Cyber City is a pioneering research project at the intersection of Artificial Intelligence and Offensive Security. It leverages the power of the Google Gemini 1.5 Pro model to create a specialized knowledge base for penetration testing, trained on a unique dataset of security walkthroughs and real-world attack patterns.
The goal of Cyber City is to build an autonomous security assistant capable of:
- Pattern Recognition: Analyzing video walkthroughs (TryHackMe, CTF recordings) to understand a hacker's "train of thought."
- Methodology Synthesis: Distilling complex attack chains into actionable, structured security data.
- Predictive Modeling: Suggesting the next logical step in a penetration test based on observed system responses.
- Gemini 1.5 Pro Integration: Utilizes long-context windows to "digest" entire walkthrough videos and identify security patterns.
- Video-to-Data Pipeline: (In Development) A custom framework for extracting and labeling security actions from video frames.
- Knowledge Base Augmentation: Continuously learning from the latest TryHackMe rooms and bug bounty write-ups.
- Safety Centric: Configured with strict AI safety settings to ensure ethical and authorized use.
- Python 3.10+
- Google Generative AI Python SDK:
pip install google-generativeai
- A valid Google Gemini API Key.
git clone https://github.com/Exploit0xfffff/cyber-city.git
cd cyber-city
python main.py- Initial Gemini API Integration and Chat Core.
- [/] Development of the Video Frame Extraction Engine.
- Manual Annotation of the first 50 TryHackMe Walkthroughs.
- Training the custom "Pentest Intelligence" model.
- Integration with real-time CLI tools for live assistance.
"Training the AI of tomorrow to defend the networks of today."