Prompt Injection Attacks on Language Models¶
Authors: Huang et al. Year: 2023 ArXiv/Link: https://arxiv.org/abs/2301.10226
Summary¶
Taxonomy and analysis of prompt injection attacks where user input can override system prompts.
Key Concepts¶
- Prompt injection
- Attack taxonomy
- Input validation
- Model robustness
- Security vulnerabilities
Impact¶
Identified critical vulnerability class in LLM systems
Category¶
Jailbreaks & Attacks