This research introduces Encrypted Prompt, a novel security mechanism that embeds permission data within user prompts to prevent unauthorized actions in LLM applications.

• Appends encrypted permission information to each user prompt
• Verifies permissions before executing any LLM-generated actions (like API calls)
• Provides stronger security guarantees than detection-only approaches
• Creates a verifiable security barrier between user input and system actions

This innovation matters because it shifts from merely detecting attacks to actively preventing unauthorized actions, creating more robust and trustworthy LLM-powered applications for enterprise use.

Encrypted Prompt: Securing LLM Applications Against Unauthorized Actions