FocusedCleaner: Sanitizing Poisoned Graphs for Robust GNN-Based Node Classification
IEEE Transactions on Knowledge and Data Engineering · 2024-06 · 36 (6) · 2476-2489
DOI: 10.1109/tkde.2023.3322129 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Knowledge and Data Engineering, 2024-06. Verfasst von Yulin Zhu und 4 weiteren Autor:innen. Die Arbeit wurde bisher 20-mal zitiert (Crossref) und nennt 56 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Hong Kong RGC.
Abstract (Originalsprache)
Graph Neural Networks (GNNs) are vulnerable to data poisoning attacks, which will generate a poisoned graph as the input to the GNN models. We present FocusedCleaner as a poisoned graph sanitizer to effectively identify the poison injected by attackers. Specifically, FocusedCleaner provides a sanitation framework consisting of two modules: bi-level structural learning and victim node detection. In particular, the structural learning module will reverse the attack process to steadily sanitize the graph while the detection module provides the “focus” – a narrowed and more accurate search region – to structural learning. These two modules will operate in iterations and reinforce each other to sanitize a poisoned graph step by step. As an important application, we show that the adversarial robustness of GNNs trained over the sanitized graph for the node classification task is significantly improved. Extensive experiments demonstrate that FocusedCleaner outperforms the state-of-the-art baselines both on poisoned graph sanitation and improving robustness.
Förderung
- National Natural Science Foundation of China
- Hong Kong RGC
Zitieren
APA
Zhu, Y., Tong, L., Li, G., Luo, X., & Zhou, K. (2024). FocusedCleaner: Sanitizing Poisoned Graphs for Robust GNN-Based Node Classification. IEEE Transactions on Knowledge and Data Engineering, 36(6), 2476-2489. https://doi.org/10.1109/tkde.2023.3322129
Vancouver
Zhu Y, Tong L, Li G, Luo X, Zhou K. FocusedCleaner: Sanitizing Poisoned Graphs for Robust GNN-Based Node Classification. IEEE Transactions on Knowledge and Data Engineering. 2024;36(6):2476-2489. doi:10.1109/tkde.2023.3322129
BibTeX
@article{Zhu2024,
author = {Zhu, Yulin and Tong, Liang and Li, Gaolei and Luo, Xiapu and Zhou, Kai},
title = {FocusedCleaner: Sanitizing Poisoned Graphs for Robust GNN-Based Node Classification},
journal = {IEEE Transactions on Knowledge and Data Engineering},
year = {2024},
volume = {36},
number = {6},
pages = {2476--2489},
doi = {10.1109/tkde.2023.3322129}
}
EUVIMED