Abstract: Spam has grown alarmingly despite many attempts to combat it. Various estimates indicate that over 60% of the world's daily email volume today (~100 billion messages) is spam.

There's no dearth of proposals to "solve the spam problem".  For email, the most popular current approach is to use spam filters that assign a "spam score" to a message.  While filters are reasonably good at reducing the amount of spam that humans must process, they have two problems.  First, spammers are creative, and periodically figure out how to get past filters.  Coping with that requires administrators to constantly tune and update filters to handle new strains of spam. Second, and more important, email has undeniably become much less reliable than before.  The utterance, "Oops, your message must have been caught by my spam filter" has now become common.

Is reduced reliability the unavoidable cost of spam?  Or is there a different approach that doesn't require software to divine the intent behind any given message?  I will discuss an alternate (and previously proposed) mechanism based on quotas and stamps that can restore email reliability while reducing spam, focusing on the several technical and socio-economic challenges that must be overcome for this solution to take root.  The technical solutions turn out to be interesting in the context of other distributed systems too, while the economic arguments require an understanding of the spammer ecosystem and how it's likely to evolve.
This talk is largely based on joint work with Michael Walfish, David Karger, and Scott Shenker.

Bio:
Hari Balakrishnan is a Professor of Computer Science and Engineering at MIT, where he works on networked computer systems.  In addition to many widely cited papers, several systems developed as part of his work are available in the public domain.  Some are in production or commercial use, including the Cricket location system, the RON and MONET overlay networks, the Chord distributed hash table, the rcc network checker, and the Aurora/Medusa stream processing engine.  Balakrishnan received a Ph.D. in Computer Science from UC Berkeley in 1998 and a B.Tech. from IIT Madras in 1993.  His honors include a Sloan Fellowship, the ACM doctoral dissertation award, and eight award-winning papers including the IEEE Communication Society's William R. Bennett Prize. He has also received awards for excellence in research and teaching at MIT (Edgerton, Spira, and Junior Bose awards).  In 2003, he co-founded StreamBase Systems, makers of a stream processing engine to process complex messages and events.

Tuesday, November 28, 2006

3:00 pm - 4:15 pm

Wu & Chen Auditorium

101 Levine Hall