Network Security Project

This project is a machine learning–based Network Security application. It can train a model and predict whether network data is safe or malicious using a web API built with FastAPI.

Below is the complete workflow explained in simple language.

---

1. Project Startup

When the project starts:

• Environment variables are loaded using .env file (for MongoDB URL).
• A secure connection to MongoDB is created.
• A FastAPI web server is started.

The API is accessible through a browser.

---

2. MongoDB Connection

• The project connects to MongoDB using a URL stored in an environment variable.
• MongoDB is used to store raw and processed network data during training.
• Database and collection names are taken from constant files to keep things organized.

---

3. API Endpoints (Main Features)

(a) Home Route /

• Redirects the user to FastAPI documentation (/docs).
• Useful for testing APIs easily.

---

(b) Training Route /train

Purpose: Train the machine learning model.

What happens internally:

1. TrainingPipeline object is created.

2. The training pipeline runs step-by-step:

   • Data ingestion (fetching data)
   • Data validation
   • Data transformation
   • Model training
   • Model evaluation
   • Model saving

3. Final trained model and preprocessor are stored in the final_model/ folder.

Output:

• A success message once training is completed.

---

(c) Prediction Route /predict

Purpose: Predict results for new network data.

Input:

• A CSV file uploaded by the user.

Workflow:

1. Uploaded CSV file is read using Pandas.

2. Saved preprocessor is loaded from preprocessor.pkl.

3. Saved trained model is loaded from model.pkl.

4. Both are combined using NetworkModel.

5. Predictions are made on the uploaded data.

6. A new column (predicted_column) is added to the data.

7. Output is:

   • Saved as a CSV file (prediction_output/output.csv)
   • Displayed as an HTML table in the browser

Output:

• Prediction results shown in a table format.

---

4. Machine Learning Model Handling

• preprocessor.pkl → Handles data scaling/encoding
• model.pkl → Trained ML model
• NetworkModel → Combines preprocessing + prediction into one step

This makes prediction easy and clean.

---

5. Error Handling & Logging

• Custom exception class NetworkSecurityException is used.
• Errors are logged properly for debugging.
• If prediction fails, error details are returned clearly.

---

6. Frontend (HTML Output)

• Predictions are shown using Jinja2 templates.
• Data is displayed as a clean table in the browser.

---

7. Overall Flow Summary

User → Upload CSV / Trigger Training
     → FastAPI
     → ML Pipeline / Model
     → Prediction
     → CSV + Table Output

---

8. Technologies Used

• Python
• FastAPI
• MongoDB
• Machine Learning
• Pandas

---