BASIS: Breach-Aware Selective Prompt Injection Shielding with Prefill Attention Probes

Abstract

Prompt injection is a critical security threat in large language model (LLM) applications, where attackers hijack model behavior by embedding malicious instructions in user or external data. Existing detection methods only detect the presence of injection and refuse to respond upon detection, overlooking the fact that for many modern aligned models, well-crafted instructions can resist most injection attacks. This means that the injection robustness varies significantly across instructions and m

Categories

Framework mappings

Suggested from the entry's categories.

@misc{qin2026basis,
  title = {{BASIS: Breach-Aware Selective Prompt Injection Shielding with Prefill Attention Probes}},
  author = {Laiqiao Qin and Tianqing Zhu and Longxiang Gao and Wanlei Zhou},
  year = {2026},
  month = aug,
  eprint = {2608.08027},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2608.08027}
}