Wearable devices are currently being considered to collect personalized physiological information, which is lately being used to provide healthcare services to individuals. One application is detecting depression by utilization of motor activity signals collected by the ActiGraph wearable wristbands. However, to develop an accurate classification model, we require to use a sufficient volume of data from several subjects, taking the sensitivity of such data into account. Therefore, in this paper, we present an approach to extract classification models for predicting depression based on a new augmentation technique for motor activity data in a privacy-preserving fashion. We evaluate our approach against the state-of-the-art techniques and demonstrate its performance based on the mental health datasets associated with the Norwegian INTROducing Mental health through Adaptive Technology (INTROMAT) Project.