Section: New Results

Middleware for Mobile Social Networks

Participants : Animesh Pathak, George Rosca.

As recent trends show, online social networks (OSNs) are increasingly turning mobile and further calling for decentralized social data management. This trend is only going to increase in the near future, based on the increased activity, both by established players like Facebook and new players in the domain such as Google, Instagram, and Pinterest. Modern smart phones can thus be regarded as social sensors, collecting data not only passively using, e.g., Bluetooth neighborhoods, but actively in the form of, e.g., “check-in”s by users to locations. The resulting (mobile) social ecosystems are thus an emergent area of interest.

The recent years have seen three major trends in the world of online social networks: i) users have begun to care more about the privacy of their data stored by large OSNs such as Facebook, and have won the right (at least in the EU) to remove it completely from the OSN if they want to; ii) OSNs are making their presence felt beyond casual, personal interactions to corporate, professional ones as well, starting with LinkedIn, and most recently with the purchase by Microsoft of Yammer, the enterprise social networking startup, and the launch of Google Plus for enterprise customers; and iii) users are increasingly using the capabilities of their (multiple) mobile devices to enrich their social interactions, ranging from posting cellphone-camera photos on Instagram to “checking-in” to a GPS location using Foursquare.

In view of the above, we envision that in the near future, the use of ICT to enrich our social interactions will grow (including both personal and professional interactions), both in terms of size and complexity. However current OSNs act mostly like data silos, storing and analyzing their users' data, while locking in these users to their servers, with non-existent support for federation; this is reminiscent of the early days of email, where one could only email those who had accounts on the same Unix machine. The knee-jerk reaction to this has been to explore completely decentralized social networks, which give the user complete control over and responsibility of their social data, while resorting to peer-to-peer communication protocols to navigate their social networks. Unfortunately, there are few techniques available to reconcile with the fact that the same user might have multiple devices, or that it is extremely resource-consuming to perform complex analysis of social graphs on small mobile devices.

Our view lies somewhere in the middle of the two extremes, taking inspiration from the manner in which users currently use email. While their inboxes contain an immense amount of extremely personal data, most users are happy to entrust it to corporate or personal email providers (or store and manage it individually on their personal email servers) all the while being able to communicate with users on any other email server. The notion of Federated Social Networks (FSNs) —already gaining some traction— envisions a similar ecosystem where users are free to choose OSN providers which will provide storage and management of their social information, while allowing customers using different OSN providers to interact socially. Such a federation can be beneficial in three major ways, among others: i) it allows users to enjoy properties such as reliability, availability, and computational power of the hosting infrastructure of their choice, while not being locked down in terms of whom they can communicate with; ii) much like spam filtering services provided by modern email providers, that are tuned by feedback from their users, FSN users can benefit from the behavior of others sharing the same OSN provider (This also gives an incentive to commercial OSN providers to provide value-added services.); and iii) this fits perfectly with enterprise needs, where ad-hoc teams can be formed across corporate OSN providers of two organizations to work on a joint project.

In [30] , we presented a set of requirements, followed by a survey of the state of the art in social networking solutions, with a special focus on their ability to support rich privacy and access control policies in federated settings. Through this extensive analysis we offer a broad vision on existing social networking platforms, protocols involved but also their privacy and access policies. By doing so, we identify the main components of a federated social platform together with presenting the current trends in standards and security paradigms underlying actual open source solutions which offers their implementation, and finally provides recommendations on constructing such systems. Our research is continually being incorporated into the Yarta middleware for mobile social networking(§  5.7 ).