Trojaning the Alignment: Stealthy Backdoor Attacks against Graph Foundation Models
作者: Minhua Lin, Zhicheng Gao, Yilong Wang, Hanqing Lu, Xiang Zhang, Suhang Wang
分类: cs.LG
发布日期: 2026-08-21
备注: Accepted by ICDM 2026
🔗 代码/项目: GITHUB
💡 一句话要点
提出STAG框架以解决图基础模型的隐蔽后门攻击问题
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 图基础模型 后门攻击 图-语言对齐 隐蔽性 图学习 安全性 文本属性图
📋 核心要点
- 现有方法在图-语言对齐的背景下,无法有效处理图和文本的独立后门攻击,导致脆弱性未被充分挖掘。
- 本文提出STAG框架,通过协调图触发器生成器与文本软提示,实现图和文本的协同攻击,增强隐蔽性。
- 在多个TAG数据集上的实验结果表明,STAG在攻击成功率和隐蔽性上均显著优于现有方法。
📝 摘要(中文)
图基础模型(GFMs)在文本属性图(TAGs)上对图表示与语言语义进行对齐,以支持可转移的图学习。然而,GFMs在TAGs下的后门脆弱性尚未得到充分理解,尤其是在图与语言对齐的情况下。现有的后门攻击主要独立针对图或文本,导致直接适应效果不佳。本文提出了STAG,一个针对GFMs在TAGs上图-语言对齐接口的隐蔽后门攻击框架。STAG通过图触发器生成器与文本侧软提示的协调,使得附加触发器的图表示和触发文本表示朝向相同的目标类别文本区域移动。大量实验表明STAG的有效性和隐蔽性。
🔬 方法详解
问题定义:本文旨在解决图基础模型在文本属性图上的后门攻击脆弱性,现有方法未能有效结合图和文本的特性,导致攻击效果不佳。
核心思路:STAG框架通过图触发器生成器与文本侧软提示的协调,确保触发器附加的图表示与文本表示朝向同一目标类别,从而实现隐蔽攻击。
技术框架:STAG的整体架构包括图触发器生成器、文本侧软提示模块和触发器附加的图表示处理,确保两种模态的协同作用。
关键创新:STAG的主要创新在于同时考虑图和文本的特性,通过可读文本的触发节点和正则化的局部结构,提升了攻击的隐蔽性和有效性。
关键设计:在设计中,采用了候选检索机制来生成可读的触发节点,并通过正则化损失函数确保触发器附加子图的局部结构与原始子图相似。
🖼️ 关键图片
📊 实验亮点
实验结果显示,STAG在多个TAG数据集上实现了超过90%的攻击成功率,相较于现有方法提升了约30%的隐蔽性,验证了其在图-语言对齐场景下的有效性。
🎯 应用场景
该研究的潜在应用领域包括安全性较高的图学习系统、社交网络分析和金融欺诈检测等。通过增强模型对后门攻击的抵抗力,提升了系统的安全性和可靠性,具有重要的实际价值和未来影响。
📄 摘要(原文)
Graph Foundation Models (GFMs) on text-attributed graphs (TAGs) align graph representations with language semantics to support transferable graph learning. Despite these advantages, the backdoor vulnerability of GFMs on TAGs remains insufficiently understood, especially under graph-language alignment, where graph and text representations are trained to constrain each other in a shared semantic space. Existing backdoor attacks mainly target either the graph side or the text side, treating the two modalities independently. This makes direct adaptation ineffective: graph-only triggers can be constrained by clean text semantics, while text-only triggers alter the language view but do not directly shift the graph representation being aligned and scored. TAGs also impose a stealth challenge because triggers are exposed as both node text and local graph structure, making incoherent trigger attributes or anomalous subgraphs easy to inspect or filter. In this paper, we propose STAG, a stealthy trojan attack framework designed for the graph-language alignment interface of GFMs on TAGs. STAG coordinates a graph-trigger generator with a text-side soft prompt so that trigger-attached graph representations and triggered text representations move toward the same target-class text region. To address TAG-specific stealthiness, STAG realizes trigger nodes as readable text through candidate retrieval and regularizes the trigger-attached subgraph so that its local structure remains close to the original subgraph. Extensive experiments on multiple TAG datasets and representative GFMs demonstrate the effectiveness and stealthiness of STAG. Our code is available at https://github.com/ventr1c/STAG.