Zusammenfassung
Smartwatches are small but powerful devices which make daily life easier and are without a doubt desirable objects for thieves. In this paper, we present a first machine learning based theft detection approach running in a user's domain, relying solely on data of his smartwatch and thus not violating privacy. Hence, we collect data from multiple persons to first prove that there is an exploitable ...
Zusammenfassung
Smartwatches are small but powerful devices which make daily life easier and are without a doubt desirable objects for thieves. In this paper, we present a first machine learning based theft detection approach running in a user's domain, relying solely on data of his smartwatch and thus not violating privacy. Hence, we collect data from multiple persons to first prove that there is an exploitable structure within data provided by a smartwatch's inertial sensors and perform user identification on the basis of that data. Then we will present and thoroughly evaluate our robust, efficient and fast (within seconds) theft detection algorithm which has both a low false rejection rate and an even lower false acceptance rate.
Altmetric