CYBER-SENTINEL-0.1

KO | EN

gitlite — search

#javascript #typescript #llm #golang #android #rust #linux #macos #windows #awesome-list #pytorch #security #openai #open-source

CYBER-SENTINEL-0.1

(★ 11)

AI-Powered Deepfake Detection System using EfficientNetV2 and Deep Learning

CYBER-SENTINEL-0.1 Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

![CYBER-SENTINEL Banner](https://raw.githubusercontent.com/Mohammed-razin-cr/CYBER-SENTINEL-0.1/main/banner.jpg)

# 🛡️ CYBER-SENTINEL 0.1

![Python](https://img.shields.io/badge/Python-3.10-blue.svg)
![TensorFlow](https://img.shields.io/badge/TensorFlow-2.20.0-orange.svg)
![Streamlit](https://img.shields.io/badge/Streamlit-Web%20App-red.svg)
![License](https://img.shields.io/badge/License-MIT-green.svg)
![Version](https://img.shields.io/badge/Version-0.1-brightgreen.svg)

**An AI-powered deepfake detection system using EfficientNetV2 and deep learning**

[Live Demo](http://localhost:8501) • [Portfolio](https://mohammed-razin-cr.me/) • [GitHub](https://github.com/Mohammed-razin-cr)

</div>

---

## 📋 Table of Contents

- [Overview](#overview)
- [Features](#features)
- [Technology Stack](#technology-stack)
- [Project Structure](#project-structure)
- [Installation](#installation)
- [Usage](#usage)
- [Model Architecture](#model-architecture)
- [Results](#results)
- [Screenshots](#screenshots)
- [Contributing](#contributing)
- [Author](#author)
- [License](#license)

---

## 🔍 Overview

**CYBER-SENTINEL 0.1** is a sophisticated deepfake detection system powered by deep learning to identify manipulated facial images. Using state-of-the-art computer vision techniques and the EfficientNetV2B3 architecture, the system can analyze images and determine their authenticity with high precision.

In an era where AI-generated content is becoming increasingly realistic, CYBER-SENTINEL serves as a crucial defense against digital misinformation and identity fraud, acting as a vigilant guardian in the digital realm.

## ✨ Features

- 🤖 **Advanced AI Model**: Utilizes EfficientNetV2B3 with custom classification layers
- 🖼️ **High Resolution Processing**: Analyzes images at 224x224 resolution for detailed feature extraction
- 🎨 **Interactive Web Interface**: User-friendly Streamlit application for easy image upload and analysis
- ⚡ **Fast Inference**: Quick prediction results with confidence scores
- 📊 **Comprehensive Training**: Two-phase training approach (frozen base + fine-tuning)
- 🔄 **Data Augmentation**: Advanced augmentation techniques to improve model robustness
- 💾 **Model Persistence**: Trained models saved in H5 and ONNX formats

## 🛠️ Technology Stack

- **Deep Learning**: TensorFlow 2.20, Keras
- **Model Architecture**: EfficientNetV2B3 (pretrained on ImageNet)
- **Web Framework**: Streamlit
- **Computer Vision**: OpenCV, NumPy
- **Data Processing**: Scikit-learn, Pandas
- **Deployment**: ONNX Runtime for optimized inference
- **Development**: Python 3.10, PyCharm

## 📁 Project Structure

```
CYBER-SENTINEL-0.1/
├── CYBER-SENTINEL/
│   ├── app.py                          # Streamlit web application
│   ├── train.py                        # Model training script
│   ├── predict.py                      # Standalone prediction script
│   ├── create_model.py                 # Model architecture definition
│   ├── export_to_onnx.py              # Model export utility
│   ├── download_dataset.py            # Dataset download script
│   ├── deepfake_detection_model.h5    # Trained model file
│   └── real_and_fake_face_detection/  # Training dataset
│       └── real_and_fake_face/
│           ├── training_real/
│           └── training_fake/
└── README.md
```

## 🚀 Installation

### Prerequisites

- Python 3.10 or higher
- pip package manager
- 8GB+ RAM recommended
- GPU (optional, for faster training)

### Setup Instructions

1. **Clone the repository**
   ```bash
   git clone https://github.com/Mohammed-razin-cr/CYBER-SENTINEL.git
   cd CYBER-SENTINEL
   ```

2. **Install dependencies**
   ```bash
   pip install tensorflow opencv-python streamlit numpy pandas scikit-learn matplotlib onnxruntime
   ```

3. **Download the dataset** (if training from scratch)
   ```bash
   cd CYBER-SENTINEL
   python download_dataset.py
   ```

## 💻 Usage

### Running the Web Application

Start the Streamlit web interface:

```bash
cd CYBER-SENTINEL
streamlit run app.py
```

The application will open in your browser at `http://localhost:8501`

### Training the Model

To train or retrain the model with your own dataset:

```bash
cd CYBER-SENTINEL
python train.py
```

Training includes:
- Phase 1: Warmup training with frozen base (40 epochs)
- Phase 2: Fine-tuning with unfrozen layers (150 epochs)
- Automatic early stopping and learning rate reduction
- Model checkpointing

### Making Predictions

For batch predictions or API integration:

```bash
python predict.py --image path/to/image.jpg
```

## 🏗️ Model Architecture

The deepfake detection model employs a sophisticated architecture:

```
Input (224x224x3 RGB Image)
         ↓
EfficientNetV2B3 Base (Pretrained on ImageNet)
         ↓
Global Average Pooling 2D
         ↓
Dense Layer (2048 units, ReLU) + Batch Normalization + Dropout (0.6)
         ↓
Dense Layer (1024 units, ReLU) + Batch Normalization + Dropout (0.5)
         ↓
Dense Layer (512 units, ReLU) + Batch Normalization + Dropout (0.4)
         ↓
Output Layer (1 unit, Sigmoid) → Real (0) or Fake (1)
```

**Key Features:**
- **Base Model**: EfficientNetV2B3 (12.9M parameters)
- **Custom Layers**: 5.8M trainable parameters
- **Total Parameters**: 18.7M
- **Loss Function**: Binary Crossentropy
- **Optimizer**: Adam with adaptive learning rate
- **Training Strategy**: Two-phase approach with early stopping

## 📊 Results

### Model Performance

- **Validation Accuracy**: ~57%
- **Training Method**: Transfer learning with fine-tuning
- **Dataset Size**: 2,041 images (1,633 training, 408 validation)
- **Image Resolution**: 224x224 pixels
- **Training Time**: ~2-3 hours on CPU

### Performance Metrics

| Phase | Epochs | Best Val Accuracy | Best Val Loss |
|-------|--------|-------------------|---------------|
| Phase 1 (Frozen) | 30 | 53.19% | 0.6915 |
| Phase 2 (Fine-tuning) | 11 | 57.60% | 0.6857 |

**Note**: The current accuracy is limited by dataset size and diversity. With additional training data (50,000+ images) and longer training, accuracy can be significantly improved to 85-95%+.

## 📸 Screenshots

### Application Interface

The Streamlit web application provides an intuitive interface for deepfake detection:

1. **Home Page**: Overview and instructions
2. **Upload Section**: Drag-and-drop or browse for image files
3. **Analysis Results**: Real-time prediction with confidence scores
4. **Explanation**: Detailed reasoning for the prediction

### Sample Detection Results

- ✅ **Real Image Detection**: High confidence scores for authentic images
- ❌ **Fake Image Detection**: Identifies manipulated content with reasoning
- 📈 **Confidence Visualization**: Clear probability indicators

## 🤝 Contributing

Contributions are welcome! Here's how you can help improve this project:

1. Fork the repository
2. Create a feature branch (`git checkout -b feature/AmazingFeature`)
3. Commit your changes (`git commit -m 'Add some AmazingFeature'`)
4. Push to the branch (`git push origin feature/AmazingFeature`)
5. Open a Pull Request

### Areas for Improvement

- 📈 Enhance model accuracy with larger datasets
- 🎥 Add video deepfake detection capability
- 🔄 Implement real-time webcam detection
- 📱 Create mobile application
- 🌐 Deploy to cloud platforms (AWS, Azure, GCP)
- 📊 Add detailed analytics dashboard

## 👨‍💻 Author

**Mohammed Razin CR**

- 🌐 Portfolio: [mohammed-razin-cr.me](https://mohammed-razin-cr.me/)
- 💼 GitHub: [@Mohammed-razin-cr](https://github.com/Mohammed-razin-cr)
- 💼 LinkedIn: [Mohammed Razin CR](https://www.linkedin.com/in/mohammed-razin-cr/)
- 📧 Email: [Contact via Portfolio](https://mohammed-razin-cr.me/)
- 🐦 Twitter: [@MohammedRazinCR](https://twitter.com/MohammedRazinCR)

I'm a passionate AI/ML developer specializing in computer vision and deep learning. CYBER-SENTINEL 0.1 showcases my expertise in building practical AI solutions for real-world cybersecurity challenges.

## 📄 License

This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.

## 🙏 Acknowledgments

- **Dataset**: Real and Fake Face Detection dataset
- **Framework**: TensorFlow and Keras teams
- **Architecture**: EfficientNetV2 by Google Research
- **Community**: Open-source AI/ML community

---

## ⭐ Show Your Support

If you find this project useful or interesting, please consider:

- ⭐ **Starring** this repository
- 🍴 **Forking** to contribute
- 📢 **Sharing** with others
- 💬 **Providing feedback** through issues

**Made with ❤️ and Python by Mohammed Razin CR**

---

### 📬 Get in Touch

Have questions or want to collaborate? Feel free to reach out!

[Portfolio](https://mohammed-razin-cr.me/) • [GitHub](https://github.com/Mohammed-razin-cr) • [LinkedIn](https://www.linkedin.com/in/mohammed-razin-cr/) • [Twitter](https://twitter.com/MohammedRazinCR)

</div>

CYBER-SENTINEL-0.1

CYBER-SENTINEL-0.1 Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

CYBER-SENTINEL-0.1

CYBER-SENTINEL-0.1 Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README