How will you be identified by likes and emoticons?

Everyone has their own reasons for hiding their real name and other online identities. Some want to protect themselves from scammers, others, on the contrary, from intelligence services. Many are trying to resist state censorship in this way - to read and write what they consider necessary. Over the past 2 years, the number of users of the Tor anonymizer , which hides information about actions on the Internet, has grown several times - to approximately 2.5 million people per day. Russia is among the top five countries in terms of popularity of this network . The attitude of Internet companies towards Tor users is changing. In the fall of 2014, Mozilla promised to take into account the wishes of the anonymizer developers in future versions of its Firefox browser. Facebook, which previously periodically blocked access to the service via Tor, also made concessions.

On October 31, 2014, Facebook officially made the site accessible via Tor . The developers launched a copy of the site inside the anonymous network itself, users can now safely log into Facebook. At the same time, the social network still requires you to indicate your real first and last name, and when registering with Tor, you also need to provide a mobile phone number or an ID card (for example, a passport). If you wish, you can bypass these restrictions and create an account under a fictitious name. But even this does not mean that the identity of the anonymous person will not be able to be identified by intelligence services, scammers or someone else.

A post on Facebook under a fictitious name can, under certain conditions, be linked to a specific author by analyzing his previous texts. Each person has a unique set of characteristic writing features - repeated errors, the number of typos, types of emoticons, policies for using capital letters, etc. Scientists from Canada's Concordia University have come up with an algorithm that allows the analysis of such features to be used as convincing evidence in court. True, it only works if there is a limited circle of suspects, among whom is the author of the anonymous message. For each, unique features of his letter are found (writeprints, by analogy with fingerprints), which are absent in other suspects, and they are compared with the anonymous text. There are two ways to protect yourself from such a check: change the character of your letter down to the smallest detail or imitate someone else’s style as accurately as possible, right down to emoticons. There is no universal method to establish the authorship of an anonymous text without a limited number of suspects. The texts of the founder of the Bitcoin cryptocurrency, whose identity the researchers were trying to establish, were studied in different ways, and they led to different results.

Scientists at the University of Texas at Austin in 2009 developed an algorithm that allows one to identify the user of an anonymous social network by analyzing the structure of contacts on another social network - a public one. They used Twitter and Flickr for the experiment. In 43%, comparison of data - mutual and non-mutual friendships, subscriptions, etc. - made it possible to correlate users of one network with users of another. Moreover, in most cases it was possible to accurately determine the identity of a particular anonymous person, and errors, as a rule, led to a person closely associated with him. At the same time, the researchers noted that the audience of these networks has little overlap with each other; otherwise, the accuracy of the method could be much higher. There is a protective technique against this method of de-anonymization, described in a 2013 paper - “identity separation.” The author suggests creating different accounts on social networks for different circles of your communication - colleagues, people with similar hobbies, and others.

Likes and ratings also help to identify the anonymous person. In 2006, the Netflix company (an online cinema at that time operated as a video rental service) made publicly available the data of half a million of its users, which contained only their ratings for films and the date of these ratings. Scientists compared these data with estimates on the imdb website, which many use under their own names. If a person rated at least 8 films on both Netflix and imdb at approximately the same time (and in two cases the ratings could be directly opposite), his identity could be determined with almost one hundred percent accuracy. Fans of rare films (outside the top 500) were identified without any ratings.