💸 FraudSniff – Smart Banking Fraud Detection System

FraudSniff is a secure, interactive banking system that performs real-time fraud detection using a trained machine learning model.
It includes OTP-based transaction verification and stores all user data in an SQLite database.
A lightweight Streamlit-based Banking Simulation App that enables users to deposit, withdraw, and transfer funds while detecting fraudulent transactions using a trained Machine Learning model.

---

🔐 Features

• 🧠 Machine Learning Fraud Detection: Predicts fraudulent transactions using Random Forest Classifier.
• 🔄 OTP Email Verification: Secure transfers with 6-digit OTP emailed to the user.
• 🧾 Transaction History: View past transfers and deposits.
• 👥 Multi-user Support: Login and manage accounts securely.
• 💾 SQLite Integration: Persistent storage of user data and transactions.

---

🏗️ Tech Stack

Component	Technology
Frontend	Streamlit
Backend	Python (Flask-style logic)
ML Model	RandomForest (scikit-learn)
Database	SQLite
Email Service	SMTP via Gmail
Deployment Ready	✅ Yes

---

📁 Folder Structure

banking\_system/
├── app.py                # Main Streamlit app
├── db.py                 # DB helper functions
├── train\_model.py        # ML model training script
├── utils.py              # Fraud detection, OTP, etc.
├── transactions.csv      # Sample dataset for training
├── model/
│   ├── fraud\_model.pkl   # Trained ML model
│   └── scaler.pkl        # Feature scaler
└── bank.db               # SQLite database

---

⚙️ Setup Instructions

1️⃣ Clone the Repo

git clone https://github.com/yourusername/fraudsniff-banking-system.git
cd fraudsniff-banking-system

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Configure Email for OTP

In utils.py, update the following lines with your Gmail and App Password:

sender_email = "[email protected]"
password = "your_app_password"

⚠️ Important: If you have 2FA enabled, generate an App Password.

4️⃣ Run the App

streamlit run app.py

---

🧪 Training Your Own Model (Optional)

To retrain the fraud detection model with your own data:

python train_model.py

Make sure your transactions.csv includes:

• amount
• timestamp
• avg_daily_tx (optional, can be derived)
• is_fraud (0 or 1)

---

🔍 Sample Fraud Detection Logic

The model uses:

• Transaction amount
• User's avg_daily_tx over past few days
• hour of transaction

Prediction is made using:

model.predict(scaler.transform([[amount, avg_daily_tx, hour]]))

---

📌 TODO / Future Work

• ✅ Add transaction reversal on fraud detection
• 🔒 Password hashing for better security
• 📈 Real-time dashboard with fraud stats
• 📊 Improve ML accuracy using XGBoost or SMOTE

---