On Mon, 2004-01-19 at 18:01, Brad Koehn wrote: > I've been thinking about a new rule, either for Bayes or for more > normal processing, and I'd like the group's opinion. It has to do with > URLs in the message. > > My original thought came to me when running SpamCop on a bunch of > messages. Taking a peek at the SC output I see that they whois the host > IP of the URLs in the message to find the email address of the netblock > owner. Running this code myself shows that the messages are coming from > netblocks in China (surprise, surprise). My question is, can we feed > use this information, either directly or through Bayes, to help predict > whether a message is spam?
I suggested a similar tactic, one that I've been using very successfully on my home machine. It uses the Perl IP::Country module to lookup the country of hosting of URLs and compares them to a list of disallowed countries (China etc). The feedback I got from the SA folks was that they'd rather make this kind of URL meta information available to the bayesian filters. Since I don't run Bayes I'd rather have it available directly. :) Derek ------------------------------------------------------- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk