ADAPTIVE WINDOW SAMPLING AND FILTERING FOR CONTINUOUS MOBILE BEHAVIORAL AUTHENTICATION

Date

2025-8-21

Author

Kaplan, Muratcan

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

4531
views

0
downloads

Cite This

Continuous behavioral authentication verifies user identity in real-time through patterns in natural device interactions, such as touch gestures and hand movements. An important challenge in this domain is identifying segments of sensor data that consistently reflect individual behavior while excluding periods of noise or inactivity. This thesis presents an end-to-end framework that uses accelerometer data temporally aligned with touchscreen events on mobile devices. The approach begins with a balanced sampling strategy that constructs triplets of same-user and different-user interaction windows across different sessions, enabling effective Siamese network training. A subsequent filtering stage evaluates the variability of each candidate window, discarding those with low motion or anomaly fluctuations. By combining session-aware sampling with proposed filtering scheme, the method constructs a dataset of sensor windows that consistently reflect distinctive user behaviors. The resulting framework improves the quality of training data for continuous authentication models and contributes toward more robust and unobtrusive security mechanisms for everyday mobile use.

Subject Keywords

Sensor-based Authentication, Behavioral Biometrics, Triplet Sampling Strategies, Siamese Neural Networks, Entropy Filtering

URI

https://hdl.handle.net/11511/115566

Collections

Graduate School of Informatics, Thesis