[rdfweb-dev] "Can Social Networking Stop Spam?"

Graham Klyne gk at ninebynine.org
Thu Mar 18 21:30:06 UTC 2004

[From ACM technews... #g]

# "Can Social Networking Stop Spam?"
NewsFactor Network (03/15/04); Martin, Mike

A new algorithm developed by UCLA professors P. Oscar Boykin and Vwani 
Roychowdhury applies social networking principles to spam filtering. "We 
routinely use our social networks to judge the trustworthiness of 
outsiders...to decide where to buy our next car, or where to find a good 
mechanic," notes Roychowdhury. "An email user may similarly use his email 
network, constructed solely from sender and recipient information available 
in the email headers, to distinguish between...'spam,' and emails 
associated with his circles of friends." The researchers' algorithm 
processes a specific user's personal email network to concurrently 
determine both the user's trusted networks of friends and spam-spawned 
sub-networks, Boykin explains, adding that the algorithm distinguished 
between spam and legitimate email with no errors or false negatives in a 
recent test. The researchers studied six weeks' worth of emails from 
assorted individuals so they could ascertain the "components" of their 
email network, a component being a series of nodes that can connect to each 
other in the network, according to Boykin; analyzing "clustering 
coefficients" in a network--provided the network is big enough--is an easy 
way to tell spam and non-spam components apart. Boykin says he and 
Roychowdhury observed that clustering coefficients run high for non-spam 
components, and are equal to zero for spam components. Roychowdhury's 
colleague attests that the algorithm can be used to train content-based 
filters to recognize words and phrases typical of spam and non-spam, once 
50 percent of email can be accurately classified as either junk or 
legitimate email. Boykin points out that the tool also produces white lists 
and blacklists used to verify that content filters are properly classifying 
Click Here to View Full Article:

Graham Klyne
For email:

More information about the foaf-dev mailing list