Collective Classification for Spam Filtering - CISIS 2011

The document discusses spam email and various techniques for detecting and filtering spam messages. It begins with a Monty Python sketch about spam food and then discusses how spam costs billions in lost productivity by infecting computers and stealing credentials. It presents various anti-spam methods like pre-sending filtering, new protocols, and increasing spammers' costs. It evaluates supervised machine learning approaches and collective classification, which leverages relationships between documents to improve spam detection without requiring extensive labeling. Evaluation results show collective techniques outperform individual classification. The document concludes by discussing using these approaches to overcome unclassified spam and potentially reduce spam by 95% by disrupting spammers' payment systems.