Spotlighting

Type: mitigation

Description: A defense mechanism that uses prompt engineering techniques to avoid indirect or direct prompt injection by highlighting the query

Version: 0.1.0

Created At: 2025-03-04 10:27:40 -0500

Last Modified At: 2025-03-04 10:27:40 -0500

External References

Defending Against Indirect Prompt Injection Attacks With Spotlighting, arXiv

--> ChatGPT (platform): Evaluation of the above mitigation strategies leveraged GPT 3.5 and GPT 4.
<-- LLM Prompt Injection (technique): By spotlighting in prompts, the LLM focuses on a specific part of the query that defines the task, thus avoiding other injected tasks.