Unsupervised Heterogeneous Graph Rewriting Attack via Node Clustering
Open Access
- 24 August 2024
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 3057-3068
Abstract
Self-supervised learning (SSL) has become one of the most popular learning paradigms and has achieved remarkable success in the graph field. Recently, a series of pre-training studies on heterogeneous graphs (HGs) using SSL have been proposed considering the heterogeneity of real-world graph data. However, verification of the robustness of heterogeneous graph pre-training is still a research gap. Most existing researches focus on supervised attacks on graphs, which are limited to a specific scenario and will not work when labels are not available. In this paper, we propose a novel unsupervised heterogeneous graph rewriting attack via node clustering (HGAC) that can effectively attack HG pre-training models without using labels. Specifically, a heterogeneous edge rewriting strategy is designed to ensure the rationality and concealment of the attacks. Then, a tailored heterogeneous graph contrastive learning (HGCL) is used as a surrogate model. Moreover, we leverage node clustering results of the clean HGs as the pseudo-labels to guide the optimization of structural attacks. Extensive experiments exhibit powerful attack performances of our HGAC on various downstream tasks (i.e., node classification, node clustering, metapath prediction, and visualization) under poisoning attack and evasion attack.Keywords
Funding Information
- National Key Research and Development Program of China (2022YFB4500300)
- National Natural Science Foundation of China (62271452)
This publication has 35 references indexed in Scilit:
- PAGCL: An unsupervised graph poisoned attack for graph contrastive learning modelFuture Generation Computer Systems, 2023
- HeGCL: Advance Self-Supervised Learning in Heterogeneous Graph-Level RepresentationIEEE Transactions on Neural Networks and Learning Systems, 2023
- Heterogeneous Graph Contrastive Learning with Meta-path Contexts and Weighted Negative Samples2023
- HeteroGuard: Defending Heterogeneous Graph Neural Networks against Adversarial Attacks2022
- Adversarial Attack and Defense on Graph Data: A SurveyIEEE Transactions on Knowledge and Data Engineering, 2022
- Attention-aware metapath-based network embedding for HIN based recommendationExpert Systems with Applications, 2021
- Leveraging Meta-path Contexts for Classification in Heterogeneous Information Networks2021
- Metapath-guided Heterogeneous Graph Neural Network for Intent Recommendation2019
- Adversarial Attacks on Neural Networks for Graph Data2018
- Non-parametric scan statistics for event detection and forecasting in heterogeneous social media graphs2014