EmailDiscussions.com  

Go Back   EmailDiscussions.com > Email Service Provider-specific Forums > FastMail Forum
Register FAQ Members List Calendar Search Today's Posts Mark Forums Read
Stay in touch wirelessly

FastMail Forum All posts relating to FastMail.FM should go here: suggestions, comments, requests for help, complaints, technical issues etc.

Reply
 
Thread Tools
Old 14 Jun 2010, 04:41 AM   #1
ezfig
Junior Member
 
Join Date: Nov 2003
Posts: 25
Frustrated - non-spam continually reported as spam

I have a paid account with a Bayes status of
Num spam learned 2827
Num non-spam learned 117
Bayes database used Global
since I cleared it several months ago, following a suggestion on this forum.

So, let's just suppose I get not one, but two daily emails I want from Casey Research. Because the Fastmail spam filter refuses to change its evil ways no matter how many times (just for the record, twice per day) I "report non-spam," I can rest assured the next day I'll have to go into the Junk Mail folder to find the daily emails.

I mentioned it to the senders, and they said to whitelist them. Which of course is only possible in Fastmail if you insist on getting emails only from people on your address book list. Not practical for me.

And then I should mention that I have several rules defined telling Fastmail to put mail from Casey or Casey Research or (____) into the IN box.

Still, every day, I have to fish them out of junk mail.

(pobox.com, looking for a new customer?)
ezfig is offline   Reply With Quote

Old 14 Jun 2010, 04:53 AM   #2
djshack
Junior Member
 
Join Date: Jun 2010
Posts: 11
If you have your spam filter settings on "Normal," this is part of what happens:
Quote:
Email from people in your address book is white-listed (always delivered)
Therefore, if you add the sender's address to your FastMail address book, the messages should never be sent to the spam folder.
djshack is offline   Reply With Quote
Old 14 Jun 2010, 05:04 AM   #3
Sherry
 Moderator 
 
Join Date: Dec 2002
Location: USA
Posts: 8,687
Quote:
Originally Posted by ezfig View Post
Num spam learned 2827
Num non-spam learned 117
Bayes database used Global
I believe you need both 200 spam "and" 200 non-spam for it to start using your personal database of what is spam and what isn't (or at least I think it used to be that way?).

Sherry
Sherry is offline   Reply With Quote
Old 15 Jun 2010, 03:50 PM   #4
emoore
Essential Contributor
 
Join Date: Apr 2002
Posts: 280
The "define rules" screen has gotten a lot more powerful over the years but it still doesn't provide as much explicit control (such as determining exactly when the spam score is tested) or support as complex rule sets as a custom sieve script. You might also find the Fastmail Sieve Tester an easier way to debug why a message didn't get handled correctly.

https://www.fastmail.fm/help/managin...ced_rules.html
http://wiki.fastmail.fm/index.php?title=SieveFAQ
emoore is offline   Reply With Quote
Old 15 Jun 2010, 09:58 PM   #5
ezfig
Junior Member
 
Join Date: Nov 2003
Posts: 25
Quote:
Originally Posted by djshack View Post
If you have your spam filter settings on "Normal," this is part of what happens:

Therefore, if you add the sender's address to your FastMail address book, the messages should never be sent to the spam folder.
Good point - I thought I had done that, but now I realize that adding the sender's email address (as they suggested) won't work because the email is actually arriving from a different address at sneakemail.com (which is awesome obtw - create disposable email addresses on the fly, etc etc).

Too clever for my own good, am I? Thanks for the suggestion. I have added the forwarded sneakemail address and we shall see, maņana....
ezfig is offline   Reply With Quote
Old 15 Jun 2010, 09:59 PM   #6
ezfig
Junior Member
 
Join Date: Nov 2003
Posts: 25
Quote:
Originally Posted by Sherry View Post
I believe you need both 200 spam "and" 200 non-spam for it to start using your personal database of what is spam and what isn't (or at least I think it used to be that way?).

Sherry
Hmmm....so maybe I should report the next 83 spam messages as non-spam...why not?
ezfig is offline   Reply With Quote
Old 15 Jun 2010, 10:03 PM   #7
ezfig
Junior Member
 
Join Date: Nov 2003
Posts: 25
Quote:
Originally Posted by emoore View Post
The "define rules" screen has gotten a lot more powerful over the years but it still doesn't provide as much explicit control (such as determining exactly when the spam score is tested) or support as complex rule sets as a custom sieve script. You might also find the Fastmail Sieve Tester an easier way to debug why a message didn't get handled correctly.

https://www.fastmail.fm/help/managin...ced_rules.html
http://wiki.fastmail.fm/index.php?title=SieveFAQ
I appreciate the links, and will dive into them if I need to, though I'd prefer not to, just as I prefer not to have to tinker under the hood of a new car. The answer, as I mentioned above with the reference to sneakemail.com, might be disturbingly simple after all....
ezfig is offline   Reply With Quote
Old 16 Jun 2010, 04:46 AM   #8
emoore
Essential Contributor
 
Join Date: Apr 2002
Posts: 280
Quote:
Originally Posted by ezfig View Post
Hmmm....so maybe I should report the next 83 spam messages as non-spam...why not?
Bad idea. It will ruin the training for your personal bayes database, which is not activated yet. At the moment you're still using the global bayes database.

Why not sign up for a couple of yahoo groups or google groups with a lot of activity where you configure it to send you a email for every single posting, rather than a day or weekly digest. That should give you a lot of good email messages within a couple of days.

You could create a filter to automatically move all of those messages to a dedicated folder to make it easy to ignore them. After you get enough mail to activate your personal bayes database quit the groups.

See http://www.fastmail.fm/help/spam_virus_protection.html
emoore is offline   Reply With Quote
Old 17 Jun 2010, 03:57 AM   #9
paleolith
Cornerstone of the Community
 
Join Date: Mar 2002
Location: Florida
Posts: 545
Or if you've kept a file of old good email, report those as non-spam, even if you have to upload them from your computer to FM. I'm pretty that old non-spam is almost as good as new non-spam for this purpose. (Old spam isn't as good as new spam because spam characteristics change over time.)

Using a mailing list will work to build up your numbers, but I would suggest only using it to reach 200 (if you don't have any other email you can use) and then stopping. Distinguishing spam from non-spam depends on statistical analysis of your email (spam and non), and your own personal email probably does not have the same statistical profile as what's in the mailing list. For this reason, I am currently whitelisting my mailing lists and NOT reporting them as non-spam, leaving the statistical analysis with the narrower problem of distinguishing my personal email from spam. If you use the mailing list technique, at least choose lists on topics you might correspond about in normal email.

Edward
paleolith is offline   Reply With Quote
Old 20 Jun 2010, 03:55 AM   #10
ezfig
Junior Member
 
Join Date: Nov 2003
Posts: 25
Quote:
Originally Posted by emoore View Post
Why not sign up for a couple of yahoo groups or google groups with a lot of activity where you configure it to send you a email for every single posting, rather than a day or weekly digest. That should give you a lot of good email messages within a couple of days.http://www.fastmail.fm/help/spam_virus_protection.html
All from the same sender - in other words, I get to report one non-spam per group, regardless of how many emails it sends?
ezfig is offline   Reply With Quote
Old 20 Jun 2010, 04:00 AM   #11
ezfig
Junior Member
 
Join Date: Nov 2003
Posts: 25
One other thing - I had 126 non-spams in the database before joining three Yahoo groups and quickly adding 3 more non-spam. End result: I have 126 non-spam...is there some kind of time lag in the database updating?
ezfig is offline   Reply With Quote
Old 20 Jun 2010, 04:23 AM   #12
paleolith
Cornerstone of the Community
 
Join Date: Mar 2002
Location: Florida
Posts: 545
The "spam learning" does statistics on the content of the message, so you can report as many as you want from a mailing list.

IIRC the database is updated daily, so it's not surprising that you didn't see your count increase immediately.

Edward
paleolith is offline   Reply With Quote
Old 20 Jun 2010, 11:47 AM   #13
n5bb
Intergalactic Postmaster
 
Join Date: May 2004
Location: Irving, Texas
Posts: 8,916
Arrow Fine-tuning spam filtering and address book whitelisting

There are several ways to improve your situation. Here are some points to consider:
  • Bayes filtering:
    • You can report any non-spam messages (even ones you received months before) to get to the 200 message requirement for the Bayes filter to be activated.
    • The Bayes filter maximum influence (BAYES_99, or considered 99% likely to be spam) is only 3.5. If you have a Normal spam setting, then there are other spammy characteristics of these messages which add up to at least 1.5. The only way to troubleshoot this is to look at the full headers of these messages (improperly being filed in Junk Mail) at the X-Spam-score and X-Spam-hits header. If the score is greater than 5 without the BAYES contribution, then reporting these messages as non-spam will never work. You will have to use whitelisting or increase your spam filter threshold (see below).
    • The Bayes filter only looks at the message content (as reported before), not the sender address.
  • Spam/Virus Protection settings:
    • You can use Custom spam protection to change the threshold for messages going to your Junk Mail folder. If you are using the spam score labeling feature, just look at the highest spam score on the good messages and set the Probable Spam setting to that value. If you don't get any true spam in your Inbox, then this has no bad side effects. You may need to compromise between getting a few false positives (good messages in Junk Mail) and allowing spam to appear in your Inbox.
    • If you use the Custom spam setting, I recommend using the Address Book Always accept setting. This allows address book whitelisting to be used (see more below).
    • If you are using a forwarding service(s), then place the domain name of their server(s) in the Trusted Hosts section. You must be using Custom settings to use this feature. For instance, I get messages forwarded through the ARRL Amateur Radio forwarding service, RoadRunner ISP, and AMSAT Amateur Radio Satellite organization. So I have my Trusted Hosts section set to:
      Code:
      interbridge.net,tx.rr.com,amsat.org
  • Address Book whitelisting:
    • Unless you modify the default Address Book whitelisting section of the Custom spam settings improperly (such as subtracting a very small value), then any message arriving from a sender whose address is in your online Fastmail Address Book will bypass your spam filtering.
    • Unless you check the Skip reject rules on known sender checkbox on the Reject tab of the Options>Define Rules screen, address book whitelisting will not affect reject rules you have set up. So if you want to be sure that address book whitelisting always works, be sure to check that box and then click the Apply all changes button at the top of the Define Rules screen.
    • You can add the exact email address, or a whole domain using the wildcard *@domain.com form. Just use an asterisk for the local part of the address, the @ symbol, and the domain name.
    • Address book whitelisting looks at the following headers:
      • From (provided by nearly all email clients)
      • Sender (used by some newsletter or advertising services or Sprint PCS email, but not by most personal email services)
      • X-Mail-from (the envelope From address used by the email service sending to the Fastmail server)
    • So for your particular messages, I would expect that the forwarding service might appear in X-Mail-from (depending on how they are configured), but the original source is probably in From. You can look at these yourself by viewing the Full Headers in the Fastmail message view screen (link at the upper right of the screen).
The key to getting this under control is to look at these headers in the messages which are being improperly filed in Junk Mail:
  • X-Spam-score (the spam score which triggers filing into Junk Mail)
  • X-Spam-hits (the individual components of your spam score)
  • X-Mail-from (the envelope address used by the sending server to send the message to the Fastmail server)
  • From (the sender address assigned by the email client used by the sender)
  • Sender (if it is present -- it's not there on most messages)
Please keep up the dialog in this thread. With a little initial setup work, you will be able to make this work perfectly. If the darned spammers weren't so determined to get their junk message to us, we wouldn't need to use these touchy spam filters!

Bill
n5bb is offline   Reply With Quote
Old 21 Jun 2010, 03:59 PM   #14
NumberSix
Cornerstone of the Community
 
Join Date: Jan 2003
Location: The Village
Posts: 599
OK, so.... I'm an enhanced user, I have my spam protection set to "normal", and I've wondered why a certain vendor I receive periodic mails from is constantly rated as spam even though I always choose "report non-spam" for them (their stuff is kinda spammy in nature, but since it's a company I've dealt with, it's not spam to me...)

I'm certain that soon after the personal Bayes filtering was implemented I checked that I was using it and knew that I was training my database... I should have reached the 200 mark a LONG, long time ago.... but this thread made me go back and look and.... lo and behold, for my Bayes status I have.... ZERO learned for both spam and non-spam, and using the Global database! What kinda crap is this? How did this get turned off?
NumberSix is offline   Reply With Quote
Old 23 Jun 2010, 02:50 AM   #15
Sierran
Junior Member
 
Join Date: Oct 2009
Location: Bessel Free Luna
Posts: 6
Me three

NumberSix-
I, too, just noticed that as a paid user who used to have nice high numbers on his Bayes DB status, I have reverted to Zeroes and 'Global' DB use. I also note that manually training it (reporting spam/non-spam) doesn't move those zeroes. I suspect a bug; I've filed a req. with support asking them whassup. Will report back on what they tell me.

-Sierran
Sierran is offline   Reply With Quote
Reply


Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is Off
HTML code is Off
Forum Jump


All times are GMT +9. The time now is 05:06 PM.

 

Copyright EmailDiscussions.com 1998-2022. All Rights Reserved. Privacy Policy