R²BN: An Adaptive Model for Keystroke-Dynamics-Based Educational Level Classification.

Area of Science:

• Behavioral biometrics
• Machine learning
• Human-computer interaction

Background:

• Keystroke pattern recognition is a growing forensic tool.
• Existing machine learning models have limitations in profiling user characteristics beyond gender and age.
• Educational level is an under-researched but valuable user characteristic for digital forensics and targeted advertising.

Purpose of the Study:

• To propose a novel machine learning model for predicting educational level based on keystroke dynamics.
• To evaluate the model's performance against existing methods.
• To investigate methods for reducing computational cost and balancing prediction accuracy with speed.

Main Methods:

• Development of a randomized radial basis function network model.
• Collection of empirical keystroke data from volunteers during daily computer use.
• Comparison of the proposed model with other machine learning models using keystroke dynamic datasets.

Main Results:

• The proposed randomized radial basis function network model achieves high accuracy in predicting educational level from keystroke dynamics.
• The model's performance was validated against established machine learning techniques.
• High computational cost was identified as a limitation, prompting further investigation into optimization.

Conclusions:

• This research presents the first model to predict educational level solely from keystroke dynamics.
• The model demonstrates significant potential for applications in digital forensics and personalized advertising.
• Further research is needed to optimize the model for practical, efficient deployment, addressing the accuracy-speed tradeoff.