDINGS OF THE 31ST USENIX SECURITY SYMPOSIUM, 2022, : 3611 - 3628
  • [27] Graph attention network via node similarity for link prediction
    Yang, Kai
    Liu, Yuan
    Zhao, Zijuan
    Zhou, Xingxing
    Ding, Peijin
    EUROPEAN PHYSICAL JOURNAL B, 2023, 96 (03):
  • [28] Graph attention network via node similarity for link prediction
    Kai Yang
    Yuan Liu
    Zijuan Zhao
    Xingxing Zhou
    Peijin Ding
    The European Physical Journal B, 2023, 96
  • [29] SupRTE: Suppressing Backdoor Injection in Federated Learning via Robust Trust Evaluation
    Huang, Wenkai
    Li, Gaolei
    Yi, Xiaoyu
    Li, Jianhua
    Zhao, Chengcheng
    Yin, Ying
    IEEE INTELLIGENT SYSTEMS, 2024, 39 (05) : 66 - 77
  • [30] Stealthy Backdoor Attack Against Speaker Recognition Using Phase-Injection Hidden Trigger
    Ye, Zhe
    Yan, Diqun
    Dong, Li
    Deng, Jiacheng
    Yu, Shui
    IEEE SIGNAL PROCESSING LETTERS, 2023, 30 : 1057 - 1061