Network analysis and privacy
According to the article linked below, Facebook has been covertly utilizing commonly known network analysis techniques to build “shadow profiles”, profiles filled with data on individuals who have never created a Facebook account, and who may have never even heard of the service. These profiles are created by analysis of a wide range of content, including phone contact information shared via the “find friends” feature and by tracking tagged people on a users photos. Utilizing these techniques Facebook can potentially circumvent an individual’s attempt at privacy and identify them within their network of friends. Going further, Facebook could even use this network of real and shadow profiles to make key predictions about shadow profiles based on their surrounding networks ( the techniques for this are referenced here: https://www.huffingtonpost.com/jules-polonetsky/curly-fries-and-big-data-_b_5522550.html ).
To explore the implications of this further, take the following network of Facebook profiles, where an edge means two people are friends:
Now lets say that all of these users decide to utilize the “find friends” feature, and in the process of searching their contacts for friends, Facebook identifies contact information for a previously unknown individual “X” on Bob, Fred, James, and Alice’s phones. Simply knowing that someone exists is not that interesting. But what if Facebook decided to check which of these users have liked Republican or Democrat leaning pages, and mapped this information onto the network? Lets add in age information too, since registering a profile requires a date of birth.
Now we’ve created a much more interesting picture. Without he/she ever having created a profile, Facebook can predict the age and political leanings of this unknown individual “X”. In addition, we can infer that this individual X may have some unique social power, in that they act as a local bridge between Bob and James and serve to link together 4 like-minded individuals. Perhaps X is a leading member in a local democratic group, or a publisher of an underground democratic zine? This is all speculation, but speculation that could be further supported with deeper digging into the “internet-public” backgrounds of these individuals.
While on the surface activities such as shadow profiling may seem unimportant and obscure, by apply some of our network analysis tools we are able to reveal a much more interesting picture of the information at stake when these profiles are created.
What’s a Facebook shadow profile, and should you be worried about it?