\"I knew they clicked when i saw them with their friends\": identifying your silent web visitors on social media

An increasing fraction of users access content on the web from social media. Endorsements by microbloggers and public figures you connect with gradually replaces the curation originally in the hand of traditional media sources. One expects a social media provider to possess a unique ability to analyze audience and trends since they collect not only information about what you actively share, but also about what you silently watch. Your behavior in the latter seems safe from observations outside your online service provider, for privacy but also commercial reasons. In this paper, we show that supposing that your passive web visits are anonymous to your host is a fragile assumption, or alternatively that third parties -- content publishers or providers serving ads onto them -- can efficiently reconciliate visitors with their social media identities. What is remarkable in this technique is that it need no support from the social media provider, it seamlessly applies to visitors who \\emph{never} post or endorse content, and a visitor's public identity become known after a few clicks. This method combines properties of the public follower graph with posting behaviors and recent time-based inference, making it difficult to evade without drastic or time-wasting measures. It potentially offers researchers working on traffic datasets a new view into who access content or through which channels.