🛡️ PhishShield – Advanced Phishing Detection System

PhishShield is a powerful AI-driven phishing detection agent that analyzes emails and web content to identify potential phishing threats in real-time. Using an intelligently trained Random Forest classifier on a dataset of 200k+ emails, PhishShield leverages linguistic patterns, link analysis, and structural metadata to predict whether a message is phishing or legitimate with over 96% accuracy.

📌 Overview

PhishShield consists of

• 🧠 Backend (Flask API): ML-powered REST API for real-time predictions.
• 🎨 Frontend (HTML/CSS/JS): Interactive UI for email inspection and manual inputs.
• 📊 ML Model: Pre-trained model using advanced feature engineering.
• 📁 Dataset: 200,000 labeled email samples with 7 rich features for training and research.

Model Working

🌟 Key Features

🚀 Real-time Detection

• REST API via Flask backend
• JSON-based predictions with confidence scores
• Auto model training fallback (train.py)

🧪 Smart Feature Engineering

• 25+ extracted features including:
  • Domain reputation
  • Link patterns
  • Urgent language indicators
  • HTML tag frequency
  • Attachment behavior

🧠 ML Model Highlights

• Random Forest Classifier
• Over 96% accuracy
• Trained on 200k labeled samples
• Explainable predictions

💡 User-Friendly Interface

• Drag-and-drop email file analysis (.eml, .msg)
• Manual content entry support
• Visual risk indicator (safe/suspicious)
• Historical detection log
• Responsive UI for all devices

---

PhishShield Demo Walkthrough

Main Dashboard

Features:

• Dual input methods (file upload + text input)
• Real-time analysis results
• Confidence visualization

Manual Analysis Mode

Workflow:

1. Paste email content
2. Optionally add metadata
3. Get instant phishing probability

File Upload Analysis

Supports:

• .eml and .msg email files
• Automatic header parsing
• Attachment detection

Analysis History

Key Functions:

• Persistent scan records
• Quick-result filtering
• One-click reanalysis

---

📊 Dataset

PhishShield includes a robust and reusable dataset of 200,000+ labeled emails for training, evaluation, and experimentation.

📑 Labels (Features):

Feature	Description
email_text	Body content of the email
subject	Email subject line
has_attachment	Binary flag (1 = yes, 0 = no)
links_count	Number of hyperlinks detected
sender_domain	Domain of sender’s email address
urgent_keywords	Binary flag (1 = urgent words found)
label	Target class: phishing or legitimate

🧠 Ideal for building and enhancing phishing classifiers or integrating into broader cybersecurity AI pipelines.

---

🛠️ Installation & Setup

📋 Prerequisites

• Python 3.8+
• Git
• Node.js (for frontend development, optional)

🔧 Backend Setup

# Clone the repository
git clone https://github.com/AtharIbrahim/Phishing-Email-Agent.git
cd Phishing-Email-Agent

# Create virtual environment
python -m venv venv

# Activate environment
# For Windows:
venv\Scripts\activate
# For macOS/Linux:
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Start Flask server
python app.py

📬 Contact

• Name: Athar Ibrahim Khalid
• GitHub: https://github.com/AtharIbrahim/
• LinkedIn: LinkedIn Profile
• Website: Athar Ibrahim Khalid

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.