A Novel GMM-Based Behavioral Modeling Approach for Smartwatch-Based Driver Authentication.

Area of Science:

• Human-Computer Interaction
• Machine Learning
• Biometric Authentication

Background:

• Driver behavior varies significantly based on individual habits and driving scenarios.
• Existing driver authentication methods may not fully capture nuanced behavioral patterns.
• Smartwatch sensors offer a convenient platform for continuous, unobtrusive data collection.

Purpose of the Study:

• To propose an improved Gaussian mixture model (GMM) for more accurate driving behavior modeling.
• To develop a novel driver authentication system utilizing smartwatch accelerometer and orientation data.
• To evaluate the performance of the proposed method in diverse driving environments.

Main Methods:

• A novel Gaussian mixture model (GMM) was developed, enhancing traditional GMM capabilities for behavior modeling.
• An experimental system was created to collect and analyze driving behavior data from smartwatch sensors.
• Driver authentication performance was assessed using the equal error rate (EER) metric.

Main Results:

• The proposed GMM-based method demonstrated improved driving behavior modeling compared to traditional GMM.
• The driver authentication system achieved an equal error rate (EER) of 4.62% in simulated environments.
• The system achieved an EER of 7.86% in real-traffic environments, confirming practical feasibility.

Conclusions:

• The novel GMM-based approach effectively models individual driving behaviors for authentication.
• Smartwatch sensor data, analyzed with the proposed method, provides a feasible basis for driver authentication.
• The system shows potential for enhancing vehicle security and personalized driving experiences.