Research And Implementation About System Of Fingerprinting Mobile Device Based On Motion Sensor

With the advent of the mobile internet,various industries based on mobile devices have flourished.However,the rapid development of underground industry has led to huge losses in all industries.Therefore,enterprises urgently need a new device identification technology to distinguish malicious users and even fraud groups,then technology of fingerprinting devices emerges as the times require.The research on technology of fingerprinting devices has been developing for a long time.Based on protecting user privacy and without users' permission,using motion sensor data as device fingerprint to identify devices is more applicable to the world scenario.However,in the world scenario,there are still two essential problems to be solved.The first problem is how to correctly identify a previously unknown device,and the second is how to update the model with a small cost after device identification.In order to solve these problems,we propose a system of fingerprinting mobile devices based on motion sensor.For this system,each device corresponds to a binary classification model based on convolutional neural network.When a piece of unknown device data comes,each device model in the system will predict it and get the predicted probability.If there are some predicted probabilities greater than threshold,the device data belongs to the device corresponding to the model with the largest predicted probability value.Otherwise,if all predicted probabilities are less than threshold,the device data belongs to a new device.And a new binary classification model of the new device will be trained based on the new device data,which means updating of system.As shown in experiments,for a certain device,the accuracy of identifying a new device can reach 87%,and all accuracies of identifying old devices are more than 90%,which proves the effectiveness of system on identifying new and old devices.For all devices,the overall identification accuracy can reach 75%.And no matter how many mobile devices are,the time of updating system is about 74s,which proves the performance of our system is great.