Currently, the number of connected devices is raising in different areas, such as smart home, smart grid and health care. The emergence of these smart things leads to the generation of a large amount of data that is manipulated by several parties. This data is usually very rich in information and includes metadata such as location, time and context which provides personal habits, user-specific behaviors and individual preferences. Therefore, it is clear that the data generated by such devices must be carefully protected by all the involved parties during the collection, transfer, storage and analysis phases. To this end, our work aims at (i) defining a semantic layer that represents the metadata as well as the data requirements considering privacy norms and laws, (ii) proposing an approach that, based on the semantic layer, protects privacy from an end to end view during the data life cycle, and (iii) providing a recommendation process that helps to exploit the collected data. Two application domains are addressed in this thesis, namely smart grid and health care.