Novel Efficient and Privacy-Preserving Protocols for Sensor-Based Human Activity Recognition

Human activity recognition (HAR) has become an important emerging field of application for sensor networks (SN) technologies. Nevertheless, the pervasiveness of SN in everyday life has given rise to new privacy concerns especially when mining personal sensed data in external environments. From that perspective, many research works have proposed cryptography-based techniques so as to tackle SN privacy issues, yet have costed significant degradations in computational-time efficiency. In this work, we propose a novel privacy-preserving Knn classification protocol to be used in HAR process, that is based on a novel privacy-preserving protocol that aims to assess similarity between personal recorded activities, external patterns using the cosine similarity metric. We build our proposals without any cryptographic schemes in order to provide a high efficient recognition service.