Touch events and human activities for continuous authentication via smartphone

Area of Science:

• Computer Science
• Cybersecurity
• Human-Computer Interaction

Background:

• Modern smartphone security relies on continuous authentication methods.
• Touch events and human activities provide silent, rich data for machine learning.
• Existing methods require user interaction, impacting user experience.

Purpose of the Study:

• To develop a continuous authentication method for smartphones.
• To authenticate users while they are sitting and scrolling documents.
• To evaluate the effectiveness of machine learning models using sensor data.

Main Methods:

• Utilized touch events and smartphone sensor features from the H-MOG Dataset.
• Incorporated the Signal Vector Magnitude feature for each sensor.
• Evaluated several machine learning models using 1-class and 2-class experiment setups.

Main Results:

• The 1-class Support Vector Machine (SVM) model achieved 98.9% accuracy.
• The 1-class SVM model achieved an F1-score of 99.4%.
• The Signal Vector Magnitude feature was identified as highly significant.

Conclusions:

• Continuous authentication using touch events and sensor data is highly effective.
• Machine learning models, particularly 1-class SVM, offer robust smartphone security.
• The proposed method provides a silent and efficient authentication solution.