Spamassassin already does figure out 419 scams. Do a find-in-page for the word "Nigeria" on their complete ruleset.

Though in the past week, suddenly half of the 419 scam letters are getting through. They must be dropping identified words on the SA ruleset.

But I want it to query Snopes for all the stuff that comes through and put a hoax/scam warning *in the body*, with a link.

The Nigerian email is really spam, so it should just be deleted. But I can see how this would be a nice feature for those friend-forwarded hoaxes like the "Shark attacks guy hanging from helicopter" photos. Also, it would be good if there was an auto-reply to the sender.

I am using PopFile, a Bayesian classifier spam filter, which I adapted for use with procmail. After I trained it with a few 419 scams that I got, it now correctly classifies Nigerian SCAMSPAM as such.

Sorry, I'm not being clear enough. This is for simps, newbies and naifs who are apt to believe in this shit, not leet hipsters like you and me. The idea is that every Nigerian spam automatically gets tagged with a warning about Nigerian spam, every fake-eBay "we need your password again" message gets tagged, etc, with a link to Snopes and a debunking.

Edited by author 11-22-2002 05:18 PM
Cool Idea. I'm another fan of PopFile, posted a link to your idea to the sourceforge boards, and to the guy who wrote Missing Attachment Catcher.

Wonder if an ISP would be interested in this, as a value-added service (and good PR). Earthlink got scads of good press, just for making a popup blocker available.

For a few months I've been mulling over how I might provide this kind of service via Purportal.com. It's an excellent idea (as is the auto-reply-to-sender suggestion). The question is, how do you get an unsophisticated user to insert you into their mail flow? All the solutions I have considered are way too much work to develop and/or deploy, and/or are unlikely to be attractive to the target audience.

There's a bit of a catch-22 here. The target market for the service is the people who don't know they need it -- the top decile of gullibility and naivete.

I am afraid you'd need to be an ISP to really do this well. "Click here to turn on the Magic Email Descammifier..."

(I love the Lazy development model by the way. I have tried it a couple times myself.)

The laziest development model would require the sender to put the comment at the top of every email - Warning, if your [fill in] enough to believe that you can get something for nothing or be cured by over-the-counter medications when prescription medications wont even do the job... etc. Using the ISP level wouldn't thrill me. Probably have to put a price tag on it. "For only $50 dollars a year we'll read your email and decide for you what is good and bad." The idea does have some merit in that it could warn when your mother is trying to put a guilt trip on you.

I sweeten the pot a little by pointing you to the company whose technology can be used to do exactly what you described:

www.attensity.com

CIA funded company doing cutting edge linguistic analysis for both commercial and "non-commercial" applications. Very hot technology out of the University of Utah. Hugely fast, hugely accurate. Actually "understands" what is happening in a text document. Not a fragile, slow, rule-based system.