Obscurix

Obscurix - Comprehensive Technical Documentation

Secure Communication & Redaction Platform with Generative AI, OCR, and PII Intelligence

---

Table of Contents

1. Project Overview
2. Features
3. Architecture & Tech Stack
4. Frontend Structure
5. API Endpoints
6. Component Architecture
7. Authentication & User Management
8. Data Flow & Integration
9. Deployment & CI/CD
10. Security Implementation
11. Development Guidelines
12. Run Locally

---

Project Overview

Obscurix is a next-generation data redaction and content moderation platform designed to help teams and enterprises securely communicate and collaborate without risking sensitive data exposure. The chatbot interface, Redactify, enables users to interact naturally while ensuring compliance, privacy, and safety—powered by Generative AI and real-time monitoring. Optical Character Recognition (OCR), and advanced PII detection techniques, Obscurix ensures sensitive data is automatically detected, redacted, or rewritten before exposure.

Key Objectives

• Data Privacy: Automated detection and redaction of sensitive content across text and images.
• AI-Powered Rewriting: Transformation of sensitive content while preserving meaning and context.
• Compliance Monitoring: Real-time risk analysis via a centralized dashboard.
• Secure Collaboration: Converts source code into readable pseudocode, protecting intellectual property.

---

Features

Feature	Description
PII Redaction Engine	Real-time masking of sensitive data such as emails, phone numbers, passwords, API keys, card numbers, and bank account details using regex and NLP.
Generative AI Rewriting	Uses Google Gemini to rewrite redacted content while preserving the original context and softening tone, especially in sensitive or negative sentiment messages.
Admin Monitoring Dashboard	Displays analytics for flagged content, user behavior trends, and PII redaction frequency in real time.
Multilingual Redaction Support	Redaction and rewriting supported for English, Hindi, and Spanish using spaCy’s NER and GenAI’s language-agnostic capabilities.
Dual-Mode Redaction (Strict & Creative)	Strict mode performs full redaction; Creative mode redacts and then rephrases the content while maintaining meaning.
Leak Prevention Browser Extension	Chrome/Edge extension that uses OCR and NLP to detect and block screenshots or copied code/text containing sensitive info before it is posted to social media.
Code Sanitization & Pseudocode Conversion	Automatically converts code to anonymized pseudocode to ensure safe collaboration without exposing internal logic or credentials.
Sentiment-Aware Moderation	Detects and softens toxic or harmful language, ensuring that content is emotionally neutral or constructive.
Cognitive Code Obfuscation via GenAI	Converts flagged code into obfuscated pseudocode while explaining the logic, allowing safe public sharing (e.g., StackOverflow).
Sensitivity Scoring System	Assigns a numeric risk score (0–100) to messages based on redacted content and sentiment, enabling proactive flagging.
Audit Trail & Reporting Dashboard	Tracks redaction activity, sentiment trends, and high-risk users or teams; helps enterprises monitor compliance.

---

Architecture & Tech Stack

Frontend

Technology	Purpose	Version
Next.js	React Framework	Latest
React	UI Library	18+
Tailwind CSS	Styling Framework	3+
Framer Motion	Animations	Latest
Lucide React	Icon Set	Latest
Axios	HTTP Client	Latest

Backend

Technology	Purpose
Django REST Framework	API Layer
Python 3.11+	Backend Logic
spaCy	Named Entity Recognition
Google Gemini API	Generative AI Engine
EasyOCR	OCR for Image Text Extraction
CI/CD	GitHub Actions + AWS EC2

Database & Storage

Technology	Purpose
MongoDB Atlas	Authentication, Session & Logs
Local Storage	Client-side State Management
Extensions	Custom Chrome/Edge extension using OCR

---

Frontend Structure

Directory Overview

frontend/
├── components/         # Reusable React components
├── pages/              # Route-based structure
│   ├── index.js
│   ├── login.js
│   ├── signup.js
│   ├── redactify.js
│   ├── profile.js
│   ├── dashboard.js
│   └── ocr.js
├── public/             # Static assets
├── styles/             # Tailwind and global styles
└── utils/              # API clients and helpers

Key Pages

• Home: Animated landing page with responsive layout.
• Login / Signup: Authentication with validation and animated visuals.
• Redactify: Main chat interface for redaction and rewriting.
• Profile: Editable user profile with real-time sync.
• Dashboard: Admin-only analytics view.
• OCR: Upload interface for image-based redaction.

---

API Endpoints

Authentication

Endpoint	Method	Description
/api/signup	POST	User registration with roles
/api/login	POST	Authenticates user, returns JWT

Core Functionality

Endpoint	Method	Description
/api/redact	POST	Handles PII detection and rewriting
/api/metrics	GET	Dashboard metrics for admin view
/api/updateUser	PUT	Update user credentials and profile

OCR & Extensions

Endpoint	Method	Description
/api/ocr	POST	OCR-based text extraction from images

---

Component Architecture

Primary Components

1. Navbar

   • Role-aware navigation with profile dropdown and route guards.

2. Hero

   • Animated intro section using Framer Motion.

3. Redactify

   • Real-time chat with strict vs. creative redaction modes.

4. Dashboard

   • Privacy metrics, user analytics, and visual trend indicators.

5. Animated Backgrounds

   • Starfield effects for aesthetic enhancement.

6. News Ticker

   • Scrollable text feed showing platform updates.

7. Footer

   • Contact form and site navigation links.

---

Authentication & User Management

Flow

1. Signup

   • Collect user credentials and role, store in MongoDB.

2. Login

   • Validate against hashed credentials and return JWT.

3. Session Handling

   • Store JWT securely in localStorage for API auth.

4. Access Control

   • Route guards enforce role-based access on frontend/backend.

User Roles

Role	Permissions
User	Redactify, OCR, Profile
Admin	Full access including Dashboard views

---

Data Flow & Integration

Core Flows

• Frontend → Backend All secure requests carry JWTs in headers for validation.

• Redaction Request Input sent to /api/redact, processed via spaCy, Gemini, and EasyOCR.

• Dashboard Metrics Admin panels fetch data from /api/metrics and visualize user risk.

• Extension Hooks OCR triggers NLP pipeline before permitting screen-based actions.

State Management

• React Hooks + Context API for runtime state.
• localStorage for JWT tokens and session persistence.

---

Deployment & CI/CD

Frontend

• Platform: Vercel
• Branch: main
• Integration: GitHub Auto-deploy
• Performance: Global CDN + asset optimization

Backend

• Platform: AWS EC2 (Amazon Linux 2023, t2.micro)
• Stack: Gunicorn + Nginx
• Database: MongoDB Atlas
• Monitoring: CloudWatch alerts (CPU > 90%)
• Security: Open ports only for SSH (22), HTTP (80), HTTPS (443), and API (8000)

CI/CD via GitHub Actions

# .github/workflows/deploy-backend.yml
on:
  push:
    branches: [main]
jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: SSH & Deploy
        uses: appleboy/ssh-action@v0.1.6
        with:
          host: $
          username: ec2-user
          key: $
          script: |
            cd /home/ec2-user/obscurix/backend
            git pull origin main
            pip install -r requirements.txt
            sudo systemctl restart gunicorn
            sudo systemctl reload nginx

---

Security Implementation

1. Authentication Security

   • Passwords are hashed using bcrypt.
   • JWT tokens expire and are validated on each request.
   • Frontend and backend routes are protected by role checks.

2. Data Protection

   • HTTPS is enforced for all communication.
   • PII is never stored in raw form. Only metadata is logged.

3. Privacy Features

   • Dashboard exposes redaction logs and scores.
   • Minimal data collection with user control over data retention.

4. API Security

   • Input sanitization, rate limiting, and validation implemented on all endpoints.

---

Development Guidelines

Code Organization

• Follow modular design with separation of components, pages, and services.

State Management

• Prefer hooks and React Context over prop drilling.
• Use localStorage only for non-sensitive session data.

Error Handling

• Use try-catch with user feedback alerts.
• Centralized error logger for debugging.

Styling & UX

• Tailwind CSS for utility-first styling.
• Maintain design consistency with shared styles and theme tokens.

Accessibility

• Use semantic HTML and ARIA attributes where appropriate.

Testing Strategy

• Unit Testing: Component-level via Jest.
• Integration Testing: API routes tested using Supertest.
• End-to-End Testing: User flow validation via Cypress.
• Performance Testing: Simulated load for critical endpoints.

---

How to Run Locally

# Frontend
cd frontend
npm install
npm run dev

# Backend
cd backend
venv/Scripts/activate 
pip install -r requirements.txt
python manage.py runserver