Does marking junk mail in Thunderbird train external filters?

Question

If I take the trouble of marking mail als junk/spam, I want to maximize the effect of that effort.

Does marking emails as spam only train my local filters, or are signals/patterns send out to other (Mozilla controlled) servers? That would benefit other people as well.

[If the effect is only local, I will switch to marking those mails as spam in Google webmail - and then they won't be downloaded over POP3 either]

Answer 1

Does marking emails as spam only train my local filters, or are signals/patterns sent out to other (Mozilla controlled) servers?

Short answer: Marking messages as junk with Thunderbird's spam filter only affects local messages. There are no upstream benefits.

Longer answer:

Just for context, there are some nuances here:

• Does Thunderbird actively do anything to affect spam beyond what it sees locally?
• Could what Thunderbird does provide indirect upstream benefits to you?
• Could what Thunderbird does provide indirect upstream benefits to others?

And for additional context, you frame this relative to the results of using Google webmail, which would imply a focus on just upstream benefits to others.

I'm not sure if you're interested in only that piece of this, so I'll address the whole picture (in not quite the same order).

Does Thunderbird actively do anything to affect spam beyond what it sees locally?

Thunderbird's spam filter deals only with its local messages. It comes with a few generic starter rules and then learns about your personal preferences from both what you mark as spam and what you clarify is not spam. These preferences are irrelevant to anyone else. You might love getting email that others consider spam and vice versa.

As far as I know, Mozilla is not an email service provider and is not connected with the blacklist side of things. Regardless, Thunderbird does not send the results of your training back to Mozilla, or anywhere else, to either modify Thunderbird's starter rules or to feed any kind of upstream spammer identification. The latter wouldn't be worth the effort, anyway, because spammers change their game too fast for that to be useful.

Could what Thunderbird does provide indirect upstream benefits to others?

Although Thunderbird does not actively feed the upstream anti-spam machine, could its results be captured and used for that purpose? There are two stages of upstream spam filtering that could affect others, blacklists and the email-provider-level filtering. Email client applications don't feed, and aren't a source for, the blacklists. That leaves email-provider-level filtering.

Email providers each have their own system (and I'm not aware that these competitors share), so there is not a one-size-fits-all answer. You mention Google. Google does data mine user's rules as one source for their provider-level filtering:

Gmail automatically helps identify spam and suspicious emails by detecting viruses, finding patterns across messages, and learning from what Gmail users like you commonly mark as spam or phishing.

Source

However, Gmail uses a rule-based filter, which works differently from Thunderbird's learning filter. The rules you create in Gmail are known to Google and in a form Google can directly use. Gmail can't tap into what Thunderbird learns. Gmail also can't learn by observing what Thunderbird does, as described next.

Could what Thunderbird does provide indirect upstream benefits to you?

Your Gmail user account has its own, personal spam filter. Could it learn from what Thunderbird learns? The answer is No, for a number of reasons, including:

• You mention using POP3. With POP3, Gmail would never see what Thunderbird does because POP3 just downloads the messages and then Thunderbird does its thing locally. However, the default protocol for Gmail is IMAP. If that applies to you, read on.
• Gmail's filter is rule based. There is no utility for extracting what Thunderbird has learned, and converting that to rules Gmail could use.
• Gmail can't learn by observing what Thunderbird has done because it doesn't operate in that way (it isn't a learning-type filter).
• Even if it was capable of learning, it wouldn't learn by observing after the fact. Filtering is done either on arrival of the message or when you manually take action via Gmail's web app. Once message disposition has occurred, Gmail doesn't recheck messages it has already dealt with.

Definitive proof

There are endless links to support the bits and pieces in this answer, but it's difficult to prove a negative. There's nothing I can find to cite that specifically documents the bottom line of Thunderbird not providing upstream benefits.

However, it is easy to demonstrate whether Gmail's user account filter (and therefore, potentially its provider-level filter), is affected by what Thunderbird does. I don't use Gmail, so I can't do this exercise, but you can prove it to yourself as follows:

• Use some disposable email account and send yourself a message.
• Mark it as junk in Thunderbird.
• Close Thunderbird.
• Send yourself another message from the same account.
• See if Gmail flags it as spam.

That will give you a definitive answer.