I don't have a better solution for server-side filtering, but for my part, I set up two filters: One that looks in the subject line for obvious words like "penis", "viagra", "vitamin", "coupon", "mortgage", "win", "free", etc. There's about 20 words in all - I don't remember all of them and I'm not going to look but I'm sure you can figure it out.
The second filter looks in the body for the word "click".
These two filters catch more spam than SA, which runs on m.org's server, ever does.
Here's a few bits from a recent spam I got, which SA caught:
1.6 RAZOR2_CF_RANGE_51_100 BODY: Razor2 gives confidence between 51 and 100
[cf: 100]
0.9 RAZOR2_CHECK Listed in Razor2 (http://razor.sf.net/)
This is completely idiotic. It assigns points for the sender being listed in Razor2, and then additional points for the "confidence" rating of Razor2. It's redundant and double-jeopardy. If the Razor2 check is that important, lump them together - if Razor2 gives a confidence of 51-100, then it's safe to assume it's listed in Razor2. Having seperate criteria for that is stupid.
Meanwhile, it also assigns points for
0.6 HTML_WEB_BUGS BODY: Image tag intended to identify you
Which is a pretty fucking obvious spam tactic, but only gets .6 points.
However, it also assigns like 2 points if the "From:" field isn't a "real name". There are tons of examples of legitimate email where the from: field is just an address, not a name, such as support@bellsouth.net, or billing@earthlink.com, or whatever else. SA tags those as spam based on the from: field, yet web-tracking ID tags get less than a point. Spare me.
Here's another oddity:
0.0 HTML_MESSAGE BODY: HTML included in message
1.1 HTML_90_100 BODY: Message is 90% to 100% HTML
In my line of work, I used to get tons of email from stupid secretaries or other computer-illiterate drooling cretins who send out HTML emails, so they can make their text pink and in some stupid font. Lots of people, unfortunately, send HTML email.
Now, SA assigns no points just for having HTML in the email, but does assign points for the ratio of HTML to plaintext. Frankly I'd be more leery of an email that was mixed plaintext and HTML, than one that was all HTML. If it's 100%, it could easily be someone just sending an HTML email - if it's only 50% HTML it's probably a spammer.
I could go on but I've made my point. SA's criteria sucks and so does it's point system. It frequently gives false positives because of this, misses other obvious spams, and my simple, client-side filters catch more than SA does.
kitten.
"What a piece of work is man..
In action, how like an angel,
In apprehension, how like a god."
[ Parent ]
|
![]](/images/header/rt_edge3.gif){kind=small}
![[XML]](/images/rssblue.gif){kind=small}
![[P]](/images/print3.gif){kind=small}
