Statistical Authorship Attribution of Mexican Drug Trafficking Online Forum Posts

Abstract

As Mexico’s drug war violence escalates, there has recently been a proliferation of online forums about drug trafficking in this country. With considerable usage statistics, some of these forums show contributions from alleged law enforcement officers, soldiers, drug dealers, and regular citizens. In this paper I will show that state of the art statistical authorship attribution techniques can be successfully applied to the contributions of prolific users in these forums. The 99% diagnostic accuracy and 93% sensitivity obtained here are higher than previously achieved equivalent figures for the method employed. These results were obtained with the statistically relevant contribution of a number of novel discriminating features. A proposal for the standardization of results obtained with discriminant analysis (DA) in authorship attribution is also presented.