Inspiration
We are inspired by the need to create a homegrown, AI-powered cybersecurity solution that defends and actively learns from attackers. With rising zero-day exploits and sophisticated intrusions, we wanted to build something adaptive, scalable, and uniquely engineered for next-gen defense.
What it does
Indiengine is an AI-driven dynamic honeypot system that creates ultra-realistic decoy environments, engages attackers in real time, and extracts intelligence on novel threats. It uses LLMs, anomaly detection, and reinforcement learning to continuously adapt, detect zero-days, and generate actionable security insights.
How we built it
We designed a modular architecture with three layers: Deception Layer with LLMs and generative AI. Sensing & Analysis Layer using Isolation Forest, Autoencoders, Random Forest, and XGBoost. Adaptation Layer with reinforcement learning for dynamic responses. We trained models on synthetic + real-world datasets, integrated them into a simulated enterprise network, and validated results against penetration testing scenarios.
Challenges we ran into
Preventing attackers from fingerprinting the honeypot.
Managing adversarial ML risks of data poisoning, evasion.
Limited high-quality labeled datasets for zero-day detection.
Accomplishments that we're proud of
Successfully built a self-adaptive honeypot prototype. Developed a system that can autonomously patch vulnerabilities in simulated environments.
What we learned
Reinforcement and Federated learning
What's next for Indiengine
We are currently working on AI based pentesting proof tool suite
Log in or sign up for Devpost to join the conversation.