Human beings are social animals and, therefore, their social interactions define themselves better than a username and a password. These social interactions can be monitored in an unattended way relying on the different kinds of devices carried by each individual thanks to wireless technologies and the Internet of Things devices. Each device has its own fingerprint which, combined with its connection context each time it consumes any cloud service or application, identifies itself in a unique way from any place and any time. This paper proposes a way to classify these social interactions for achieving to guess an identification for each individual based on their device fingerprint information gathered. The proposed solution is validated and evaluated using a use case with a real and completely anonymized dataset.