Network Intrusion Detection (NSL-KDD)
ML pipeline for network anomaly detection: EDA on ~125k records, Random Forest with Bayesian tuning, and an interactive Streamlit dashboard.
Overview
Built as an R&D member of the Securinets FST cybersecurity club: a complete ML pipeline for intrusion detection on the NSL-KDD dataset (~125,000 records, 42 features).
The work covers data cleaning and exploratory analysis with correlation-guided feature selection, then a Random Forest classifier tuned with Bayesian hyperparameter optimization (skopt), evaluated with per-class precision, recall, and F1 via confusion matrices.
Results are delivered through a 3-page interactive Streamlit dashboard — Overview, Feature Importance, and Model Comparison — designed so non-technical stakeholders can explore model behavior.