README.md

metadata

title: Network Intrusion Detection System
colorFrom: blue
colorTo: red
sdk: gradio
sdk_version: 4.0.0
app_file: app.py
pinned: false
license: mit
tags:
  - ml-intern

🛡️ Network Intrusion Detection System (NIDS)

Real-time network intrusion detection using ML & Deep Learning models trained on the NSL-KDD dataset.

Models

• XGBoost (best) — Accuracy 76.18%, AUC-ROC 95.75%
• Random Forest — Accuracy 73.10%, AUC-ROC 95.34%
• MLP — Deep learning baseline
• Autoencoder — Unsupervised anomaly detection
• LSTM — Sequential feature modeling
• Transformer — FlowTransformer-style architecture

Dataset

NSL-KDD — 185K records, 41 features, binary classification (normal/anomaly)

Usage

REST API (FastAPI)

pip install fastapi uvicorn
uvicorn api:app --host 0.0.0.0 --port 8000

Endpoints:

• GET /health — Health check
• GET /models — List models
• POST /predict — Single flow detection
• POST /predict/batch — Batch detection
• GET /stats — Usage statistics

Gradio UI

python app.py

API Example

curl -X POST http://localhost:8000/predict \
  -H "Content-Type: application/json" \
  -d '{"features":[0,1,45,0,491,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,0,0,0,0,1,0,0,0,2,2,0,0,0,0,1,0,0,0,0.17], "model":"XGBoost"}'

Results

Model	Accuracy	Macro F1	AUC-ROC	Type
XGBoost	76.18%	76.04%	95.75%	Supervised
RandomForest	73.10%	73.05%	95.34%	Supervised
MLP	73.28%	73.21%	89.33%	Supervised
Autoencoder	71.84%	71.34%	73.60%	Unsupervised
LSTM	70.65%	70.58%	87.80%	Unsupervised
Transformer	57.94%	57.26%	80.29%	Supervised
IsolationForest	56.55%	55.96%	65.24%	Unsupervised

Project Structure

nids_project/
├── api.py           # FastAPI REST API
├── app.py           # Gradio web UI (this Space)
├── data_loader.py   # Dataset loading
├── ml_models.py     # ML implementations
├── dl_models.py     # DL implementations
├── realtime.py      # Real-time pipeline
├── realtime_batch.py # Batch throughput benchmark
├── Dockerfile       # Container image
└── requirements.txt # Dependencies