Configuration-Based Fingerprinting of Mobile Device Using Incremental Clustering

Device fingerprinting has lately received great attention due to its effectiveness in fraud detection, secure authentication, and user tracking. Whereas fingerprinting performs well on labeled device data using classification methods, there are several researches concentrated on unlabeled mobile device data and existing methods often lack in precision and efficiency. To overcome this challenge, we focus on the use of the mobile device’s configuration-related characteristics as a mean to build a device fingerprint, which allows to distinctively and reliably characterize each device. In addition, an incremental clustering approach is proposed to classify unlabeled device data into clusters on the basis of their similarity. Moreover, we customize individual distance threshold for each user according to their device configurations’ modifying frequency, in order to construct a precise authentication mechanism between users and devices. The proposed clustering model and device fingerprinting system are evaluated on 8220 fingerprints from 815 different devices. The experimental results demonstrate the effectiveness and efficiency of our algorithms.