On the Use of Mobile Calling Patterns for Soft Biometric Classification

摘要

In this paper, we present an analysis of soft biometric classification and user identification using six months of calling logs from 46 smartphone users. We introduce a ranking methodology to quantify the discriminating strength of thirteen soft biometric classes. These ranked classes are then used to reduce the search space during the biometric identification task for potentially improving recognition accuracy. However, findings indicate that the dynamic nature of calling habits significantly complicates user identification, particularly when soft biometric information is incorporated.