@inproceedings{huang-etal-2026-toposhield,
    title = "{T}opo{SHIELD}: Reshaping the Flow of Malice via Spatio-Temporal Risk-Aware Topological Evolution in Multi-Agent Systems",
    author = "Huang, Ruiyang  and
      Wang, Chenxi  and
      Zhang, Tinghe  and
      Liu, Fengrui  and
      Sun, Jiayan  and
      Wang, Haocheng  and
      Wu, Yifan",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.findings-acl.426/",
    pages = "8740--8768",
    ISBN = "979-8-89176-395-1",
    abstract = "While LLM-based Multi-Agent Systems (MAS) demonstrate remarkable problem-solving capabilities, their interconnectivity acts as a conduit for the rapid spread of malicious injections. Addressing the limitations of static defenses, we present TopoSHIELD, a framework that reshapes the flow of malice via risk-aware topological evolution. Our approach utilizes a spatio-temporal graph neural network to monitor interaction dynamics, calculating node risk entropy (NRE) and edge attack conductivity (EAC) to pinpoint vulnerabilities. Guided by these metrics, TopoSHIELD executes precise structural interventions, pruning high-risk edges and isolating compromised communities to block attack diffusion. Empirically, TopoSHIELD reduces toxicity by 58{\%} on GPT-4o while preserving high utility ({\ensuremath{>}}90{\%} success rate), outperforming existing baselines in both suppression efficiency and scalability."
}