.png){kind=icon}
.png){kind=icon}
Return
AnGLEAuth: Sensor-Based Continuous Authentication via Adaptive Sample Generation and Global–Local Feature Encoding
.png){kind=icon}
PRE
AI
DOI：10.1109/JIOT.2026.3656890.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
With the rapid advancement of communication technologies, mobile devices have become indispensable in our daily life. As dependence on mobile and internet of things (IOT) technologies continues to grow, ensuring the robust protection of sensitive data becomes increasingly critical. Traditional authentication methods, such as personal identification numbers (PINs) and biometrics, remain vulnerable to adversarial attacks and are typically limited to one-time verification. To overcome these limitations, we propose AnGLEAuth, a sensor-based continuous Authentication system that leverages Adaptive sample construction and Global–Local feature Encoding for effective feature extraction using data from the accelerometer, gyroscope, and magnetometer sensors of mobile devices. Specifically, we design an adaptive fusion-enhanced sample construction algorithm that improves generalization capability and noise tolerance by integrating multiple data augmentation strategies, thereby enhancing overall accuracy and robustness. Moreover, AnGLEAuth incorporates both a global feature encoder and a local feature encoder to simultaneously capture long-term dependency patterns and fine-grained behavioral characteristics, enabling adaptive fusion of authentication policies and multidimensional user features. Extensive experiments on our dataset demonstrate the superiority of AnGLEAuth, achieving an average accuracy of 99.06% and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F1$ </tex-math></inline-formula>-score of 99.07% across ten unseen users.
Keywords:
Adaptive sample generation
continuous authentication
contrastive learning
global–local feature encoding
Journal
IF:
8.9
Papers:
1.4W
Citations:
7.8W