Hybrid Intrusion Detection System using Machine Learning & Deep Learning

A hybrid Intrusion Detection System (IDS) that combines Machine Learning and Deep Learning techniques to accurately identify malicious network traffic while reducing false positives. The system integrates XGBoost-based feature selection, Logistic Regression, LSTM, and fuzzy logic to improve detection performance on network intrusion datasets.

---

Overview

Cyberattacks continue to evolve in complexity, making traditional rule-based intrusion detection systems less effective. This project proposes a hybrid intrusion detection approach that combines the strengths of classical Machine Learning and Deep Learning algorithms for improved attack detection.

The system performs data preprocessing, feature engineering, feature selection, model training, and performance evaluation to classify network traffic as normal or malicious. By integrating fuzzy logic for uncertainty handling, the model reduces false-positive predictions while maintaining high detection accuracy.

---

Key Features

• Hybrid Machine Learning and Deep Learning architecture
• Network intrusion classification
• Data preprocessing and feature engineering
• XGBoost-based feature selection
• Logistic Regression and LSTM models
• Fuzzy logic for uncertainty handling
• Performance comparison of multiple models
• High detection accuracy with reduced false positives

---

Project Workflow

Network Traffic Dataset
          │
          ▼
 Data Cleaning & Preprocessing
          │
          ▼
 Feature Engineering
          │
          ▼
 XGBoost Feature Selection
          │
          ▼
 Machine Learning Model
(Logistic Regression)
          │
          ▼
 Deep Learning Model
(LSTM)
          │
          ▼
 Fuzzy Logic Integration
          │
          ▼
 Intrusion Classification
          │
          ▼
 Performance Evaluation

---

System Architecture

Network Dataset
      │
      ▼
Preprocessing
      │
      ▼
Feature Selection
 (XGBoost)
      │
      ▼
Hybrid Detection Engine
 ├── Logistic Regression
 └── LSTM Network
      │
      ▼
Fuzzy Logic Decision Layer
      │
      ▼
Normal / Attack Prediction

---

Methodology

The proposed intrusion detection system follows these stages:

• Data preprocessing and cleaning
• Feature engineering
• XGBoost feature importance analysis
• Logistic Regression training
• LSTM model training
• Prediction fusion using fuzzy logic
• Model evaluation using classification metrics

---

Technology Stack

Programming Language

• Python

Machine Learning

• Scikit-learn
• XGBoost

Deep Learning

• TensorFlow
• Keras
• LSTM

Data Processing

• Pandas
• NumPy

Visualization

• Matplotlib
• Seaborn

---

Performance

Achievements

• Detection Accuracy: 98.4%
• Reduced False Positives by 18%
• Improved intrusion detection using a hybrid ML-DL architecture

Evaluation Metrics

• Accuracy
• Precision
• Recall
• F1-Score
• Confusion Matrix

---

Project Structure

Hybrid-Intrusion-Detection/
│
├── dataset/
├── notebooks/
├── models/
├── results/
├── requirements.txt
├── README.md
└── LICENSE

---

Installation

Clone the repository

git clone https://github.com/GowthamRavuri01/hybrid-intrusion-detection-system.git

Navigate to the project

cd hybrid-intrusion-detection-system

Install dependencies

pip install -r requirements.txt

Run the notebook

Launch Jupyter Notebook and execute the notebooks in sequence for preprocessing, model training, and evaluation.

---

Future Improvements

• Real-time network traffic monitoring
• Explainable AI (XAI) for attack interpretation
• Lightweight deployment for edge devices
• Cloud-based intrusion detection
• Ensemble learning techniques
• Live dashboard for monitoring alerts

---

Author

Gowtham Ravuri

AI Engineer | Computer Vision Engineer | Software Engineer

• GitHub: https://github.com/GowthamRavuri01
• LinkedIn: https://www.linkedin.com/in/gowthamravuri

---

License

This project is licensed under the MIT License.