PhishGuard AI

Overview

PhishGuard AI is an open-source, AI-powered phishing detection system hosted on Hugging Face with a lightweight browser extension for real-time protection. It uses cloud-based AI analysis to detect phishing attempts in emails, URLs, and websites before harm occurs.

Novelty

• 56-Feature Analysis – Evaluates URLs using 56 parameters like domain age, entropy, URL length, SSL validity, suspicious keywords, and more.
• AI-Enhanced Threat Detection – Goes beyond blacklists with ML-based risk analysis.
• Cloud-Powered Processing – Performs computations on secure cloud servers for fast detection.
• Seamless Browser Integration – Lightweight extension with minimal performance impact.
• Adaptive Learning – Continuously improves with new threat data.

How It Works

1. The browser extension captures the URL or web content.
2. 56 features are extracted for risk assessment.
3. Data is sent securely to the Hugging Face-hosted AI model.
4. The model predicts phishing probability and returns a response.
5. The extension blocks or warns users accordingly.

56 Features Used for Detection

The model uses structural, lexical, and security-based features such as:

• URL length, domain age, entropy of URL
• Presence of IP address in URL
• HTTPS/SSL certificate validity
• Number of dots, hyphens, suspicious patterns
• WHOIS registration info (age, validity)
• Presence of phishing keywords (e.g., "login", "verify")
• Redirect count and presence of shortened URLs
• And 48+ additional attributes for fine-grained detection

Tech Stack

• Frontend (Extension): JavaScript, HTML, CSS
• Backend (Cloud): Python, FastAPI/Flask
• AI Models: PyTorch / TensorFlow hosted on Hugging Face Spaces

Deployment on Hugging Face

The AI model is deployed on Hugging Face Spaces for real-time inference.

API Example

import requests

API_URL = "https://huggingface.co/spaces/ascend-x/phishing_detector"
data = {"url": "http://suspicious-site.com"}

response = requests.post(API_URL, json=data)
print(response.json())