EmpowerNet AI

Multi-modal deepfake & scam detection for browsers, video calls, and digital media.

---

Overview

EmpowerNet AI is a full-stack forensic intelligence platform designed to detect AI-generated scams, deepfakes, voice cloning, and synthetic media across text, images, audio, and video.

It combines a React dashboard, FastAPI backend, Chrome Extension, machine learning inference pipelines, and a cryptographic evidence ledger into a unified workflow for digital threat analysis.

Blockchain note: The evidence ledger generates a deterministic SHA-256 proof hash for every scan and stores it in a persistent local log. It is architected to submit transactions to Polygon Amoy when PRIVATE_KEY and CONTRACT_ADDRESS are configured — currently running in local simulation mode by default.

---

Features

Area	Capability
🔍 Multi-modal Scanning	Text, Image, Audio, and Video deepfake/scam detection
🌐 Browser Extension	Real-time protection on any webpage (Chrome Manifest V3)
🎭 Deepfake Detection	Vision ensemble (EfficientNet-B5, ViT, Organika/sdxl-detector)
🎙️ Voice Clone Detection	MFCC + F0 analysis + optional Whisper STT
⛓️ Evidence Ledger	SHA-256 proof hashing, Polygon Amoy-ready (simulation mode default)
📄 Forensic Reports	PDF export + direct submission workflow to cybercrime.gov.in
🛡️ Meeting Shield	Real-time frame + audio analysis for Google Meet / Zoom
👁️ Social Media Scanner	Profile image and post content analysis
👶 Child Safety	NSFW and cyberbullying content filtering

---

Architecture

flowchart LR

    subgraph Clients
        A["React Dashboard"]
        B["Chrome Extension"]
    end

    subgraph Backend
        C["/api/scan"]
        D["/realtime/video"]
        E["/realtime/audio"]
        F["/analyze/*"]
    end

    subgraph ML
        G["Text Pipeline"]
        H["Image Pipeline"]
        I["Audio Pipeline"]
        J["Video Pipeline"]
    end

    subgraph Evidence
        K["SHA-256 Hasher"]
        L["Blockchain Service (Simulation / Polygon Amoy)"]
    end

    A --> C
    B --> C
    C --> G
    C --> H
    C --> I
    C --> J
    D --> J
    E --> I
    F --> G
    G --> K
    H --> K
    I --> K
    J --> K
    K --> L

Loading

ML Pipeline Detail

flowchart TD
    Input["User Input\n(Text / Image / Audio / Video)"]

    Input --> T["Text Infer\nDistilBERT + RoBERTa + FinBERT\n+ keyword heuristic engine"]
    Input --> Im["Image Infer\nELA + Vision Ensemble\n(EfficientNet-B5, ViT × 3)\nor Gemini Flash (cloud mode)"]
    Input --> Au["Audio Infer\nMFCC + F0 + Whisper STT\n+ motheecreator model"]
    Input --> V["Video Infer\n5-point frame sampling\n+ FFmpeg audio extract"]

    T --> R["Risk Engine\nConfidence Scoring + Labelling"]
    Im --> R
    Au --> R
    V --> R

    R --> H["SHA-256 Hasher"]
    H --> BC["BlockchainService\n(Simulation / Polygon Amoy)"]
    BC --> Resp["Structured API Response"]

Loading

---

Tech Stack

Category	Technologies
Frontend	React 18 · TypeScript · Vite · Tailwind CSS · Zustand · Framer Motion
Backend	FastAPI · Python 3.10+ · Uvicorn · Pydantic v2
ML — Vision	PyTorch · HuggingFace Transformers · EfficientNet-B5 · ViT (×3) · OpenCV · MediaPipe
ML — Audio	Librosa · Whisper-Tiny · motheecreator/Deepfake-audio-detection
ML — Text	Fine-tuned transformer · unitary/toxic-bert · ProsusAI/FinBERT · Keyword heuristic engine
Evidence Ledger	SHA-256 · web3.py · Solidity · Polygon Amoy Testnet (simulation mode default)
Browser Extension	Chrome Manifest V3 · Service Worker · Offscreen API
Deployment	Docker · Railway · Vercel

---

Project Structure

EmpowerNet-AI/
├── backend/
│   ├── api/
│   │   └── routers/        # Modular FastAPI route handlers
│   │       ├── scan.py
│   │       ├── realtime.py
│   │       └── websocket.py
│   ├── blockchain/         # Evidence ledger (simulation-ready Polygon Amoy)
│   ├── config/             # Centralized settings (Settings class)
│   ├── ml/                 # Inference pipelines per modality
│   │   ├── text_infer.py
│   │   ├── image_infer.py
│   │   ├── audio_infer.py
│   │   ├── video_plus_infer.py
│   │   ├── ml_service.py   # Vision ensemble + real-time video/audio
│   │   └── child_safety.py
│   ├── models/             # Pydantic request/response schemas
│   └── main.py             # App entrypoint — registers all routers
│
├── frontend/
│   ├── pages/              # React page components
│   ├── services/           # API client (fetch wrappers)
│   └── store/              # Zustand global state
│
├── empowernet-extension/   # Chrome Extension (Manifest V3)
│   ├── background.js
│   ├── content.js
│   ├── popup.html / popup.js
│   └── manifest.json
│
├── docs/                   # Architecture diagrams and documentation
├── .env.example            # Environment variable template
├── Dockerfile              # Production Docker image (Railway)
├── CONTRIBUTING.md
├── SECURITY.md
└── CHANGELOG.md

---

Run Locally

Prerequisites

• Python ≥ 3.10
• Node.js ≥ 18.x

Backend

cd backend
pip install -r requirements.txt

# Copy and populate environment variables
cp ../.env.example ../.env

uvicorn main:app --reload --port 8001

Frontend

cd frontend
npm install
npm run dev
# → http://localhost:3001

Chrome Extension

1. Open chrome://extensions
2. Enable Developer mode
3. Click Load unpacked → select the empowernet-extension/ directory

---

Environment Variables

# ML Configuration
LOAD_MODELS=false          # Set to true to enable local heavy models (GPU recommended)
GEMINI_API_KEY=            # Required for cloud vision inference when LOAD_MODELS=false

# Evidence Ledger — leave blank to use simulation mode (default)
POLYGON_RPC_URL=https://rpc-amoy.polygon.technology/
PRIVATE_KEY=               # Optional: Polygon wallet private key
CONTRACT_ADDRESS=           # Optional: Deployed EvidenceRegistry address

---

Screenshots

---

Contributing

See CONTRIBUTING.md for setup instructions, code style guidelines, and the pull request process.

Security

See SECURITY.md for the vulnerability reporting process and security disclosures.

License

This project is licensed under the MIT License — see LICENSE for details.