Privacy Guard AI Chat System

An AI-powered privacy protection platform that masks Personally Identifiable Information (PII) from user conversations and uploaded documents before cloud processing.

Live Demo

Frontend Application: https://privacy-guard-ai-chat-system-2tmzwgyo5-ramvignesh-rs-projects.vercel.app/

• BERT model is disabled due to free tier cloud deployment hence feedback loop is not functional in this deployment.

Features

AI Chat Privacy Protection

• Real-time PII masking before sending data to cloud AI models
• Sensitive information detection and anonymization
• Gemini AI integration
• Interactive chat interface
• Active learning feedback system for improving masking quality

Document OCR + Masking

• Upload and process:

  • PDF
  • TXT
  • PNG
  • JPG
  • JPEG
  • DOCX

• OCR support using Tesseract

• Automatic sensitive data masking

• Secure document processing

Deep Learning Engine

• BiLSTM-based custom masking model
• Transformer-based NER pipeline
• Continuous learning workflow
• Known entities memory storage

---

Tech Stack

Frontend

• React.js
• Vite
• Lucide React
• CSS
• Vercel Deployment

Backend

• FastAPI
• Uvicorn
• SQLAlchemy
• Pydantic
• Python

AI / Machine Learning

• PyTorch
• HuggingFace Transformers
• BiLSTM
• TorchCRF
• Scikit-learn
• SpaCy
• Gemini API

OCR / Document Processing

• Tesseract OCR
• PyMuPDF
• pdf2image
• Pillow
• python-docx

Deployment

• Vercel (Frontend)
• Railway (Backend + OCR Services)
• Docker

---

System Architecture

Frontend (Vercel)
    ↓
Chat Backend (Railway + FastAPI)
    ↓
PII Detection & Masking
    ↓
Gemini AI Processing

Document OCR Service (Railway + Flask)
    ↓
OCR Extraction
    ↓
PII Masking

---

Project Structure

Privacy_Guard_AI_Chat_System/
│
├── backend/
│   ├── model_output/
│   ├── main.py
│   ├── pii_pipeline.py
│   ├── bilstm_model.py
│   ├── database.py
│   ├── models.py
│   ├── requirements.txt
│   └── .env
│
├── DocumentMasker/
│   ├── utils/
│   │   └── ocr_extractor.py
│   ├── web/
│   │   ├── app.py
│   │   └── requirements.txt
│   └── Dockerfile
│
├── frontend/
│   ├── src/
│   │   ├── components/
│   │   ├── App.jsx
│   │   └── main.jsx
│   ├── package.json
│   └── vite.config.js
│
└── README.md

---

Installation Guide

1. Clone Repository

git clone https://github.com/Ramvignesh-R/Privacy_Guard_AI_Chat_System.git

cd Privacy_Guard_AI_Chat_System

---

Frontend Setup

cd frontend
npm install
npm run dev

Frontend runs on:

http://localhost:5173

---

Backend Setup

Install Dependencies

cd backend
pip install -r requirements.txt

Create .env File

GOOGLE_API_KEY=your_gemini_api_key

Run Backend

uvicorn main:app --reload

Backend runs on:

http://localhost:8000

---

OCR Service Setup

Install Dependencies

cd DocumentMasker/web
pip install -r requirements.txt

Install OCR Requirements

Ubuntu / Debian

sudo apt-get update
sudo apt-get install -y tesseract-ocr poppler-utils

macOS

brew install tesseract poppler

Windows

Install:

• Tesseract OCR
• Poppler

Run OCR Service

python app.py

OCR service runs on:

http://localhost:5000

---

Environment Variables

Frontend (.env)

VITE_CHAT_API_URL=http://localhost:8000
VITE_OCR_API_URL=http://localhost:5000

Production Environment Variables

VITE_CHAT_API_URL=https://your-railway-chat-url.up.railway.app
VITE_OCR_API_URL=https://your-railway-ocr-url.up.railway.app

---

Deployment

Frontend Deployment (Vercel)

1. Push frontend code to GitHub
2. Import repository into Vercel
3. Set Root Directory:

frontend

4. Add environment variables
5. Deploy

---

Backend Deployment (Railway)

1. Create Railway project
2. Connect GitHub repository
3. Set Root Directory:

backend

4. Add environment variables
5. Deploy FastAPI service

---

OCR Service Deployment (Railway)

1. Create second Railway project
2. Set Root Directory:

DocumentMasker

3. Use Docker deployment
4. Deploy OCR service

---

API Endpoints

Chat Endpoint

POST /chat

Request

{
  "text": "My phone number is 9876543210"
}

---

Feedback Endpoint

POST /report_bad_masking

---

Screenshots

Chat Interface

• Real-time PII masking
• Gemini integration
• Secure AI interaction

Document OCR

• Upload documents
• OCR extraction
• Masked output generation

---

Security Highlights

• Sensitive information masking before AI processing
• Privacy-first architecture
• Local preprocessing workflow
• Entity masking pipeline
• Active learning retraining system

---

Future Improvements

• Multi-language OCR support
• Speech-to-text privacy masking
• Real-time voice anonymization
• Advanced entity detection
• User authentication system
• Analytics dashboard
• PDF export support

---

Contributors

Ramvignesh R and Ritujaa BG

Integrated M.Sc Data Science PSG College of Technology

---

License

This project is developed for educational and research purposes.

---

Acknowledgements

• Google Gemini API
• HuggingFace Transformers
• PyTorch
• FastAPI
• Railway
• Vercel
• Tesseract OCR