@inproceedings{chang-etal-2026-know,
    title = "Know Thy Enemy: Securing {LLM}s Against Prompt Injection via Diverse Data Synthesis and Instruction-Level Chain-of-Thought Learning",
    author = "Chang, Zhiyuan  and
      Li, Mingyang  and
      Huang, Yuekai  and
      Jiang, Ziyou  and
      Jia, Xiaojun  and
      Xiong, Qian  and
      Wang, Junjie  and
      Li, Zhaoyang  and
      Wang, Qing",
    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.173/",
    pages = "3540--3561",
    ISBN = "979-8-89176-395-1",
    abstract = "Large language model (LLM)-integrated applications have become increasingly prevalent, yet face critical security vulnerabilities from prompt injection (PI) attacks. Defending against PI attacks faces two major issues: malicious instructions can be injected through diverse vectors, and injected instructions often lack clear semantic boundaries from the surrounding context, making them difficult to identify. To address these issues, we propose InstruCoT, a model enhancement method for PI defense that synthesizes diverse training data and employs instruction-level chain-of-thought fine-tuning, enabling LLMs to effectively identify and reject malicious instructions regardless of their source or position in the context. We evaluate InstruCoT across three critical dimensions: Behavior Deviation, Privacy Leakage, and Harmful Output. Experimental results across four LLMs demonstrate that InstruCoT significantly outperforms baselines in all dimensions while maintaining utility performance without degradation."
}