XAI-EncryptedTrafficClassifier

This repository contains an Explainable AI (XAI) model developed for the classification of encrypted network traffic using a combination of machine learning and deep learning techniques. The model leverages SHAP and LIME to interpret and visualize the contribution of features during classification.

🔍 Project Overview

Encrypted traffic classification is essential for modern cybersecurity systems to detect and monitor potential threats while preserving user privacy. This project:

• Trains a hybrid model to detect types of encrypted traffic (e.g., Tor, VPN, I2P)
• Applies SHAP and LIME to explain model predictions
• Demonstrates visualizations and interpretable results

🚀 Features

• Binary and multi-class classification support
• Model interpretability using SHAP & LIME
• Lightweight feature preprocessor
• Ready for plug-and-play with your own dataset

🗂️ Project Structure

data/ # Dataset samples or links models/ # Trained model files explainability/ # SHAP, LIME scripts and results src/ # Core Python scripts notebooks/ # Jupyter notebooks results/ # Output results and figures

🛠️ Getting Started

pip install -r requirements.txt
python src/train.py


📸 Screenshots
🔹 Streamlit Dashboard

    Upload CSV, toggle between Binary/Multi-Class, view results in real time.

🔹 SHAP Summary Plot – Binary

    Global feature importance from SHAP.

🔹 SHAP Summary Plot – Multi-Class

    Multi-class SHAP visualization.

🔹 LIME Local Explanation

    Per-instance interpretability via LIME.

### 📸 Screenshots

#### 🔹 Binary Confusion Matrix
![Conf Matrix Binary](results/conf_matrix_binary.png)

#### 🔹 SHAP Summary – Binary
![SHAP Binary](results/shap_summary_binary.png)

#### 🔹 SHAP Summary – Multi-Class
![SHAP Multi](results/shap_summary_multi.png)

#### 🔹 LIME Explanation – Binary Instance
(Open the `.html` file to interact)
➡️ [`lime_binary_instance.html`](results/lime_binary_instance.html)

#### 🔹 LIME Explanation – Multi-Class Instance
➡️ [`lime_multi_instance.html`](results/lime_multi_instance.html)