Spam and Webmail

[Bitstream]

I've had an increase in spam slipping through the spam filter lately, so I went in to the webmail interface to see if I could, perhaps, tweak some settings. However, I can't seem to find any spam controls. I know there used to be some.

Worse, the folder list on the side bar no longer loads. It just has an error message in that frame:

ERROR: Could not complete request.
Query: CREATE "~/Trash"
Reason Given: CREATE failed: Can't create mailbox node /home/bitstream/Trash: File exists


Any thoughts?

[Xepher]

First off, the "spam filters" in squirrelmail are completely different than the site-wide controls implemented through the account interface. Also, they were broken. If you need spam filtering, using the account-management system to set it... I know it's got problems though, and doesn't work extremely well. That's being resolved by a learning-based system on the new server. I know you hate hearing "just wait" but well, just wait a bit. :-)

As for the ~/Trash error... it's a known/common problem with squirrel mail. What you need to do is go to folder/subscribe options and subscribe to the existing Trash folder. You may also need to do that for the Drafts and Sent folders if they're not already subscribed. After that, the folder bar should work properly.

[Bitstream]

That did indeed fix the webmail interface problem.

Beyond that, I don't think I've ever messed with the Account Management stuff. It says it's for Phase 2 accounts only. As far as I know, mine isn't one, so I guess it's not an option. Still, I've got -something- marking things as spam for me, so I guess it's running in some capacity. I suppose I've just been targetted by a more difficult to detect than usual spammer of late.

[Xepher]

You've got a .procmailrc file that's doing it. It looks like you set it up manually, so it filters on certain addresses. You have it passing the email through spamc, (which tags it) but no filter in place to actually delete the stuff that's spam. You basically want to add the following block right after the spamc filter.

Code Select Expand

:0:
* ^X-Spam-Flag: YES
/dev/null


Which will make it delete any email where "X-Spam-Flag: Yes" exists. Or you can just wait for the new system, which will be totally different. :-)

[Bitstream]

You've got a .procmailrc file that's doing it. It looks like you set it up manually, so it filters on certain addresses. You have it passing the email through spamc, (which tags it) but no filter in place to actually delete the stuff that's spam. You basically want to add the following block right after the spamc filter.

Yeah, I did. You helped me out with that a little actually, which I thank you for.

Actually, I don't want it to delete the spam. It's being tagged, and I have it routed in to my local Trash folder so I can review and delete as needed. I've gotten a few false positives over time, so I like the option.

The sort of settings I was looking for were ones to, in some way, make it detect spam more accurately. The system still catches more than it misses. I probably should consider myself lucky I get as little as I do, heh.

[Xepher]

Yeah... there's no real "tuning" options for spamassassin, at least not easy ones. You have to dig deep into the thing to figure out what particular tests are good "hits" for you, and then change the scoring values for each. Your best bet is looking at the X-Spam-Level flag in the headers, then you can match at some custom level above/below, rather than the simple default "Spam=yes" settings in the procmail file. Or, you can setup a custom spamassassin config file and tune anything, but as the new system is different, I think that's not worth the effort.

The new system will be ENTIRELY learning-based. It'll start off knowing nothing, and delivering mail to an "Unsure" folder. You can drag/drop (move) mail from there to "Spam" (where it will be learned as spam) or some other folder (like the inbox) where it will be learned as non-spam. As it learns, it'll start putting detected spam into the Spam folder directly, and it will be auto-deleted after 30 days. If you pull it out of there into another folder, it will correct the training. Ditto something it puts straight in the inbox that actually IS spam... just drop it in spam and it'll learn. I've been using this system for about 2 months on my personal account, and it's pretty accurate. Of course, as spammers come out with new tactics/spams, you'll probably get a few that slip through into "Unsure" but it quickly corrects itself based on your training. Usually when I get a new spam, one or two of them is all I have to train before it catches on.

I particularly like the idea of an "Unsure" folder. It provides a buffer for those near-misses, so it's a lot less likely to have a fully false positive. That means I can sort through only a few "unsures" for potential mistakes, rather than the entire spam folder. My admin email address is probably one of the worst case scenarios for spam filtering possible. It's a public address, used by hundreds of people, all in very different places in life. Thus, the emails I get range from professional inquiries from other system administrators, all the way down to requests to "plz giv me ur free site hosted thx" I get emails in other languages, with blank subjects, misspelled words, bad formatting, etc... all of which are still "legit." I've had one false positive since I started using this system, and it had more ads tagged onto the bottom by the "free" email service the sender used than most spam usually does.

[Bitstream]

The new system will be ENTIRELY learning-based. It'll start off knowing nothing, and delivering mail to an "Unsure" folder. You can drag/drop (move) mail from there to "Spam" (where it will be learned as spam) or some other folder (like the inbox) where it will be learned as non-spam.

Hrm, it sounds like I won't be able to use the spam filtering properly then, since I access my mail entirely through POP.

I've heard nothing but good things about the efficacy of systems like that though.

[Xepher]

Well... I think you must've missed the announcement where I asked if anyone still used POP3. Email access will be IMAP or Web only on the new server, specifically to implement the spam filtering. Sorry if you're just now hearing about that, I know not everyone keeps up with every forum post and all, but I couldn't exactly lock down everyones site to ask every question I thought of either, and until the verification process gets completed, I didn't have an email for most people either.

[Bitstream]

Hrm. I did miss that, indeed. I mostly just check the front page for announcements. Though I did see a preceeding thread asking about things that users use on the site. I did express my preference for POP mail there, I believe.

Skimming over posts, I suppose I can probably get used to the IMAP thing, though it does add a few layers of complication on things, at least compared to how I've been doing things.