Agentic Security: LLM Vulnerability Scanner and AI Red Teaming Kit

Summary

Agentic Security is an open-source vulnerability scanner designed for Agent Workflows and Large Language Models (LLMs). It helps protect AI systems from various threats, including jailbreaks, fuzzing, and multimodal attacks. This Python-based tool is essential for developers and security teams to proactively identify and mitigate risks in AI deployments, ensuring safer and more reliable AI systems.

Use at your own risk

OSRepos shares public repositories for knowledge and discovery only. Any installation, execution, configuration, or use of code from these repositories is the user's own responsibility. Always review the repository, source code, dependencies, licenses, and security implications before running or installing anything. OSRepos is not responsible for issues, damages, or losses resulting from third-party repositories.

Introduction

Agentic Security is a powerful open-source vulnerability scanner for Agent Workflows and Large Language Models (LLMs). Developed by msoedov, this Python-based tool provides a comprehensive kit for AI red teaming, enabling users to identify and mitigate risks such as jailbreaks, fuzzing, and multimodal attacks. It offers robust features to ensure the safety and reliability of AI systems, making it an invaluable asset for developers, researchers, and security teams.

Key features include:

Multimodal Attacks: Probe vulnerabilities across text, images, and audio inputs.
Multi-Step Jailbreaks: Simulate sophisticated, iterative attack sequences.
Comprehensive Fuzzing: Stress-test any LLM with randomized inputs.
API Integration & Stress Testing: Seamlessly connect to LLM APIs and push their limits.
RL-Based Attacks: Leverage reinforcement learning to craft adaptive, intelligent probes.

Installation

To get started with Agentic Security, simply install the package using pip:

pip install agentic_security

Examples

After installation, you can quickly launch the Agentic Security server:

agentic_security

This will start a local server, often with a UI accessible via a web browser for interactive testing. For more advanced usage, you can initialize a configuration file and run a CI check to integrate security scanning into your development pipeline:

agentic_security init
agentic_security ci

Agentic Security also supports probing various modalities like image and audio, and integrating with LLM APIs using custom HTTP specifications, allowing for flexible and comprehensive testing.

Why Use Agentic Security?

Agentic Security is crucial for anyone involved in developing, deploying, or securing AI systems. It provides the tools to proactively identify and mitigate vulnerabilities in LLMs and agent workflows, protecting against emerging threats like sophisticated jailbreaks and multimodal attacks. By integrating this scanner into your development lifecycle, you can ensure safer, more robust, and reliable AI deployments, safeguarding against potential misuse and enhancing trust in your AI applications. Its focus on real-world threats and adaptive attack strategies makes it a vital component for modern AI security.