Cuckoo Watchtower

Project Overview

Cuckoo Watchtower (CW) is a tool designed to assist malware forensics investigations by leveraging the power of GPT-4, an advanced LLM by OpenAI. It processes large Cuckoo Sandbox JSON reports into accessible, chatbot-style insights via a Retrieval Augmented Generation (RAG) pipeline, all within a Google Colab notebook. CW aids analysts by dynamically responding to queries about malware behavior, pivoting investigation paths, and generating tailored reports.

Features

• Data Integration: Processes Cuckoo Sandbox JSON outputs into organized text files covering general, network, signature, and static analysis.
• Retrieval Augmented Generation (RAG): Enhances context understanding for the LLM to generate highly accurate and relevant responses.
• Vector Database: Employs Weaviate to store and retrieve analysis vectors efficiently.
• Chatbot Interaction: Provides dynamic, question-based analysis for streamlined investigations.

Use Cases

• Comprehensive Reports: Generates detailed, audience-specific malware analysis reports.
• Interactive Queries: Enables users to ask follow-up questions or clarify malware behaviors.
• Pivot Assistance: Facilitates rapid data retrieval for efficient investigative pivoting.

Limitations

• Data Size Constraints: Limited by the JSON file size and OpenAI's API token constraints.
• Privacy Concerns: May require alternative, private LLMs (e.g., LLaMA) for sensitive data processing in operational environments.

Future Work

• Integrating additional data sources, such as memory snapshots or BSON files.
• Adding the ability to run Cuckoo scans directly within the tool for streamlined workflows.

Technologies Used

• Cuckoo Sandbox: Malware analysis tool generating JSON reports.
• Python: Primary programming language for text processing and integration.
• Weaviate: Vector database for efficient context storage and retrieval.
• Google Colab: Notebook environment for tool deployment and user interaction.