Anomaly Detection for Mobile Device Comfort

In this work, we utilize the data that can be collected via some of the sensors found on modern smartphones to find anomalies in the behaviour of a smartphone user and the current context. This can be done to prevent the mobile phone from being compromised physically or to warn the user when he/she is behaving "unusually."We apply the "time slice" notion to existing anomaly detection methods, evaluate our approach on two published data sets, and confirm that it is feasible to use our approach on smartphones with modest hardware.Our work is part of Marsh et al.'s Device Comfort paradigm, which is an application of computational trust to mobile device security, where the mobile device is able to have a varying level of "trust" in its user based on behavioural biometrics, and the contexts in which the user interacts with the mobile device.