🐶 DeepSafe: Enterprise-Grade Deepfake Detection Platform

---

DeepSafe is a modular, high-performance, and containerized platform designed for the robust detection of deepfakes in digital media. By aggregating state-of-the-art detection models into a unified ensemble, DeepSafe provides enterprise-grade accuracy and reliability.

🚀 Why DeepSafe?

Platform Agnostic & Dependency Isolated

DeepSafe adopts a microservices architecture where each detection model runs in its own isolated Docker container. This design choice is deliberate and critical:

• No Dependency Hell: Each model can have its own specific version of PyTorch, CUDA, or other libraries without conflicting with others.
• Platform Independent: Whether you are on Linux, macOS, or Windows, if you have Docker, DeepSafe works.
• Scalable: Scale individual models based on load.

Key Features

• Enterprise-Grade Authentication: Secure Login and Registration system to protect access.
• Multi-Modal Detection: Analyzes both Images and Videos for manipulation.
• Ensemble Intelligence: Combines multiple state-of-the-art models (NPR, UniversalFakeDetect, CrossEfficientViT) for superior accuracy.
• Meta-Learning Engine: Dynamically stacks model outputs using advanced meta-learners to reduce false positives.
• Premium UI/UX: A modern, dark-themed React dashboard with interactive charts and real-time feedback.
• Dockerized Architecture: Fully containerized services for easy deployment and isolation.
• RESTful API: Robust FastAPI backend with health checks, batch processing, and detailed logging.

📸 UI Preview

🏗️ Architecture

DeepSafe orchestrates a fleet of specialized detectors via a central API gateway:

graph TD
    Client[Web Client / User] --> Nginx[Nginx / Frontend]
    Nginx --> API[DeepSafe API FastAPI]
    API --> Meta[Meta-Learner Ensemble]
    
    subgraph "Image Detectors"
        Meta --> NPR[NPR Deepfake]
        Meta --> UFD[Universal Fake Detect]
    end
    
    subgraph "Video Detectors"
        Meta --> CEV[Cross Efficient ViT]
    end

Loading

🛠️ Quick Start

Prerequisites

• Docker & Docker Compose
• Git

Installation

1. Clone the repository:

   git clone https://github.com/siddharthksah/DeepSafe.git
   cd DeepSafe

2. Launch the Platform:

   make start

   This will build all containers and start the services. Initial build may take a few minutes.

3. Access the Dashboard: Open http://localhost:8888 in your browser.

4. API Documentation: Visit http://localhost:8000/docs for the interactive Swagger UI.

📦 Available Models

Model	Type	Status	Description
NPR Deepfake	Image	✅ Active	Neural Pattern Recognition for subtle artifact detection.
Universal Fake Detect	Image	✅ Active	Generalizable detection for unseen deepfake types.
Cross Efficient ViT	Video	✅ Active	High-efficiency video analysis using Vision Transformers.
FakeSTormer	Video	✅ Active	Vulnerability-Aware Spatio-Temporal Learning for Generalizable Deepfake Video Detection.

🧪 Testing & Verification

DeepSafe includes a comprehensive test suite to ensure system integrity.

# Run health checks and basic functionality tests
make test

🤝 Contributing

We welcome contributions! Please see our Contributing Guidelines and Code of Conduct.

📄 License

Distributed under the MIT License. See LICENSE for more information.

📖 Citing DeepSafe

If you find DeepSafe useful in your research or work, please consider citing it:

@misc{deepsafe,
  author = {Siddharth Kumar},
  title = {DeepSafe: Enterprise-Grade Deepfake Detection Platform},
  year = {2025},
  publisher = {GitHub},
  url = {https://github.com/siddharthksah/DeepSafe}
}

🏆 Credits

DeepSafe integrates and builds upon the following excellent open-source research:

• NPR Deepfake: GitHub
• Universal Fake Detect: GitHub
• Cross Efficient ViT: GitHub
• FakeSTormer: GitHub

We thank the original authors for their contributions to the community.