🔒 Real-Time Risk Scoring System Based on Login Data (Deep Learning)

---

🚀 Project Overview

This project aims to develop a deep learning-based system that calculates a real-time risk score for user logins, based on behavioral and contextual login data. The model compares each login attempt with the user's historical patterns and outputs a risk score between 0 and 100.

---

📁 Project Structure

risk_skorlama_projesi/
│
├── data/                # Mock/generated datasets
├── src/                 # Source code modules
│   ├── data_generation.py   # Mock data generation
│   ├── labeling.py          # Risk labeling & scoring
│   ├── model.py             # Model training & evaluation
├── requirements.txt     # Python dependencies
├── README.md            # Project documentation
├── main.py              # Main pipeline script
└── app.py               # Streamlit web app

---

✨ Features

• Realistic Mock Data Generation: Simulates user login behavior with configurable randomness.
• Flexible Risk Labeling: Rule-based and statistical risk scoring for each login event.
• Deep Learning Model: Predicts risk scores using a neural network (TensorFlow/Keras).
• Interactive Web UI: Modern Streamlit interface for real-time scoring and data exploration.
• Modular & Reusable: Clean, well-documented, and easy to extend.

---

🖥️ Quick Start

1. Install dependencies:

   pip install -r requirements.txt

2. Run the main pipeline:

   python main.py

3. Launch the web app:

   streamlit run app.py

---

📊 Screenshots

---

⚙️ How It Works

• Data Generation: Produces mock login data with realistic user patterns and anomalies.
• Risk Labeling: Assigns a risk score (0-100) to each login based on time, IP, MFA, device, and behavioral deviations.
• Model Training: Trains a deep learning model to predict risk scores from login features.
• Web Interface: Allows users to simulate logins, get instant risk scores, and explore the dataset interactively.

---

📚 Technologies Used

• Python, Pandas, NumPy
• TensorFlow & Keras
• Scikit-learn
• Streamlit
• Faker (for data generation)

---

🙋‍♂️ Author & Contact

• Developed by Yusuf Uzun
• yusfuzn@hotmail.com

Feel free to contribute, open issues, or suggest improvements!