SmartGuard: AI-Powered SMS Threat Intelligence

"A security tool, not just a classifier."

SmartGuard is a next-generation SMS firewall that uses Deep Learning to intercept and neutralize phishing attacks in real-time.

Live Demo

Launch SmartGuard Secure Chat (Works on Mobile & Desktop)

---

Project Overview

Traditional spam filters are passive—they sort mail after it arrives. SmartGuard acts as an Active Sentinel Agent. It sits between the incoming data stream and the user interface, analyzing message tokens in under 100ms.

If a threat is detected (e.g., "URGENT," "WINNER," "FREE"), the agent automatically obfuscates the content before it renders, protecting the user from psychological triggers used in social engineering attacks.

Key Features

• Active Interception: Simulates a middleware layer that filters traffic before display.
• Blur-to-Protect UI: Spam messages are visually blocked with a "Show Anyway" override, prioritizing user safety.
• Real-Time Analysis: Powered by a TensorFlow/Keras model achieving 98.5% accuracy on the UCI SMS Spam Collection.
• Modern Aesthetic: Built with a "Cybersecurity Terminal" design language using Tailwind CSS and Lucide Icons.

---

Interface

The Threat Scanner

A terminal-like interface for manual message analysis.

The Active Agent Chat

A live simulation of a secure messaging app.

---

Technical Architecture

The Brain (Machine Learning)

• Framework: TensorFlow / Keras
• Architecture: Feed-Forward Neural Network (Sequential)
• Layers:
  1. Embedding: Maps 1000+ vocabulary tokens to dense vectors.
  2. Global Average Pooling: Condenses vector sequences into context-aware summaries.
  3. Dense (ReLU): Extracts non-linear features.
  4. Dropout (0.2): Prevents overfitting to specific keywords.
  5. Sigmoid Output: Returns a probability score (0.0 - 1.0).

The Body (Frontend)

• Framework: React (Vite)
• Styling: Tailwind CSS (Dark Mode, Glassmorphism)
• State Management: React Hooks (useState, useEffect) for simulating async API latency.

---

Local Installation

Want to run the "Active Agent" locally?

1. Clone the repository

   git clone https://github.com/natinew77-creator/SMS-Spam-Classifier-NLP.git
   cd SMS-Spam-Classifier-NLP

2. Install Client Dependencies

   cd client
   npm install

3. Run the Development Server

   npm run dev

   Open http://localhost:5173 in your browser.

---

Author

Natneal B.

• LinkedIn

---