Overview
DASE was my Master's project for my M.S. in Cybersecurity. The core focus was to create **dynamically generated scenarios** for incident response training. Unlike traditional static simulations, DASE adapts to the specific technology stack of a company, providing highly relevant and realistic training environments.
How It Works
The engine utilizes Company Profiles as input for a Large Language Model (LLM). The LLM interacts with the end-user to understand their training needs and then generates a bespoke attack scenario.
Because the LLM is grounded in the specific company profile, it significantly reduces the risk of hallucination and ensures the generated content is technically accurate for that environment. Dynamic generation at runtime ensures that the scenarios are fresh and tailored, rather than using generic, pre-canned simulations.
Key Features
- Dynamic Scenario Generation: Creates unique attack vectors based on real-time inputs.
- Context-Aware: understands specific tech stacks to generate relevant threats.
- Interactive Feedback: The LLM provides detailed feedback to users after the simulation concludes.
- Customizable Review: Users can select difficulty levels and the number of adversary actions.
Methodology
As part of this project, I conducted extensive user testing surveys to refine the user experience. A thorough literature review was also performed to validate the methodology against current research in adversary simulation and automated training.