Empirical evaluation of authorship obfuscation using JGAAP

Authors:
Patrick Juola;Darren Vescovi
Affiliations:
Duquesne University, Pittsburgh, PA, USA;Duquesne University, Pittsburgh, PA, USA
Venue:
Proceedings of the 3rd ACM workshop on Artificial intelligence and security
Year:
2010

Citing 3
Cited 2

Mining e-mail content for author identification forensics

ACM SIGMOD Record
Whose thumb is it anyway?: classifying author personality from weblog text

COLING-ACL '06 Proceedings of the COLING/ACL on Main conference poster sessions
Authorship attribution

Foundations and Trends in Information Retrieval

Adversarial stylometry: Circumventing authorship recognition to preserve privacy and anonymity

ACM Transactions on Information and System Security (TISSEC)
Detecting stylistic deception

EACL 2012 Proceedings of the Workshop on Computational Approaches to Deception Detection

Quantified Score

Hi-index	0.01

Visualization

Abstract

Authorship attribution is an important emerging security tool. However, just as criminals may wear gloves to hide their fingerprints, so authors may choose to mask their style to escape detection. Most authorship studies have focused on cooperative and/or unaware authors who do not take such precautions. Using a newly published corpus (the Brennan-Greenstadt Obfuscation corpus), we use the JGAAP system (www.jgaap.com) to test different methods of authorship attribution against essays written in deliberate attempt to mask style. We confirm that this is an issue.