Visualizing authorship for identification

  • Authors:
  • Ahmed Abbasi;Hsinchun Chen

  • Affiliations:
  • Department of Management Information Systems, The University of Arizona, Tucson, AZ;Department of Management Information Systems, The University of Arizona, Tucson, AZ

  • Venue:
  • ISI'06 Proceedings of the 4th IEEE international conference on Intelligence and Security Informatics
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

As a result of growing misuse of online anonymity, researchers have begun to create visualization tools to facilitate greater user accountability in online communities. In this study we created an authorship visualization called Writeprints that can help identify individuals based on their writing style. The visualization creates unique writing style patterns that can be automatically identified in a manner similar to fingerprint biometric systems. Writeprints is a principal component analysis based technique that uses a dynamic feature-based sliding window algorithm, making it well suited at visualizing authorship across larger groups of messages. We evaluated the effectiveness of the visualization across messages from three English and Arabic forums in comparison with Support Vector Machines (SVM) and found that Writeprints provided excellent classification performance, significantly outperforming SVM in many instances. Based on our results, we believe the visualization can assist law enforcement in identifying cyber criminals and also help users authenticate fellow online members in order to deter cyber deception.