AccountGuard AI

AccountGuard AI is a specialized security tool designed for cybersecurity analysts working in financial institutions. It enables analysts to detect suspicious activity by analyzing transaction data and phishing emails—all from a single interface.

---

🔍 Key Features

1. Email Phishing Detection

• Upload any suspicious email.

• Analyzes content using HuggingFace models:

  • cybersectony/phishing-email-detection-distilbert_v2.4.1
  • CrabInHoney/urlbert-tiny-v4-phishing-classifier

• Use case: Analysts can quickly check user-reported emails that may be part of an attack campaign.

2. Transaction Fraud Detection

• Upload financial transaction data.

• Identifies anomalous transactions based on patterns in:

  • Amount, Product Code, Card details, Address, Distance, Device Info, etc.

• Enables analysts to correlate suspicious emails with transaction activity.

---

🌟 What Makes This Unique

We trained our own machine learning model on the IEEE-CIS Fraud Detection dataset, making this one of the first real-time tools to use that dataset for live fraud detection.

Unlike off-the-shelf solutions, this project features:

• A custom IsolationForest model tuned for practical anomaly detection
• Data preprocessing and transformation pipelines
• Integration with a real-time API via FastAPI

No existing model from HuggingFace or other public repositories was trained on this type of transactional data for fraud detection. This is a ground-up, custom-trained model tailored for financial institutions.

---

⚙️ Tech Stack

• Frontend: Next.js, Tailwind CSS
• Backend: FastAPI, Python
• ML Models: Custom-trained IsolationForest (Scikit-learn), HuggingFace Transformers for phishing detection

---

📂 Project Structure

AccountGuard-AI/
├── frontend/        # Next.js + Tailwind frontend
├── backend/         # FastAPI backend with custom ML models
├── .gitignore
└── README.md

---

🚚 Setup Guide

1. Clone the repository

git clone https://github.com/sameerwire/AccountGuard-AI.git
cd AccountGuard-AI

2. Run Frontend

cd frontend
npm install
npm run dev

3. Run Backend

cd ../backend
python -m venv venv
venv\Scripts\activate  # On Windows
pip install -r requirements.txt
uvicorn main:app --reload

---

🔢 Example Usage

Email:

Subject: Urgent Account Verification
Body: Please confirm your credentials at http://fakebank.com

Transaction:

{
  "amount": 120.50,
  "product_cd": "w",
  "card1": 11109,
  "card2": 404.0,
  "card3": 150.0,
  "card4": "visa",
  "card5": 226.0,
  "addr1": 330.0,
  "dist1": 10.0,
  "device_type": "desktop",
  "device_info": "windows"
}

---

📤 .gitignore Notes

• node_modules/, .next/, venv/, .env, *.log, and large dataset/model files are excluded
• Model training datasets from IEEE-CIS are not pushed to GitHub

---

📖 License

This project is open source under the MIT License. Feel free to fork, modify, and share!