Leveraging Personalization to Facilitate Privacy

Abstract

Online social networks have enabled new methods and modalities of collaboration and sharing. These advances bring privacy concerns: online social data is more accessible and persistent and simultaneously less contextualized than traditional social interactions. To allay these concerns, many web services allow users to configure their privacy settings based on a set of multiple-choice questions.

We suggest a new paradigm for privacy options. Instead of suggesting the same defaults to each user, services can leverage knowledge of users' traits to recommend a machine-learned prediction of their privacy preferences for Facebook. As a case study, we build and evaluate MyPrivacy, a publicly available web application that suggests personalized privacy settings. An evaluation with 199 users shows that users find the suggestions to be appropriate and private; furthermore, they express intent to implement the recommendations made by MyPrivacy. This supports the proposal to put personalization to work in online communities to promote privacy and security.

Keywords: privacy, social networks, machine learning, personalization, access control