The proposed ZeroGAR is the first systematic benchmark for evaluating the adversarial robustness of ZGMs and reveals that strong clean zero-shot performance does not guarantee adversarial robustness, with three key findings.
Abstract
Zero-shot graph models (ZGMs), which learn transferable knowledge from source graphs and directly apply to unseen target graphs without any adaptation, have achieved promising performance and attracted considerable attention. Despite their proliferation, existing ZGMs are predominantly evaluated on clean graphs, while existing graph robustness benchmarks mainly focus on supervised settings, leaving a fundamental question largely unexplored: How robust are ZGMs when their unseen target graphs are exposed to adversarial manipulation? In this paper, we answer this question by proposing ZeroGAR, the first systematic benchmark for evaluating the adversarial robustness of ZGMs. ZeroGAR evaluates 13 representative ZGMs from 3 different paradigms on 8 graph datasets across 4 domains, covering both in-domain and cross-domain transfer under structural, textual, and node injection attacks with multiple perturbation budgets. It further investigates whether existing graph defenses remain effective in the zero-shot setting. Extensive experiments reveal that strong clean zero-shot performance does not guarantee adversarial robustness, with three key findings: (1) Vulnerability patterns are related to model prediction mechanisms: GNN-based methods are particularly vulnerable to structural and node injection attacks, whereas LLM-based methods are more vulnerable to textual attacks; (2) Stronger LLM backbones introduce a structure-text robustness trade-off; (3) Existing graph defense methods do not consistently improve zero-shot robustness and may compromise clean performance. We hope that ZeroGAR will facilitate rapid, equitable evaluation and inspire further innovative research in ZGM security.
This work designs a dual-consistency graph augmentation method that enriches the graph topology to mitigate structural vulnerabilities and presents an adversarial masked autoen-coder to mitigate model overfitting to the perturbed graph.
Qi-Qi Zhang, Chuan-Jin Liu, Gen Liu et al.· Proceedings of the Thirty-Fi...· 0 citations
Strong standard-metric performance alone is insufficient to establish robustness after unlearning and motivate adversarial stress-testing as a complementary component of unlearning evaluation, showing a substantial gap between clean-query forgetting and adversarial robustness.
A. Gupta, Hima Varshini Surisetty, Sreevidya Bollineni et al.· 1 citation
Kernel-Complexity Edge Sanitization (KCES), a training-free and model-agnostic framework for defending against structural attacks, provides a principled and efficient framework for securing GNNs.
Ya-Ning Jia, Shenyang Deng, Yaoqing Yang et al.· ACM International Conference...· 0 citations
This work proposes a transferable graph purification scheme, named ProGAP, to bridge adversarial defense knowledge via vulnerability-aware graph prompt learning, and achieves 1%-9% improvement, and reduces the time consumption by up to 2.2x.
Shuo-Min Xue, Jing-Yuan Li, Ju Jia et al.· ACM International Conference...· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduOct 13, 2026
Machine-learning networks can learn without detailed instructions from a designer. Those that do may offer insight into how the brain reorganizes itself during learning.
MIT News · Artificial Intelligence· news.mit.eduSep 29, 2026
Professor Sherry Turkle’s new book, “Artificial Intimacy,” offers a withering critique of chatbots and the antisocial dynamics she believes they encourage.