🛡️ AutoPhish — Intelligent Email Threat Classifier

AutoPhish is an AI-powered email classification prototype that detects phishing attempts using an LLM backend. It's designed to be seamlessly integrated into enterprise email systems and security workflows.

📷 Demo

🚀 Features

• 📧 Paste Email Content: Simulate an incoming email.
• 🤖 LLM-based Detection: Uses a large language model to classify emails as "Phishing" or "Not Phishing".
• ⚡ Instant Action:
  • ✅ If not phishing: Archive + auto-response simulated.
  • ⚠️ If phishing: Simulated report to IT Security.

🧠 How It Works

• Built with Streamlit for rapid prototyping.
• Email content is passed to an LLM running on an OpenAI Comptible Endpoint
• The model returns a one-word decision: "Phishing" or "Not Phishing".

🔧 Tech Stack

• Streamlit
• Ollama or LM Studio
• Python, requests

📦 Local Setup

git clone https://github.com/yourusername/autophish.git
cd autophish
streamlit run app.py

🧭 Future Vision

• 🧩 Outlook Plugin: Direct integration with enterprise Outlook environments

• 🛠️ Tool-Calling LLM: To trigger actions like auto-reply, quarantine, IT report

• 🧠 MCP server integration: The LLM can directly have access to the emails and tools for further actions

• 🗂️ Archival & Search: Automatic archiving and tagging of safe conversations

• 🌃 Multimodal support : Enhancing understanding by accepting image/attachment inputs too

• 🌐 SOC Integration: Plug into existing security orchestration systems (SOAR)