Continuous Authentication in Resource-Constrained Devices via Biometric and Environmental Fusion

Area of Science:

• Computer Science
• Cybersecurity
• Biometrics

Background:

• Continuous authentication enhances device security beyond single logins.
• Existing methods are vulnerable to spoofing, data breaches, and environmental factors.
• Environmental conditions significantly impact user online security and susceptibility to attacks like impersonation and replay.

Purpose of the Study:

• To develop a lightweight and secure continuous authentication system.
• To integrate face recognition with environmental sensing for adaptive security.
• To improve energy efficiency and reduce authentication time.

Main Methods:

• A convolutional neural network converts faces into 128-bit codes, combined with a nonce and hashed.
• A camera-microphone module monitors ambient light and sound to trigger re-authentication.
• The system was verified using formal security tools (Scyther v1.1.3) against various attacks.

Main Results:

• The system demonstrated resistance to replay, interception, deepfake, and impersonation attacks.
• Median decision time decreased from 61.2 ± 3.4 ms to 42.3 ± 2.1 ms.
• Data usage per cycle dropped by 24.7% ± 1.8%, and mean energy consumption decreased from 21.3 mJ to 19.8 mJ.

Conclusions:

• Smart-sensor-triggered face recognition provides secure and energy-efficient continuous verification.
• The system adapts to environmental changes, enhancing user online security.
• This approach supports advancements in smart imaging and deep-learning-based biometrics.