mbstone writes "Arvind Narayana writes: What if authors can be identified based on nothing but a comparison of the content they publish to other web content they have previously authored? Naryanan has a new paper to be presented at the 33rd IEEE Symposium on Security & Privacy. Just as individual telegraphers could be identified by other telegraphers from their 'fists,' Naryanan posits that an author's habitual choices of words, such as, for example, the frequency with which the author uses 'since' as opposed to 'because,' can be processed through an algorithm to identify the author's writing. Fortunately, and for now, manually altering one's writing style is effective as a countermeasure."
In this exploration the algorithm's first choice was correct 20% of the time, with the poster being in the top 20 guesses 35% of the time. Not amazing, but: "We find that we can improve precision from 20% to over 80% with only a halving of recall
. In plain English, what these numbers mean is: the algorithm does not always attempt to identify an author, but when it does, it finds the right author 80% of the time. Overall, it identifies 10% (half of 20%) of authors correctly, i.e., 10,000 out of the 100,000 authors in our dataset. Strong as these numbers are, it is important to keep in mind that in a real-life deanonymization attack on a specific target, it is likely that confidence can be greatly improved through methods discussed above — topic, manual inspection, etc."