HoneyShield AI
Explainable behavioral anomaly detection system combining multiple modelling approaches with a usable dashboard.
Inspect the architecture signal.
Model map uses only the supplied HoneyShield stack and delivery facts.
Streamlit
Dashboard layer delivered with the hackathon submission.
Problem
Build an explainable behavioral anomaly detection system for the Honeywell Campus Connect hackathon.
Constraints
Solo submission.
Hackathon setting required a complete technical submission, not just a model.
Decisions / trade-offs
Combine Isolation Forest, a GRU autoencoder, and XGBoost rather than presenting a single-model pipeline.
Use SHAP for explainability and Streamlit for the dashboard layer.
What broke / changed
No specific failed experiment was supplied in the project history.
Verified project facts
Honeywell Campus Connect hackathon finalist.
Produced README, system design document, technical report, charts, and architecture diagram.