The problems that social media users have in appropriately controlling access to their content has been well documented in previous research. A promising method of providing assistance to users is by learning from the access control decisions made by them and making future recommendations.
In this paper, we present REACT, a learning mechanism which utilizes information available in the social network in conjunction with information about the content to be shared to provide users with access control recommendations. We demonstrate the highly accurate performance of REACT through a detailed empirical evaluation and also discuss ways of personalizing it for different users in order to improve performance even further.