Using Social Networks for Filtering Spam

With the recent vast prevalence of, many times, nefarious e-mail distribution (aka spam), there is an ever-increasing need for research into more effective methods of controlling, or more precisely filtering, these spam e-mails. A great deal of attention has been paid to recent breakthroughs in algorithmic spam filtering techniques. This survey highlights two particularly unique and ground breaking techniques for filtering spam e-mails.

There are two realms of spam filtering that have received particularly close attention in the research that is currently being done. Considering that each mechanism has its own associated strengths and weaknesses, it appears, based on two fairly recent, groundbreaking approaches, that a technique which utilizes a combination of these two ideas might be the most effective. One mechanism that is being given such attention, and is used quite often in practice is the idea of 'whitelist' filtering. Whitelist filters, filter e-mail by only delivering e-mails from user-approved addresses, while all other e-mails are placed in an alternate, lower-priority, folder. While this method is almost 100% effective at filtering spam, it has the obvious problem of filtering out e-mails which are not spam (false positives). Additionally, this technique requires active participation from the user, forcing them to create the 'whitelist' used to filter addresses. The second realm of spam filtering utilizes a social network and local clustering concepts to filter spam as valid based on certain clustering thresholds.

There are two predominately unique spam filtering algorithms which have recently risen to the forefront. Both algorithms use a somewhat modified collaborative approach for identifying e-mail messages as spam.