Social Network Analysis for E-mail Filtering

Existing approaches to e-mail filtering measure the likelihood of an e-mail spam based on the blacklist, whitelist, and the spaminess of the words in the content. However, false negative rates and false positive rates cannot be effectively improved without adopting new technologies. In this paper, we propose a social network analysis approach to e-mail filtering based on the previous communication history and the spaminess of the content. For a specific user, the senders of received mails and the recipients of sent mails can be extracted to form a partial social network. With the information about the spaminess score of each e-mail, we can automatically identify the “reputation” of the e-mail sender. The recipients of sent mails can also be a good source of the “closeness” between the user and the recipients. Finally in the experiments, the effectiveness of such an approach is verified.