S3: An AI-Enabled User Continuous Authentication for Smartphones Based on Sensors, Statistics and Speaker Information

Area of Science:

• Computer Science
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Continuous authentication aims for non-intrusive smartphone user verification.
• Existing systems struggle with data requirements, training time, and false alerts.
• The role of voice in continuous authentication requires deeper analysis.

Purpose of the Study:

• To develop and evaluate the S3 platform, an AI-enabled system for continuous smartphone authentication.
• To assess the relevance of voice, sensor, and application data for user authentication.
• To determine optimal data combination strategies and necessary training durations.

Main Methods:

• Developed the S3 platform integrating AI with sensor, application statistics, and voice data.
• Conducted experiments to evaluate individual data source relevance and combination strategies.
• Assessed the impact of training duration on user profile accuracy.

Main Results:

• Voice data demonstrated significantly higher relevance for authentication compared to sensors and application statistics.
• Combining individual data models proved to be the most effective strategy.
• The S3 platform achieved robust performance with just five days of user data for training.

Conclusions:

• Voice is a critical modality for effective continuous smartphone authentication.
• The S3 platform offers an efficient and accurate solution using multi-modal data fusion.
• A publicly available dataset was created to facilitate further research in continuous authentication.