All projects14 / 43
AI & Data
IoT Intrusion Detection
Vulnerability-aware network security with ML
At a glance
- Discipline
- AI & Data
- Key features
- 5 delivered
- Screens
- 4 in gallery
PythonScikit-learnXGBoostPandasSHAPParquet
IoT Intrusion Detection
Overview
Tree-based models trained on 6.6 million network flows, with XGBoost reaching 98.22% accuracy and 99.07% ROC-AUC.
The pipeline aggregates CSE-CIC-IDS2018 Parquet data, removes leakage-prone identifiers, and standardises 78 encoded features before a stratified 80/20 split.
Beyond classification, the most influential flow features are mapped to real attack patterns such as brute force, command-and-control and denial of service.
Key features
- 016.6M network flows processed end to end
- 02Random Forest, XGBoost and soft-voting ensemble
- 0399.69% precision with far fewer false positives
- 04Leakage-aware preprocessing
- 05Explainability layer linking features to vulnerabilities
Gallery
Inside the product
4 screens
Up next15 / 43
Race Performance Predictor
Explainable ML for top-3 finishes
