How often social objects meet each other? Analysis of the properties of a social network of IoT devices based on real data Hamid Zargari Asl , Antonio Iera University of Reggio Calabria {name.surname}@unirc.it Luigi Atzori University of Cagliari l.atzori@diee.unica.it Giacomo Morabito University of Catania giacomo.morabito@dieei.unict.it Abstract— Internet of Things (IoT) applications will be based on the interactions between smart objects. In many applications such interactions are possible (or meaningful) when objects are close to each others, i.e., there is a co-presence. Unfortunately, to date there are no data traces providing information about the co- presence of smart objects. Indeed, several mobility traces reporting humans’ movements are available, but none of them contains information about the interactions between their objects. Objective of the work reported in this paper is to fill this gap. To this purpose, we start from user mobility patterns available from several datasets. We associate to each user a set of objects, based on a survey we have carried out over around 450 users. Accordingly, we analyze the statistics about the co- presence of objects. We carry out our analysis by exploiting the tools developed for the analysis of complex networks. Our objective is to identify the objects which are likely to play a key role in the interactions between smart objects in the IoT. Keywords—IoT, social graph, mobility data set I. INTRODUCTION In the web 2.0 era, social awareness has become a key enabler of significant changes in applications and protocols designed in the view of increasing the system performance and the Quality of Experience offered to users. In this context, relevant examples include Social oriented forwarding in Delay Tolerant Networks (DTN) [1], realistic mobility models based on social networks for VANET [2], social networking in peer- to-peer systems [3]. The Internet of Things domain is no exception to this trend. In the future Internet, the majority of connections will not be established among humans but among devices [4] (things, more or less “smart”). Therefore, typical notions, rules, modes of interaction, and dynamics of social networking must inevitably be extended to the networks of objects. Smart objects will need to operate in tremendously multifaceted contexts and it seems unlikely that a single (even very smart) object will ever have the capabilities to face the deriving complexity by themselves. A new generation of social objects may possibly interact with other objects to deal with this complexity. In this vision, “Social IoT” (SIoT) [5] networks use concepts and technologies of social networks to foster resource visibility, service discovery, object reputation assessment, source crowding, and service composition. To users, this approach gives the possibility of enjoying new powerful applications, which rely on social interactions between mobile devices and environmental sensors to hide the complexity of the underlying IoT and to offer a more natural access to complex ICT environments (by also interacting with Social Networks of humans). To the underlying network layers, the approach gives the opportunity to improve the efficiency of policies for the management and selection of communication resources based on social concepts such as community and centrality. A first approach to follow is to establish social relationships between devices (PC, mobile phones, etc.) that show a certain customary in exchanging data. It is easy to find examples of studies focusing on the analysis of social networks established between cellular phones. These rely on the availability of real logs of phone calls, which means that the social relationships of devices are derived from social interactions among their owners [6]. A complementary approach is used to build so called detected social networks [7] (i.e. devices that have a high probability to come into short-range coverage and establish social links), which have proven to be highly effective in supporting information exchange in DTN or opportunistic communications [8]. This paper starts from the following observations: (i) the cited studies are reductive because usually the only things considered are cellular phones that either remotely call one another or are mutually detected through short-range technologies (ii) the availability of real data relevant to the probability that different things come into contact is limited and restricted to specific events [9] (iii) in no study there is a real notion of social relationships among devices, on which IoT protocols and applications should rely. The main focus of this paper is thus on interactions among things. Our objective is to understand if, and to which extent, This is a preprint version of a paper accepted for presentation at the IEEE Globecom 2013 Conference. The published version will be made available by IEEE in the IEEEXplore Digital Library.