This project is an advanced cybersecurity defense mechanism designed to deceive attackers and extend dwell time using Reinforcement Learning (RL). Unlike static honeypots, this system dynamically adapts its responses based on the attacker's behavior, classifying threats in real-time and routing them to high-interaction containers.
- Core Engine: Python (Flask)
- Containerization: Docker
- ML/RL: XGBoost (Classification), Q-Learning (Response Strategy)
- Frontend Dashboard: React.js (Threat Visualization)
- Dynamic Adaptation: Changes system output/errors to confuse attackers.
- Real-Time Classification: Detects SQLi, XSS, and Brute Force using XGBoost.
- Smart Routing: Redirects sophisticated actors to "Tarpit" environments to waste their time.
The full source code is currently being refactored for public release following the completion of the ICACCS '26 review process. Core modules will be uploaded progressively.
This project was engineered by a team of 4 researchers as part of our final year Capstone.
- Kartik Nair: Lead on Deception Module & Dashboard Integration.
- Md Sahis NP: Core Engine & Backend Logic.
- Adhityan S: Dockerization & Deployment.
- Janisha K: Data Analysis & Reporting. (Research Paper submitted to ICACCS '26)