Secure and Usable User-in-a-Context Continuous Authentication in Smartphones Leveraging Non-Assisted Sensors.

Area of Science:

• Computer Science
• Cybersecurity
• Human-Computer Interaction

Background:

• Smartphones utilize sensors for environmental and usage data.
• Continuous user authentication is crucial for security against impersonation and robbery.
• Existing methods often rely on assisted sensors, limiting effectiveness and neglecting usability.

Purpose of the Study:

• To explore non-assisted smartphone sensors for continuous user authentication.
• To develop an approach with a tunable security-usability trade-off.
• To assess accuracy, immediacy, usability, and readiness for practical application.

Main Methods:

• Leveraging data stream mining techniques.
• Utilizing four non-assisted sensors: battery, transmitted data, ambient light, and noise.
• Evaluating performance across 50 users over 24 months.

Main Results:

• Battery sensor alone achieved 97.05% accuracy.
• Combined audio, light, and battery sensors reached 81.35% accuracy.
• Robbery detection within 100 seconds (battery) and 250 seconds (audio, light, battery) was achieved.

Conclusions:

• Non-assisted sensors offer a viable alternative for continuous authentication.
• The proposed method balances security and usability effectively.
• The approach demonstrates moderate training/storage needs, suitable for current devices.