An evaluation of Naive Bayesian anti-spam filtering

TL;DR

This paper thoroughly evaluates the effectiveness of Naive Bayesian classifiers for spam filtering, analyzing various factors affecting performance and highlighting the need for additional safety measures for practical deployment.

Contribution

It provides a comprehensive evaluation on a publicly available corpus and explores the impact of multiple factors on filter performance, which had not been previously examined.

Findings

Naive Bayesian filter requires safety nets for practical use.

Attribute-set size and training data significantly affect performance.

Lemmatization and stop-lists influence spam filter effectiveness.

Abstract

It has recently been argued that a Naive Bayesian classifier can be used to filter unsolicited bulk e-mail ("spam"). We conduct a thorough evaluation of this proposal on a corpus that we make publicly available, contributing towards standard benchmarks. At the same time we investigate the effect of attribute-set size, training-corpus size, lemmatization, and stop-lists on the filter's performance, issues that had not been previously explored. After introducing appropriate cost-sensitive evaluation measures, we reach the conclusion that additional safety nets are needed for the Naive Bayesian anti-spam filter to be viable in practice.