CONTINUOUS AUTHENTICATION BASED ON SMARTPHONES APP ACCESS EVENT DATA

Abstract

Authentication in mobile phones is done by using knowledge-based authentication schemes (passwords or numeric codes known as personal identification numbers, PINs) or biometric authentication factors (face recognition or finger-print recognition) which are usually used to unlock the device or at the entry-point of the apps. After the initial authentication, it is assumed that the user is the same till the app is closed or the screen is unlocked, which makes smartphones vulnerable to unauthorized access. Therefore, there is a need for continuous authentication of the user at the application level after the initial entry-point authentication. One such motivation is, Smart homes are an emerging area of attraction and comfort for people where they can control their entire home on the tip of their fingers. So, security becomes an important factor so that any outsider or attacker does not login into the smart home server to steal or misuse user information. App access events are utilized to build user behavior patterns which can be used to identify the registered user of a particular device, requesting login to the smart home server. To enhance the model training many techniques have been used, along with that some existing and some additional features have been used. To evaluate the model performance, initially two metrics have been compared and for the final evaluation only the best suitable metric out of the two is used.