Smartphones have become an integral component of our everyday lives. Activity record of our entire lives is present on our smartphones due to the advent increase in the use of social networking applications. Security of smartphones has become more necessary than ever.
Authentication is the process of recognizing a user’s identity. Authentication in a smartphone is performed with the use of a security PIN or facial/thumbprint recognition. In this research work, an additional layer of user authentication is implemented with the help of onboard device sensors such as accelerometers. The goal of the research was to develop a secondary authentication mechanism that is “continuous” and “implicit.”
Each user holds and uses their smartphones in a particular way. This research aims to model the way the user holds their smartphone and authenticate the user using that information. With the help of onboard direction sensors such as the accelerometer – which measures the coarse-grained motion of the user, and proximity sensor – which measures the distance of the user from the smartphone, the behavior of the user is modeled and classified using a Support Vector Machine (SVM) classifier.
The behavior of the user is continuously monitored, and a One-Class SVM classifier in which a hyperplane separates the fed data into two planes.
The above model was prototyped on a Raspberry Pi board. Using the data from the accelerometer, the accuracy of the model was around 85%, with a 200ms sampling time. The accuracy was further increased to around 90% with the help of data from a combination of sensors like an accelerometer and gyroscope. The results of the work were presented in the Deep Learning Cybersecurity Research Forum organized by the School of Computing, National University of Singapore, and sponsored by industry giants like PayPal and Blackhat Asia.
A travel grant of SGD 1200 was granted by the National University of Singapore for speaking at the research forum.