|
FastMail Forum All posts relating to FastMail.FM should go here: suggestions, comments, requests for help, complaints, technical issues etc. |
|
Thread Tools |
22 Feb 2007, 11:00 PM | #1 |
Essential Contributor
Join Date: Oct 2003
Posts: 327
|
Over-aggressive SpamAssasin rules?
One of the newsletters I subscribe to just got tagged with an abnormally high spam score:
Code:
X-Sieve: CMU Sieve 2.3 X-Spam-score: 9.9 X-Spam-hits: BAYES_50 0.001, FORGED_RCVD_HELO 0.135, HOT_NASTY 0.157, HTML_30_40 0.374, HTML_MESSAGE 0.001, ME_GEOCITIES 1, ME_GEOCITIES_MANY 6, MIME_HTML_MOSTLY 1.102, SPF_FAIL 1.142 Code:
<A HREF="http://www.geocities.com/ResearchTriangle/Lab/1131/eng/safe.html" TARGET="other"> http://www.geocities.com/ResearchTriangle/Lab/1131/eng/safe.html</A> Rob, if you're reading it, please lower that score to something reasonable or, even better, adjust the SpamAssasin rule to not trigger on two URLs within a single <A HREF>. Thank you. |
23 Feb 2007, 01:32 PM | #2 |
Intergalactic Postmaster
Join Date: Oct 2001
Location: Melbourne, Australia
Posts: 6,102
Representative of:
Fastmail.FM |
It was a bunch of rules I added a while back to catch the bloody geocities "watch" spam and other related stuff that was flooding inboxes for a while and ******* people off.
Here's the complete ruleset I added. Yes, it's all pretty arbitrary, but it did the job. Code:
uri ME_GEOCITIES /\.geocities\./i score ME_GEOCITIES 1.00 body __ME_GEOCITIES_WATCH /\b(?:watch|tag h[uer]+\b|cart[ier]+\b|wristwatch)/i score __ME_GEOCITIES_WATCH 0.05 body __ME_GEOCITIES_CHRONO /\b(?:chronometers?|timekeepers?)\b/i score __ME_GEOCITIES_CHRONO 0.05 body __ME_GEOCITIES_TIMEPIECE /\b(?:time ?pieces?|masterpieces?|swiss)\b/i score __ME_GEOCITIES_TIMEPIECE 0.05 body __ME_GEOCITIES_CHIC /\b(?:chic|stainless steel|studded diamond|classic one|precious beauties|utterly delighted|sheer elegance|elegent|fashion|opulent|vogue)\b/i score __ME_GEOCITIES_CHIC 0.05 body __ME_GEOCITIES_APPAREL /\b(?:apparel|accoutrements|)\b/i score __ME_GEOCITIES_APPAREL 0.05 meta __ME_GEOCITIES_WATCHTERM (__ME_GEOCITIES_WATCH*2 + __ME_GEOCITIES_CHRONO*3 + __ME_GEOCITIES_TIMEPIECE*2 + __ME_GEOCITIES_CHIC*2 + __ME_GEOCITIES_APPAREL*2) >= 3 score __ME_GEOCITIES_WATCHTERM 0.05 body __ME_GEOCITIES_PRESCRIPTIONS /\b(?:prescriptions?)\b/i score __ME_GEOCITIES_PRESCRIPTIONS 0.05 body __ME_GEOCITIES_DRUG /\b(?:drugs?|medical|remedy|health issues?|meds)\b/i score __ME_GEOCITIES_DRUG 0.05 meta __ME_GEOCITIES_DRUGTERM (__ME_GEOCITIES_PRESCRIPTIONS*2 + __ME_GEOCITIES_DRUG*2) >= 2 score __ME_GEOCITIES_DRUGTERM 0.05 body __ME_GEOCITIES_DISCOUNT /\b(?:discount|min costs?|cost[- ]conscious|afford it|reasonable price|less money|spending less|bargain|economical|worry.{1,15}bills)/i score __ME_GEOCITIES_DISCOUNT 0.05 body __ME_GEOCITIES_SATISFACTION /\b(satisfaction|satisfied|money[- ]back)\b/i score __ME_GEOCITIES_SATISFACTION 0.05 body __ME_GEOCITIES_INVENTORY /\binventory\b/i score __ME_GEOCITIES_INVENTORY 0.05 body __ME_GEOCITIES_SAFE /\b(safe|best service)\b/i score __ME_GEOCITIES_SAFE 0.05 body __ME_GEOCITIES_AFFORDABLE /\bafford(?:able)\b/i score __ME_GEOCITIES_AFFORDABLE 0.05 body __ME_GEOCITIES_STORE /\bstore\b/i score __ME_GEOCITIES_STORE 0.05 body __ME_GEOCITIES_SHIP /\b(shipped|world wide)\b/i score __ME_GEOCITIES_SHIP 0.05 meta __ME_GEOCITIES_MONEYTERM (__ME_GEOCITIES_DISCOUNT || __ME_GEOCITIES_SATISFACTION || __ME_GEOCITIES_INVENTORY || __ME_GEOCITIES_SAFE || __ME_GEOCITIES_AFFORDABLE || __ME_GEOCITIES_STORE || __ME_GEOCITIES_SHIP) score __ME_GEOCITIES_MONEYTERM 0.05 meta ME_GEOCITIES_MANY ME_GEOCITIES && (__ME_GEOCITIES_WATCHTERM || __ME_GEOCITIES_DRUGTERM) && __ME_GEOCITIES_MONEYTERM score ME_GEOCITIES_MANY 6.0 1. A geocities link in it 2. A "watch" term or a "drug" term 3. A "money" term I wonder what was in your newsletter that triggered it? Rob |
23 Feb 2007, 10:45 PM | #3 | ||
Essential Contributor
Join Date: Oct 2003
Posts: 327
|
Quote:
Quote:
http://techsupportalert.com/issues/issue142.htm Incidently, I have rules based on individual SA tags. It would help if you could make your SA config available so I could check the criteria for those tags. For example, one email that I got was marked with KAM_STOCKOTC and STRONG_BUY. I could not find the actual rules for these tags anywhere on the web. Finally, quite a lot of email that I get (spam and non-spam) is marked with FORGED_RCVD_HELO. A quick google check produced this: http://www.mailarchives.org/list/spa...msg/2003/00001 |
||
26 Feb 2007, 01:12 PM | #4 |
Intergalactic Postmaster
Join Date: Oct 2001
Location: Melbourne, Australia
Posts: 6,102
Representative of:
Fastmail.FM |
I've reduced the score, but it should be pretty easy to add the From address of the newsletter to your address book to whitelist it as well.
Rob |
26 Feb 2007, 01:15 PM | #5 |
Intergalactic Postmaster
Join Date: Oct 2001
Location: Melbourne, Australia
Posts: 6,102
Representative of:
Fastmail.FM |
All the scores we use are pretty standard. Either their base SA ones, or ones from sa-update, or from the SARE repository. Any ones we add ourselves we always prefix them with ME_
Rob |