arXiv：cs.LG· Honglin Gao, Junhao Ren, Lan Zhao, Yue Yang, Peilin Du, Jindong Chang, Gaoxi Xiao·· 3 小时前
Blackknife：面向异质图神经网络的免查询黑盒重连攻击
Blackknife: Query-Free Black-Box Rewiring Attacks on Heterogeneous Graph Neural Networks
AI 导读
Blackknife 是一种免查询、严格黑盒的逃逸攻击，仅利用被攻击节点的一跳结构、真实标签和原始特征，无需受害模型输出即可训练代理模型并生成扰动。它通过连续松弛与投影梯度上升优化结构扰动，每次重连删除一条边并添加同关系类型的另一条边，保持目标节点总度数与各关系类型度数不变。
正文
Abstract:Heterogeneous graph neural networks (HGNNs) perform well on graphs with multiple node and relation types, but their robustness to attacks with restricted structural access remains insufficiently understood. We propose Blackknife, a query-free, strict black-box evasion attack that observes only the one-hop structures of the attacked nodes, their ground-truth labels, and raw features within the permitted action space. No victim-model outputs are used to train the surrogate, generate perturbations, or select this http URL trains a shared surrogate using only observable edges and relation types, then optimizes structural perturbations through continuous relaxation and projected gradient ascent. Each rewiring operation deletes one edge and adds another of the same relation type, preserving the target node's total degree and degree for each relation type. A cosine-similarity penalty and an adaptive structural classifier further regularize the perturbations through bilevel optimization. Experiments on different datasets demonstrate competitive attack success rates across five victim architectures and three defense settings. The results expose HGNN vulnerabilities under jointly restricted graph access and query-free attack generation.
| Subjects: | Machine Learning (cs.LG); Social and Information Networks (cs.SI) |
| Cite as: | arXiv:2606.29240 [cs.LG] |
| (or arXiv:2606.29240v3 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2606.29240 arXiv-issued DOI via DataCite |
Submission history
From: Honglin Gao [view email]
[v1]
Sun, 28 Jun 2026 07:11:05 UTC (737 KB)
[v2]
Tue, 13 Oct 2026 07:52:15 UTC (1,883 KB)
[v3]
Wed, 14 Oct 2026 05:34:14 UTC (1,862 KB)
来源：arXiv：cs.LG · arxiv.org