NeuralSentinel is a comprehensive solution for detecting and defending against deepfakes using a multi-layered approach. It combines watermarking, traditional image analysis, and advanced machine learning techniques to provide robust media authentication and manipulation detection.
This system provides a robust framework for:
- Watermarking authentic images to verify their integrity
- Detecting potential deepfakes using multiple analysis methods
- Managing and tracking verified versus manipulated media
- Providing a user-friendly dashboard for monitoring and analysis
- Watermarking System: Embeds secure digital signatures into images
- Deepfake Detection Engine: Utilizes traditional image analysis and ensemble machine learning for reliable detection
- White-listing Mechanism: Maintains a registry of known authentic images to reduce false positives
- User Dashboard: Web interface for uploading, testing, and monitoring images
- Robust API: RESTful endpoints for seamless integration with other applications
-
Clone the repository:
git clone https://github.com/MLNexusHub/NeuralSentinel.git cd NeuralSentinel -
Create and activate a virtual environment:
python -m venv myenv source myenv/bin/activate # On Windows: myenv\Scripts\activate
-
Install dependencies:
pip install -r requirements.txt
-
Set up environment variables:
cp .env.example .env # Edit the .env file with your configuration as needed
Run the Flask-based dashboard server:
python dashboard.pyAccess the dashboard at: http://localhost:5000
| Endpoint | Description |
|---|---|
/api/upload |
Upload and watermark an image |
/api/test |
Test an image for watermark and deepfake detection |
/api/assets |
Retrieve all media assets |
/api/alerts |
Get detection alerts |
/api/logs |
Fetch system logs |
Implemented in watermarking.py, featuring:
- Invertible Neural Network (INN) for watermark encoding and decoding
- Error Correction Code (ECC) to improve watermark robustness
- Methods to embed and extract secure digital signatures
Implemented in deepfake_detector.py, including:
- Traditional image analysis (noise patterns, compression artifacts, face consistency)
- Ensemble machine learning for enhanced detection accuracy
- White-listing of verified authentic images
Implemented in dashboard.py:
- Flask web server providing RESTful APIs
- User interface for image upload, testing, and monitoring
- Optional Firebase integration for backend data management
This project is licensed under the MIT License.
- Prajwal Jahagirdar (@jahagirdarPrajwal)
- Aditya Singh (@EchoSingh)