Authorship Attribution Using Writeprints

Authorship attribution in online domains has become increasingly relevant and important for exposing anonymous identities and cybercrime. In this paper I present an implementation of Writeprints [1], a state-of-the-art authorship attribution method. Writeprints is a Karhunen-Loéve-transforms-based technique that uses pattern disruption algorithm with individual author-level feature sets. The implementation is evaluated on the Extended Brennan-Greenstadt Adversarial Corpus [2] in both standard and adversarial settings.