I have what may seem like a humerous question, at first. I was wondering if anyone knew of a mailing list that receives only spam... The purpose would be to treat this list as a corpus for training the Bayesian filters.
Unfortunately, I don't believe there are any. Part of the problem is that such a list would need the full headers of the original spam to be useful to the most number of people, and getting people to report full headers is a nightmare (lots of our customers can't even do it -- often just sending the Outlook pseudo-headers (such as "Sent:", which isn't a header), and sometimes partial headers). Another problem is that people with spamtraps (E-mail addresses designed to collect just spam) often don't want to share (partly because they don't want spammers finding their spamtraps). And, the list owner would need to be careful not to get spammers who send mail to the list, in order to taint the data.
Right now, I am having a hard time convincing our 1,000 or so users to put the time and effort into classifying email as spam. The next best this, I figure, is to get novel spam pre-classified from an outside source.
Note that naive Bayes works best when done on an individual level, and is much less accurate with generic spam (since each E-mail address and domain has its own unique "spam profile"). Also, it will be less effective if you have 10,000s of spams analyzed but only hundreds/thousands of legitimate E-mails. So if you can convince your users, that's the best option.
What do you guys think? If no one knows of anything like this, I might be willing to set something up.
That could be a good idea. If it is just being used for naive Bayesian analysis, the headers problem isn't an issue. And if you are careful as to who can submit spam, it could become a very useful resource.
-Scott
---
Declude JunkMail: The advanced anti-spam solution for IMail mailservers.
Declude Virus: Catches known viruses and is the leader in mailserver vulnerability detection.
Find out what you've been missing: Ask about our free 30-day evaluation.
--- [This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)]
To Unsubscribe: http://www.ipswitch.com/support/mailing-lists.html List Archive: http://www.mail-archive.com/imail_forum%40list.ipswitch.com/ Knowledge Base/FAQ: http://www.ipswitch.com/support/IMail/
