Satvanetra - Deepfake Detection System (Audio / Video / Image)

Hackathon Project | CIDECode 2025 | 2nd Place Winner | ₹50,000 Prize

Satvanetra is an advanced deepfake detection system designed to analyze audio, video, and image content using both AI/ML models and manual digital signal processing techniques. This project was created for CIDECode 2025, an annual hackathon organized by the Cybersecurity Department of CID Karnataka, where it secured 2nd place.

---

Project Overview

Satvanetra is built to detect and analyze manipulated multimedia content using a hybrid approach:

1. Automated AI/ML Detection

   • Users upload an object (image, video, or audio) via the Next.js frontend (hacknext).
   • Uploaded content is stored on Supabase, and a pre-signed URL is sent to the Flask backend (flaskBack).
   • The backend connects to machine learning models to analyze the content.
   • Results are compiled into a PDF report, stored on Supabase, and a pre-signed URL is sent back for display.

2. Manual Signal Analysis

   • Deep analysis of audio, video, and images using Fourier transforms and other digital signal processing techniques.
   • Provides a secondary verdict for a deeper understanding of potential manipulations.

3. Admin Panel

   • Complete control and monitoring of uploads, analysis results, and system activity.

---

⚡ Key Features

• Full-stack Next.js and Flask architecture
• Supabase storage and pre-signed URL integration
• AI/ML models for automated deepfake detection
• Manual digital signal processing for enhanced verification
• PDF report generation and secure distribution
• Admin panel for monitoring and management
• Robust handling of audio, video, and image data

---

🛠 Tech Stack

• Frontend: Next.js, React
• Backend: Flask (Python)
• Storage & Database: Supabase
• ML / AI: Custom deepfake detection models
• Data Processing: Fourier transforms, signal analysis
• Deployment: Cloud-hosted endpoints and storage

---

How It Works

1. User Upload

   • Upload multimedia content via the frontend.
   • Frontend stores the file on Supabase and generates a pre-signed URL.

2. Automated AI Analysis

   • Flask backend retrieves content using pre-signed URL.
   • Runs deepfake detection ML models.
   • Generates verdict and PDF report.

3. Manual Analysis

   • Performs signal processing on content for additional verification.
   • Complements AI verdict for more reliable results.

4. Results Delivery

   • PDF report and results sent back to frontend via pre-signed URL.
   • Users and admins can view results securely.

---

🏆 Hackathon Achievement

• Event: CIDECode 2025, Karnataka (Cybersecurity Department)
• Prize: 2nd Place | ₹50,000
• Scope: Demonstrated advanced AI/ML hybrid system for deepfake detection in real-world scenarios.

---

📁 Repository Structure

Satvanetra/
├─ hacknext/ # Next.js full-stack frontend
├─ flaskBack/ # Flask backend with ML integration
└─ README.md # Project documentation

---

💡 Why Satvanetra?

Satvanetra stands out by combining automated AI detection with manual signal analysis, providing a reliable, robust, and practical solution to combat deepfake threats. Its architecture is scalable, secure, and recruiter-ready, showcasing both your technical skills and innovation in AI/ML.

---

Next Steps / Improvements

• Expand AI models to cover more multimedia types
• Implement real-time detection streaming
• Enhance admin analytics dashboard
• Containerize backend for easier deployment

---