Secure Keystroke Authentication

Secure Keystroke Authentication is a biometric-based authentication system that leverages keystroke dynamics to enhance security. By analyzing typing patterns, such as key press durations and transition times, this project aims to provide an additional layer of authentication beyond traditional password-based methods.

Features

• Collects keystroke timing data to create a unique user profile.
• Trains a machine learning model (Support Vector Machine) for user authentication.
• Detects unauthorized users based on typing behavior.
• Lightweight and does not require additional hardware.

Technologies Used

• Python
• NumPy, Pandas
• Scikit-learn – logistic regression model
• TensorFlow / Keras – simple feedforward neural network

How It Works

1. Keystroke Timing Feature Extraction

• Each typing session records:
  • Hold time (key down → key up)
  • Latency (delay between keys)
  • Flight time (key up → next key down)

2. Data Preprocessing

• Clean input data
• Normalize timing features
• Balance dataset if needed

3. Model Training

• Logistic Regression (scikit-learn)
• Neural Network (TensorFlow)
• Compare performance on accuracy & generalization

4. Prediction

• Real-time or simulated user input is passed into the model
• Model returns a binary classification: “user” or “intruder”

Project Structure

├── collected_data.py       # Script to collect keystroke data
├── train_model.py          # Script to train SVM model for authentication
├── keystroke_auth.py       # Main authentication script
├── requirements.txt        # Required dependencies
├── README.md               # Project documentation
└── research_paper.pdf      # Research paper detailing the project

Installation

Clone the Repository

 git clone https://github.com/miyoku1226/secure-keystroke-auth.git
 cd secure-keystroke-auth

Install Dependencies

pip install -r requirements.txt

Usage

Collect Keystroke Data

Run the script to record typing behavior for different users:

python collected_data.py

Train the Authentication Model

After collecting sufficient data, train the SVM model:

python train_model.py

Authenticate a User

Once the model is trained, run the authentication process:

python keystroke_auth.py

Example Results

Model	Accuracy	False Accept Rate (FAR)	False Reject Rate (FRR)
SVM	85%	5%	10%
KNN	81%	7%	12%

Future Improvements

• Improve accuracy by testing with a larger dataset.
• Implement additional classification models (e.g., neural networks).
• Develop a GUI for user-friendly interaction.
• Test the system on different keyboard layouts and devices.