Anomaly detection for mobile device comfort

Abstract

As part of the Device Comfort paradigm, we envision a mobile device which, armed with the information made available by its sensors, is able to recognize whether it is being used by its owner or whether its owner is using the mobile device in an “unusual” manner. To this end, we conjecture that the use of a mobile device follows diurnal patterns and introduce a method for the detection of such anomalies in the use of a mobile device. We evaluate the accuracy of our method with two publicly available data sets and show its feasibility on two mobile devices.