EUVIMEDEUVIMED entdecken
Crossref

SOPA: Sensitivity-Oriented Poisoning Attack for Self-Supervised Graph Embedding Model via Bilevel Evolutionary Optimization

Shen You, Kai Zhou, Zhongshen Li, Kay Chen Tan, Qiuzhen Lin, Xiangtao Li, Ka-Chun Wong

IEEE Transactions on Evolutionary Computation · 2026-06 · 30 (3) · 1108-1122

DOI: 10.1109/tevc.2025.3586128 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Evolutionary Computation, 2026-06. Verfasst von Shen You und 6 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 50 Literaturangaben. Gefördert von: Hong Kong RGC, Research Grants Council of the Hong Kong Special Administrative Region, National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Despite the popularity of graph neural networks (GNNs), perturbed graph data is still a serious threat toward its inherent vulnerabilities. Adversarial examples can easily manipulate the output of GNNs across various attack scenarios. Meanwhile, studying attacks on graph networks is crucial, as it can help model designers enhance the robustness of their models. In this study, we propose a sensitivity-oriented poisoning attack for self-supervised graph embedding model through bilevel optimization, which employs different optimization methods at each level. To further enhance attack effectiveness, we analyze graph structure to identify sensitive nodes and edges that guide attack directions, combining gradient-based and query-based methods to target both edge connections and node attributes. Besides, according to the defects of existing graph masked autoencoders models, we design the feature sensitivity and feature variance to reduce the feature differentiability, which impairs the performance of the downstream model. Ablation studies validate the effectiveness of our operator on three citation datasets. And benchmark-based experiments support the effectiveness of our method on three different graph tasks. Specifically, our approach achieve an average reduction of 3% in the accuracy of node classification compared to existing methods for attacking graph structures alone. When attacking both graph structures and attributes, our model has even achieved an average reduction of 4.5% for the node classification task, outperforming the existing methods.

Förderung

  • Hong Kong RGC
  • Research Grants Council of the Hong Kong Special Administrative Region
  • National Natural Science Foundation of China
  • Innovation and Technology Commission
  • City University of Hong Kong

Zitieren

APA

You, S., Zhou, K., Li, Z., Tan, K. C., Lin, Q., Li, X., & Wong, K. (2026). SOPA: Sensitivity-Oriented Poisoning Attack for Self-Supervised Graph Embedding Model via Bilevel Evolutionary Optimization. IEEE Transactions on Evolutionary Computation, 30(3), 1108-1122. https://doi.org/10.1109/tevc.2025.3586128

Vancouver

You S, Zhou K, Li Z, Tan KC, Lin Q, Li X, et al. SOPA: Sensitivity-Oriented Poisoning Attack for Self-Supervised Graph Embedding Model via Bilevel Evolutionary Optimization. IEEE Transactions on Evolutionary Computation. 2026;30(3):1108-1122. doi:10.1109/tevc.2025.3586128

BibTeX
@article{You2026,
  author = {You, Shen and Zhou, Kai and Li, Zhongshen and Tan, Kay Chen and Lin, Qiuzhen and Li, Xiangtao and Wong, Ka-Chun},
  title = {SOPA: Sensitivity-Oriented Poisoning Attack for Self-Supervised Graph Embedding Model via Bilevel Evolutionary Optimization},
  journal = {IEEE Transactions on Evolutionary Computation},
  year = {2026},
  volume = {30},
  number = {3},
  pages = {1108--1122},
  doi = {10.1109/tevc.2025.3586128}
}
Kostenlos starten →