Accelerometer-Based Mobile Device Identification System for the Realistic Environment

Abstract

With the rapid development of the Internet and mobile technology, the frequency of online fraud incidents continues to increase, and the fraudulent means are increasingly diverse. Traditional antifraud methods can no longer effectively deal with the serious security situation at present because of their own various shortcomings. In this paper, a device identification system based on acceleration sensor is constructed which takes into consideration the imperfections of the sensor in the manufacturing process and assembley process. The system can identify the user's device by taking advantages of the data collected in a flat static state, and thus confirm the user's identity. The acceleration data of the device is first collected, and then the data in flat static state is extracted through the state discrimination algorithm. After the feature extraction process, a 20-dimensional feature vector is obtained. Finally, a device identification model combining one-class classifier and multi-class classifier is trained. The model can identify unknown devices and classify known devices. The method proposed in this paper solves the problem that the device state cannot be determined in the real world. Further on we propose a multi-layer device identification model that can identify unknown device without training threshold. Compared with the previous studies in the literature, the biggest innovation of our method is to consider the user authentication of device fingerprint in the real world.

Smartphone Camera Identification from Low-Mid Frequency DCT Coefficients of Dark Images

“…In addition to this, the recent literature has considered smartphone identification using distinct sensors, e.g., microphone [ 59 , 60 ], loudspeaker [ 61 , 62 , 63 ], gyroscope [ 64 ], battery consumption [ 65 ], accelerometers [ 66 , 67 ], etc.…”

Fingerprinting Smartphones Based on Microphone Characteristics From Environment Affected Recordings

“…Mobile devices identification based on 20 features in time and frequency domain, extracted from accelerometer data is proposed in [34]. A more rarely employed sensor for fingerprinting is the magnetometer.…”

Cross Layer Attacks and How to Use Them (for DNS Cache Poisoning, Device Tracking and More)

“…A recent survey of sensor-based tracking device identification techniques is presented in [20,Table 1]. As can be seen, none of the techniques surveyed is both practical and performant.…”

“…As can be seen, none of the techniques surveyed is both practical and performant. The new technique introduced in that paper also yields results [20,Table 5] which make it impractical when the device is not in a "flat static case".…”