Seminar on Computational Learning and Adaptation

  Spam and Phishing
  A study of text classification in an adversarial domain

Jon Oliver
MailFrontier, Inc.

One of the most interesting features of the Spam and Phishing domains is their adversarial nature. This creates many problems for statistical filters (Bayesian and others), keywords filters and other traditional approaches to filtering email.

In this talk, I will highlight some of the problems for both filters and people at this classification task. I will present results on the difficulty people have with discriminating between phishing and legitimate email. I will examine statistical approaches which overcome some of these obstacles.

Return to the seminar schedule