On De-anonymization of Single Tweet Messages

Abstract

In this work, we address the question of whether the authorship of a single tweet can be successfully identified (and in a mixed set with other authors). Here, we present a new authorship identification scheme, which is useful in detecting authorship of short texts such as tweets, in case where only single messages are available. Our authorship identification scheme relies on selecting features that work for the special setting and combine them in order to obtain a better accuracy. This technique demonstrates significant results through out our experiments. Our results can be used to detect authors of illegitimate tweets, fake tweets in a Twitter account or break the privacy of a multi-user account by showing the authors who participate in it.