As online social network (OSN) providers such as Facebook
deal with tremendous amounts of user information, they can
obtain a deep insight into their users’ personal opinions, social
relationships, and economical or political preferences, raising
severe privacy and security concerns. Facebook, for example,
recently reached the one billion user mark and already controls
the personal data of more than 10% of the world’s population.
Yet, the hunt for user data is not over and users are often helpless
when imposed with changes of any kind, as demonstrated
by Facebook’s recent acquisition of the Instagram user base
with a billion-dollar agreement. Moreover, centralized control
often results in the misuse of user data. For example,
LinkedIn leaked millions of passwords of its users, and
the Facebook Beacon application passed sensitive shopping
information of users to Facebook without their consent.
While it remains to be seen whether OSN providers would
give up a major source of income and grant comprehensive
security and privacy means to their users (e.g., encryption of
data), decentralized OSNs are becoming more promising for
better user data security and privacy. Instead of relying on a
central data repository, a decentralized OSN can allow users to
regain control over their data. For instance, users can encrypt
their data according to their own need, enforce access control
of their data at their discretion, or distribute their data to their
designated storage facilities or user devices across the OSN.
With SOUP, we design and evaluate a data placement
solution for decentralized OSNs that makes all data of all users
available at all times in the absence of a central OSN provider.
Whereas one might consider classical peer-to-peer (P2P) data
storage approaches, the distributed OSN requires a unique
data placement solution for a number of reasons. First, the
online patterns of OSN users are much more bursty than those
in traditional decentralized applications. Second, social
networking platforms are increasingly employed by mobile
devices, whose storage capacity is often meager. Third, an
OSN needs to be robust as a whole to ensure all data of interest
are highly available, including those from users who are not
active or contributing.
Recently, researchers have proposed a wide range of alternative
systems that consider the characteristics of an OSN in
their design. In particular, different from
the tit-for-tat in classical P2P networks, the inherent social
relation among OSN users offers potential incentives for users
to host data for each other. While some of these solutions
provide means to realize data storage in decentralized OSNs,
each introduces new shortcomings, ranging from technical and
economical deployability issues to limited robustness
and discrimination of users due to a dependency on nodes to
which the user is connected in the social graph.