Privacy on Reddit? Towards Large-scale User Classification

Reddit is a social news website that aims to provide user privacy by encouraging them to use pseudonyms and refraining from any kind of personal data collection. However, users are often not aware of possibilities to indirectly gather a lot of information about them by analyzing their contributions and behaviour on this site. In order to investigate the feasibility of large-scale user classification with respect to the attributes social gender and citizenship this article provides and evaluates several data mining techniques. First, a large text corpus is collected from Reddit and annotations are derived using lexical rules. Then, a discriminative approach on classification using support vector machines is undertaken and extended by using topics generated by a latent Dirichlet allocation as features. Based on supervised latent Dirichlet allocation, a new generative model is drafted and implemented that captures Reddit's specific structure of organizing information exchange. Finally, the presented techniques for user classification are evaluated and compared in terms of classification performance as well as time efficiency. Our results indicate that large-scale user classification on Reddit is feasible, which may raise privacy concerns among its community. 1 Introduction The popularity of social media and social networking has risen continuously since its first appearance. Users generate massive amounts of content on Facebook, Twitter, and similar social networking sites as well as on blogs, video sharing platforms, etc. Many companies are searching for opportunities in analyzing social media entries (Kaplan and Haenlein, 2010). This ranges from automatic processing of product reviews and opinions especially with the social components of online retailers (Oelke et al., 2009; Popescu and Etzioni, 2007) to predictions of the financial market (Ferguson et al., 2009) and even mining companies' market structures (Netzer et al., 2012). Also social sciences increasingly use the information contained in social websites. Especially political movements such as the Arab Spring (Lotan et al., 2011) or Occupy Wall Street (Tremayne, 2014) have been studied with the help of content analysis of social media. Many of these analyses require knowledge of certain demographics such as the origin of the authors of messages. However, many users might begin to share less information about their profiles as anonymity and privacy are gaining more attention during the last years. In particular, the so called NSA leak in which highly confidential documents about online user surveillance by the US