Graduate student Anirudh Ramachandran's work on filtering spam usingnetwork-level properties will appear at the ACM Conference on Computerand Communications Security (CCS), ACM's top security conference, atthe end of October. Ramachandran and his advisor, Assistant ProfessorNick Feamster, have been working with Professor Santosh Vempala todevelop next-generation spam filtering techniques.

Spam is becoming increasingly virulent as it makes use of images andPDFs to evade content-based filters. To make matters worse, spammersare sending spam from "fresh" machines every day, which makes itdifficult to maintain static blacklists of known bad senders.

To get a step ahead, the researchers have taken a different approach:
rather than filtering spam based on content or an ephemeral identity ofthe sender (e.g., an IP address), the researchers have invented a newtechnique called "behavioral blacklisting". Behavioral blacklistingaims to learn and "fingerprint" spammers' sending patterns---forexample, the set of recipients a particular sender is targeting---andblacklist senders based on their sending behavior, rather than a fixedidentity.

The researchers developed their first behavioral blacklisting techniqueby applying Professor Vempala's novel spectral clustering algorithms,which have also successfully been applied to other areas (e.g., Websearch).

You can read the paper here.