Race Conditions in Distributed Enforcement at the Example of Online 
Social Networks


In today's society, online social networks (OSNs) play an increasingly 
important role in sharing common interests between individuals. It is 
favourable to give the users of OSNs more control over their sensitive 
data so that the data owners can limit not only the access to the data, 
but also the future usage of data by other users. These security policies 
can sensibly base on inter-personal relationships between contacts. 

Enforcing such security policies can be achieved by dynamically monitoring 
the social network and intervening upon detecting unacceptable actions. 
In distributed social networks (DSNs), determining whether or not an 
action is acceptable usually requires communication between different 
servers. During the decision-making, changes in the state of servers, 
such as changing the type of relationships between contacts, can occur 
that lead to incorrect decisions. Detecting and preventing race conditions 
are  challenging research topics.

This thesis studies the soundness of runtime enforcement of owner-obeying 
security policies in DSNs. We formally define such a security policy, 
present an enforcement mechanism at runtime for it, and discuss some 
approaches for achieving a sound enforcement, in context of time of check 
to time of use (TOCTTOU) race conditions. We also present an 
implementation of our enforcement on the distributed social network 
Diaspora*.