ML Deepfake Detection & Defense System

A comprehensive solution for detecting and defending against deepfakes using a multi-layered approach combining watermarking, traditional image analysis, and machine learning techniques.

Overview

This system provides a robust framework for:

1. Watermarking authentic images to verify their integrity
2. Detecting potential deepfakes using multiple analysis methods
3. Managing and tracking verified vs. manipulated media
4. Providing a user-friendly dashboard for monitoring and analysis

Key Features

• Watermarking System: Embeds secure digital signatures into images
• Deepfake Detection Engine: Utilizes multiple analysis techniques to identify manipulated images
• White-listing Mechanism: Maintains a registry of known authentic images
• User Dashboard: Web interface for uploading, testing, and monitoring images
• Robust API: RESTful endpoints for integration with other systems

Installation

1. Clone the repository:

git clone https://github.com/JahagirdarPrajwal/ML-DeepFake-Detection-Defense--BAM-.git
cd ML-DeepFake-Detection-Defense--BAM-

2. Create and activate a virtual environment:

python -m venv myenv
source myenv/bin/activate  # On Windows: myenv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Set up environment variables:

cp .env.example .env
# Edit .env with your configuration

Usage

Starting the Server

python dashboard.py

The dashboard will be available at: http://localhost:5000

API Endpoints

• /api/upload - Upload and watermark an image
• /api/test - Test an image for watermarks and deepfake detection
• /api/assets - Get all media assets
• /api/alerts - Get detection alerts
• /api/logs - Get system logs

Components

Watermarking Module

The watermarking.py file implements:

• Simple Invertible Neural Network (INN) for encoding/decoding
• Error Correction Code (ECC) for robust watermark extraction
• Methods for embedding and extracting watermarks

Deepfake Detection Module

The deepfake_detector.py file implements:

• Traditional image analysis methods (noise, compression artifacts, face consistency)
• Ensemble approach for more reliable detection
• White-listing system for known authentic images

Dashboard

The dashboard.py file provides:

• Flask web server
• API endpoints for image processing
• Integration with Firebase (optional)
• File serving and management

License

MIT

Contributors

• Prajwal Jahagirdar