Cloud-Based Vehicle Telematics Anomaly Detection

A real-time telematics platform for simulating CAN-like vehicle data, streaming it securely, and detecting anomalies using machine learning. Built for hands-on experience with modern telematics, API security, and cloud deployment.

Features

• CAN-like Data Simulation: Generates realistic vehicle telemetry (RPM, temperature, voltage, speed, etc.) with occasional injected faults.
• Real-Time Streaming: Streams live data to clients using FastAPI WebSocket endpoints.
• JWT Authentication: Secures API access and data transmission with industry-standard JWT tokens.
• Anomaly Detection: Uses an Isolation Forest model to flag abnormal vehicle behavior in real time.
• Cloud Deployment: Backend and ML inference deployed on AWS EC2 for scalable, remote access.
• Dashboard Integration: (Planned) Live dashboard for visualization and monitoring.

Installation

1. Clone the repository:

   git clone https://github.com/Architrawat25/telematics-anomaly-detection.git
   cd telematics-anomaly-detection
   

2. Set up a virtual environment:

   python3 -m venv venv
   source venv/bin/activate
   

3. Install dependencies:

   pip install -r requirements.txt
   

4. Train the anomaly detection model (optional, if you want to retrain):

   • Open ml/model_training.ipynb in Google Colab or Jupyter.
   • Run all cells to generate and save anomaly_detector.joblib in the ml/ folder.

Running Locally

1. Start the FastAPI backend:

   python -m uvicorn backend.main:app --reload
   

2. Test the WebSocket stream:

   • Use the provided test_ws.py script or any WebSocket client.
   • Obtain a JWT token by sending a POST request to /token.
   • Connect to ws://localhost:8000/ws/vehicle-data and send the token as the first message.

Deployment

• The backend can be deployed on AWS EC2 (Free Tier).
• After deployment, update security groups to allow HTTP/WebSocket traffic.

Project Structure

├── backend/ # FastAPI app and JWT logic ├── simulator/ # CAN-like data generator ├── ml/ # Model training notebook and saved model ├── requirements.txt └── test_ws.py # WebSocket test client

Tech Stack

Python, FastAPI, WebSocket, JWT, Scikit-learn, Pandas, AWS

• Real-time CAN-like data simulation
• Secure streaming with JWT authentication
• ML-based anomaly detection
• AWS cloud deployment
• Dashboard integration (coming soon)

Author

Archit Rawat