基于强化学习的图神经网络节点注入标签特异性攻击（英文版）

英文标题：Single Node Injection Label Specificity Attack on Graph Neural Networks via Reinforcement Learning中文摘要：提出了一种基于强化学习的无梯度通用对抗攻击方法，称为 G$^2$-SNIA，它可以开展黑盒回避攻击并仅注入单个恶意节点至目标节点来操纵其分类结果，优于现有最先进的基线。英文摘要：Graph neural networks (GNNs) have achieved remarkable success in variousreal-world applications. However, recent studies highlight the vulnerability ofGNNs to malicious perturbations. Previous adversaries primarily focus on graphmodifications or node injections to existing graphs, yielding promising r