Rebuff

Rebuff 0.1.1 is an open-source AI security framework designed to help protect Large Language Model (LLM) applications against prompt injection attacks. It provides developers with tools for identifying potentially malicious prompts and adding an additional security layer to AI-powered applications.

Rebuff can be used when developing LLM applications that process user-provided or untrusted input. It is intended to help developers detect suspicious prompt-injection attempts and improve the security posture of AI workflows.

Features:

• Prompt-injection detection for LLM applications
• Multiple security checks for identifying suspicious inputs
• Python-based SDK for integration into AI applications
• Designed for protecting applications that process untrusted prompts
• Suitable for experimentation, development, and AI security research

Usage:

# Activate virtual environment: 
$ sudo su
$ cd /usr/src
$ cd Python-3.12.12
$ source /opt/rebuff/venv/bin/activate

#Verify the RedBuff version: $ python -m pip show rebuff

Disclaimer: Rebuff is an open-source security project and should not be considered a complete or guaranteed defense against prompt injection or other AI security threats. Detection results may be incomplete or inaccurate. The project is archived upstream and is no longer actively maintained. Users should evaluate Rebuff within their own security architecture and implement additional application-level security controls, validation, monitoring, and access restrictions as appropriate.