Malware Generation using LLM
Created a controlled environment to study AI model behavior and safety bypass mechanisms.
Project Overview
This project involved creating a controlled environment to investigate the behavior of AI models, specifically focusing on safety bypass mechanisms. The goal was to understand how generative models could potentially be misused to generate malware and how to defend against such threats.
Technologies Used: FastAPI, Hugging Face, Docker, Django.
Key Learnings: Model fine-tuning, prompt engineering, AI safety and ethics.
Research Dashboard
The research interface allows for fine-grained control over model parameters and provides visualizations for identified security bypass attempts.