Who Used My Smart Object? A Flexible Approach for the Recognition of Users

This paper deals with the authentication of the user of a connected object. We propose a flexible and nonintrusive method based on the use of two categories of everyday connected objects (i.e., smart watch and remote control). Data were collected during participants' interactions with a smart TV. The discrete cosine transform algorithm was used to extract the most informative features. Based on these features, four classification algorithms (deep neural network, support vector machine, Naïve Bayes classifier, and C45) were applied to the data in order to detect the user's identity. The classification was performed based on the recognition of four types of human activities (sitting, standing, walking, and lying down) through building four databases. Following this, a second classification was made for each data set activity type in order to identify the users. The results show that it is possible to discriminate between users according to their activities. The accuracy of recognition reached 91% for some participants within a certain activity configuration.