Mark As SPAM No Longer Functioning As Expected

I have SS installed on two Macs (two licenses); on one Mac, it behaves as expected; on the second, it used to behave the same, but now it doesn’t.

When I’m reviewing mail SS marked as Spam in the Spam folder, it appears in various colors. I’m in the habit of declaring (‘Train as Spam’) anything other than Gray as definite Spam; this action generally does the following:

• Marks the item as Read
• Colors the item Blue
• Moves the message to Trash
• (Presumably adds the sender to the Block List, adds message data to file for future positive filtering)

Suddenly, it only does the following:

• Colors the message Gray
• (no idea if it is adding to Block or other filters)

I also tested the ‘Train as Spam’ action on a few messages already marked Blue; it changes it to Gray and leaves it in place.

I have started/restarted SS; verified Settings/Preferences; confirmed Rules are active; checked the FAQ.

How can I restore the expected function?

Are you using one of the approved multi-Mac setups?

That is not recommended.

I’m not sure why you were getting blue before (maybe the rule was getting re-applied?), but the “Train as Spam” command is supposed to color the message gray.

You can use the Change Settings command to control what the training command does, and the log shows what training occurred.

The two Macs do not share email accounts.

I have no idea how, but the Settings script revealed the target folder was set to Spam, rather than Deleted Messages as it was from setup until a week or so ago. Thanks for the easy fix.

I’ve reviewed the dos and don’ts article; but I’m struggling to understand why I should have to tolerate seeing the same “somewhat spammy” messages and senders over and over. Am I wrong that declaring them Spam definitely eliminates them by subsequently marking them blue (which I have set to move straight to trash)? Are you saying that if I just keep not disputing the reds, yellows, etc., that SS will eventually mark them blue and I won’t be bothered?

During and after a careful and lengthy training period, my experience in reviewing blue items prior to recently shunting to trash was that they included all the lesser spam value senders/message content I’d previously declared.

I’ve not seen a false positive marked blue (ever) or gray (in many, many weeks, possibly months), though I do see a rare yellow or red once in awhile I have to declare as Good. Again, declaring them Spam results in blue of the same.

I get 150-200 pieces of Spam every single day; I’m greatly relieved to not see a great chunk of them as they are marked blue and deleted without further review.

I’ve been reluctant to pull the trigger, but I’m super tempted to shunt gray to Trash as well. I might instead create a grays folder, and script a rule to purge items 30 days and older automatically. (Read: feature request)

That could happen if the preferences file was deleted or restored to an older version.

I’m saying that training it with every spam message is not great for the overall long-term filtering accuracy. However, you are correct that messages with the other colors will never turn blue without manual training because blue is reserved for messages that match the blacklist. Those types of messages will move to the more spammy colors, but not blue.

I’m actually somewhat surprised you’re asking about this because I don’t think it’s common to get so many messages from the same senders. Are these actual spam messages or perhaps legitimate promotional or mailing list messages that you just don’t want to see?

Assuming they are actual spam, I think what I would do in your situation is make a rule to separate the grays from the other colors. But I guess if the manual training is working for you, feel free to continue…

That’s on my to-do list.

I have two situations that open me to landing repeatedly on Spam and subscriptions lists; one is a business address that is public facing and often publicly distributed and registered to lists, sometimes without my consent. I also have to make frequent purchases from countless vendors who can be very aggressive and unscrupulous with privacy. The address is easily scraped by marketers, and while some unwanted mail looks reasonably legit and a reply or unsubscribe request would be/is respected, most isn’t. The disposable alias techniques are effective only to a degree and are arguably more work than Spam filters.

The second is that I own several original.Mac (mac.com) addresses for highly common and sought after usernames; as the service evolved, the aliases for .me and .icloud went into effect, and now any idiot who thinks he’s giving out a fake email address (verypopularname@me.com) or inexplicably believe they can just claim any iCloud address they desire, use my addresses to sign up for “free” and trial services, contests, porn and dating sites, etc, and a staggering amount of Android and iOS join-this-game requests, from around the world.

Sorting out who to trust to respect unsubscribe requests in multiple languages is not a task I choose to endure. I can waste my day trying to sort and judge which will honor an unsubscribe request, or I can block them.

If all commercial email owners were smart enough, kind enough, or legally required to validate addresses and respect Do Not Spam registries before selling them off to distribution lists, I wouldn’t need SS.

I also receive a shocking amount of unsolicited <ahem> personal selfies and desperate pleas for additional contact or requests for unpaid loans to be fulfilled from obviously used victims of a playah who thinks he’s given out a fake address; the equally sad ones are from clearly older relatives trying to get in touch with kin (again, money is often involved). These kinds of messages deserve a (usually form letter) reply explaining what has happened to them, so I’m (ethically, I guess) prevented from just deleting all misdirected mail. Honestly, I also feel compelled to protect myself legally at times.

Related, but not entirely germane, about every two-four weeks or so I see a wave of otherwise legit subscriptions and registration responses and contact follow ups (vehicle repair invoices, hotel receipts and offers, furniture and clothing purchases and satisfaction surveys, etc, etc.), followed quickly by multiple notifications of attempts to retrieve my password from iCloud and AppleID (almost certainly by a sadly confused individual who thinks they own my account name). Within a few days of this, new waves of spam assaults from related market types begin an entirely new cycle of garbage.

The ability to Block Senders was an advertised feature that prompted me to buy SS in the first place; if blocking them by declaring them Spam negatively affects the Bayesian features, please separate the functions so I can simply block repeat offenders that will either not respect unsubscribe requests or have already sold my name to countless identifiable spam services. I had disabled a long running set of rules to filter out known garbage domains and individuals; must I return to maintaining those too?

Must I throw out all the training I’ve done and start over? Is it possible to extract and preserve blocked senders, but reset the content filters (if damaged from unintentional misuse)?

I’m going to filter grays as soon as I can find the time to set it up, but it’s tedious creating rules and folders for a dozen accounts under the SS setup structure.

Insert another plea here for a rewrite of the setup and preferences interface; there are just too many instructions and user-built steps and choices for an average user to deal with, who’d like to use the more robust features but whose eyes instantly glaze over reading the advanced setups.

No one should have to manually obtain and install deep in the invisible User Library, let alone manually edit an AppleScript; multiple Scripts/functions/subroutines that use a single .plist written from a non-modal dialog, single or multiple-tab interface that can gather the needed info from Apple Mail are pretty easy to write using simple languages like Smile (is that still a thing?) or REALBasic, AppleScript Studio (is that still a thing?); I used those for this sort of thing often before I “retired.”

I don’t want to pretend to tell you your business, but I’m not sure why a single rule can’t quickly execute a script tree that splits off tasks to rate, filter and move messages from multiple accounts based on a single list of criteria. If multiple Boolean queries per message/account/filter/action/target create too much process overhead, at most a single rule for each account, still drawing from the same plist, should be able to handle Remote Training, multiple folders to scan, move to, etc. Or am I entirely misinformed?

Cheers and Thanks for what has been an otherwise highly effective weapon against my daily woes. Please accept this feedback in the spirit intended; I want to use your product in the most effective manner, without weakening its potential, and greatly appreciate your time and attention.

Would it help if you had a separate menu command to add the sender of selected message to the blocklist? If so, I could write you an AppleScript to do this. Turning off “Auto-train with incoming mail” and using that script for these kind of messages would fully separate the blacklist from the Bayesian training.

Would it help to have an option for SpamSieve to auto-create blocklist rules for all received spam messages?

I don’t think that’s necessary if you are currently getting good filtering accuracy.

The blocklist and corpus are stored separately and can be reset independently.

I’m working on it.

I don’t think what you’re describing is really feasible, but I have in mind a different solution that I think would address the same concerns, i.e. make the advanced configurations easier to set up.

Absolutely, yes, please; I think that would be very helpful, as the content isn’t necessarily spam, but the contact is definitely unwanted. The ability to assign a keystroke without third party utilities would also be essential long term, but as long as it appears in a menu, the System Shortcuts feature is sufficient. I also share with a user that is a right-click fiend (unable to commit shortcuts to memory), so hopefully it can appear there, too. A Service could suffice for the near term, if a plugin update is further down the road, so even just an accessible script would do for now.

Turning off “Auto-train with incoming mail” and using that script for these kind of messages would fully separate the blacklist from the Bayesian training.

I believe I understand what you’re saying, but am I losing out by not auto-training further? It’s been about six months of training.

Or will it still train if I manually declare as Spam or use TrainSpam remote?

Would it help to have an option for SpamSieve to auto-create blocklist rules for all received spam messages?

I think so? It might be too aggressive, though, unless it’s easily undone and false positives are easily retrieved.

I presume auto-blocked addresses would be colored and easily identified for review, but where would they go? I’d like to see manually blocked items go straight to Trash, else there’s not much point; how would auto-blocked items be elevated to ‘never show me this sender again’ status?

I’m working on it.

I don’t think what you’re describing is really feasible, but I have in mind a different solution that I think would address the same concerns, i.e. make the advanced configurations easier to set up.

Awesome. If there’s anything I can do to assist with testing or UX review, I’m happy to assist. Any approach that reduces the number of rules needed to be manually created, and automates or assists in creating/selecting folders for scanning and targeting, while providing easy access checkboxes/radio selection to enable features is absolutely welcome. Hopefully a new installer/updater package to add advanced scripts is coming as well.

I’ve lots of people who need this kind of tool, but it’s beyond them to setup and manage in its current state, and I just can’t support many people anymore.

Cheers

Frederico

Please try the new Apple Mail - Block Sender script.

You’re not missing much since auto-training is most useful in the beginning if you didn’t do much initial training.

Right. So it will still learn from any mistakes.

I’m not sure what would make sense there. Right now there’s no distinction between messages that were automatically or manually blocked.

Please send me an e-mail and I’ll get you a beta version when it’s ready to test.