buttercup - Buttercup finds and patches software vulnerabiliti

项目说明

Buttercup Cyber Reasoning System (CRS)

![Tests]()
![Component Integration Tests]()
![System Integration]()

Buttercup is a Cyber Reasoning System (CRS) de by Trail of Bits for the DARPA AIxCC (AI Cyber Challenge). Buttercup finds and patches software in open-source code repositories like example-libpng. It starts by running an AI/ML-assisted fuzzing campaign (built on oss-fuzz) for the program. When are found, Buttercup analyzes them and uses a multi-agent AI-dri to repair the . Buttercup system consists of several components:

System Requirements

Minimum Requirements

Note: Buttercup uses third-party AI pro (LLMs from companies like OpenAI, Anthropic and Google), which cost money. Please ensure that you manage per-deployment costs by using the built-in LLM budget setting.

Note: Buttercup works best with access to models from OpenAI and Anthropic, but can be run with at least one API key from one third-party provider. You can use a combination of OpenAI, Anthropic, and Google LLMs.

Supported Systems

Required System Packages

Before setup, ensure you ha packages installed:

Supported Targets

Buttercup works with:

Quick Start

1. Clone the repository with submodules:

1. Run automated setup (Recommended)

This script will install all dependencies, configure the en, and guide you through the setup process.

Note: If you prefer manual setup, see the Manual Setup Guide.

1. Ensure Docker is running before deploying:

1. Start Buttercup locally

1. local deployment:

When a deployment is successful, you should see all pods in "Running" or "Completed" status.

1. Send Buttercup a simple task

Note: When tasked, Buttercup will start consuming third-party AI resources.

This command will make Buttercup pull down an example repo example-libpng with a known . Buttercup will start fuzzing it to find and patch .

1. Access Buttercup's web-based GUI

Run:

Then na to in your web browser.

In the GUI you can monitor acti and see when Buttercup finds bugs and generates patches for them.

1. Stop Buttercup

Note: This is an important step to ensure Buttercup shuts down and stops consuming third-party AI resources.

Accessing Logs

Buttercup includes local SigNoz deployment by default for comprehensi obser. You can access logs, traces, and metrics through the SigNoz UI:

Then na to in your web browser to view:

If you configured LangFuse during setup, you can also monitor LLM usage and costs there.

For additional log access methods, see the Quick Reference Guide.

Additional Resources

项目信息

开发语言：Python

Star 数：1716

技术要点

该项目属于漏洞知识库方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕Buttercup finds and patches software vulnerabilities展开，提供了可运行的实现，对理解相关安全机制、工具原理与攻防技术具有较高的研究价值。

从技术实现角度看，此类项目的核心在于对底层原理的深入理解与工程化落地。开发者在研究过程中可以结合官方文档与社区讨论，梳理项目的初始化流程、核心模块职责与数据流转，逐步掌握其中的技术细节与设计思路。

使用与学习建议

1. 下载源码压缩包，解压到本地环境。

2. 根据项目类型准备对应的运行环境与依赖。

3. 按照 README 中的步骤完成配置与部署。

4. 在隔离的测试环境中运行验证，避免影响生产系统。

免责声明

本资源来源于开源社区，仅供安全研究、技术学习与合法授权范围内的攻防演练使用，请勿用于任何非法或侵害他人权益的用途。使用本项目产生的一切后果由使用者自行承担。