Phishing Detector - BERT + LSTM Deep Learning UI

Production-ready phishing email detector with BERT Transformer + Bidirectional LSTM deep learning models. Features live UI with model agreement, confidence comparison, final verdict, risk indicator, and processing time.

🎯 Live Demo Features

🤖 BERT Prediction 
🧠 LSTM Prediction 
📊 Model Agreement Status (✅ Agree / ⚠️ Disagree)
📈 Average Confidence Score
🎯 Final Combined Verdict
🟢 Risk Level Indicator

🚀 Quick Start (Local)

1. Clone & Install

git clone https://github.com/kasmya/phising-detector-dl.git
cd phising-detector-dl
pip install -r requirements.txt

2. Run Server

python app.py

Open: http://localhost:5001

3. Test Analysis

Legitimate Invoice:

Invoice #1234
Dear Customer,
Your invoice #1234 is ready. Download: invoice1234.pdf
Questions? support@company.com
Thanks, Billing Dept

Expected: LEGITIMATE ✅ | Low Risk | Models Agree

Phishing Alert:

URGENT: Your account is LOCKED! Verify immediately:
http://fake-bank-security.com/verify-account
24hrs or PERMANENT loss!

Expected: PHISHING 🔴 | High Risk | Models Agree

🤖 Model Details

Model Training

• BERT: Fine-tuned on phishing emails (HuggingFace)
• LSTM: Custom bidirectional RNN with class weights
• Vocab: 8K tokens (lstm_vocab_fixed.json)
• Max Len: 80 tokens

📁 Project Structure

phising-detector-dl/
├── app.py                 # Flask backend + models
├── templates/index.html   # Responsive UI
├── requirements.txt       # Dependencies
├── lstm_phishing_fixed.pth # LSTM model (91% acc)
├── lstm_vocab_fixed.json # LSTM vocab
├── phishing_model/        # BERT model files
├── README.md             # This file