InjecAgent: Benchmarking Indirect Prompt Injections in Tool-Integrated LLM Agents

Qiusi Zhan, Zhixiang Liang, Zifan Ying, Daniel Kang

Abstract

Presents InjecAgent, a benchmark for evaluating indirect prompt injection attacks against LLM agents that use tools, showing most agents are highly vulnerable.

Categories

Tags

Framework Mappings

Cite This Resource

@article{llmsec202400013,
  title = {InjecAgent: Benchmarking Indirect Prompt Injections in Tool-Integrated LLM Agents},
  author = {Qiusi Zhan and Zhixiang Liang and Zifan Ying and Daniel Kang},
  year = {2024},
  journal = {ACL 2024 Findings},
  url = {https://arxiv.org/abs/2403.02691},
}