Performance Evaluation of Mobile Sensor for Context Awareness User Authentication

Abstract

With the increase of smart devices and their capacities, their use for different services have also increased. As much as this is an advantage, it has posed additional risks because of the confidential information stored on them. This has increased the need for additional security on these systems. Most of the methods used for user authentication pose certain drawbacks that are either easy to circumvent or cumbersome to use. As a result, multi-level means of authentication is needed to improve the security of mobile devices. Sensors are playing a vital role in the mobile ecosystem to enhance different services. These sensors can be leveraged upon as a solution for user authentication. This research analyzed and evaluated different mobile device sensors for continuous and transparent user authentication. The mobile data used includes gyroscope, accelerometer, linear accelerometer, proximity, gravity, and magnetometer sensors’ data. Using a Feedforward Neural network for data classification after extracting features from the different sensors available in the mobile device; the most effective was selected by evaluating performance of the different sensors. The best sensor, the accelerometer was further experimented on. The experiment showed that smartphone accelerometer sensor exhibits sufficient discriminability, stability, and reliability for active and continuous authentication, by achieving a performance of 6.55% for the best overall EER.