Formalizing and Benchmarking Prompt Injection Attacks and Defenses

Abstract

Proposes defense mechanisms against prompt injection in LLM systems including isolation-based approaches, input/output filtering, and detection methods.

Categories

Framework mappings

@inproceedings{liu2024formalizing,
  title = {{Formalizing and Benchmarking Prompt Injection Attacks and Defenses}},
  author = {Yupei Liu and Yuqi Jia and Runpeng Geng and Jinyuan Jia and Neil Zhenqiang Gong},
  year = {2024},
  booktitle = {USENIX Security 2024},
  eprint = {2310.12815},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2310.12815}
}