Collaborative E-Mail Filtering

Acknowledgements I would like to thank my advisor, Dr. Susan Gauch, for guidance throughout my undergraduate and graduate years at the University of Kansas. Also, thanks to my thesis committee: Dr. Victor Frost, to whom I also owe my many opportunities at ITTC, and Dr. Perry Alexander, who served on my committee on short notice. I would like to thank my family: Ralph, Charlene, Jeff, and Denise for always supporting all of the educational endeavors that I have chosen. Finally, thanks to my friends and co-workers at ITTC. Some helped by volunteering their e-mail for the data set, and others lent an ear when needed. Abstract The concept of e-mail as a quick, free method of information communication for business and personal use may soon be overshadowed by the high percentage of SPAM infiltrating user's inboxes. As of May 2004, two-thirds of the world's e-mail is SPAM. Users must now sort through this high quantity of SPAM to find legitimate messages. Filtering techniques are needed to reduce the amount of SPAM that has to be manually sorted by the user. Several statistical methods have been used, and have shown great performance, excelling in adapting to the ever changing content of SPAM e-mail. This thesis explores using statistical methods, along with collaboration between users, to further reduce SPAM. Collaboration is a fairly new concept in e-mail filtering, but may become the next technology to save e-mail communication as we know it.