Feeding Untangle A Diet of SPAM

[Netopia]

Greetings!

When I've setup my own Spam Filters on Linux, one of the things I was able to do was to feed spam that got through back into the filter so that it could be analyzed by SpamAssassin to help with Bayesian scoring.

Is there any way to feed spam that's gotten through back into the Untangle box so it can see where it's failing?

Joe

[dmorris]

Its trains itself automatically off the "super spam."

[Netopia]

That's fine for spam that it identifies, but that doesn't answer the question of what to do with spam that gets through. How can this be learned by the system?

There's a reason that Spamassassin has the sa-learn command. It can have false negatives and needs a way to address them.

Joe

BTW... I don't know why, but your avatar is both weird and amazing.... makes me smile.

[Netopia]

So, is the answer that there currently is no way to have Untangle learn from false negatives?

I suppose one could enable SSH, copy examples of SPAM marked as HAM to a folder on the UT box via SCP, and then run sa-learn manually.

Could I have a hint from the developers as to whether doing this would adversely affect any UT optimizations? I can't see how it could do more than simply make the filter more precise, but I thought I should at least ask.

Joe

Joe

[dmorris]

As a developer, my advice would be if you want something you can tinker and play with and turn knobs, the untangle spam solution isn't going to be a good fit for you.
Sorry to be blunt but its designed to be the exact opposite of what you're trying to do. Its meant to be hands off and self-managing.

Yes, you could use SSH and sa-learn.

IMO, however, its likely to make no difference. The "knowledge" you are contributing to the spam engine by doing this is virtually nothing compared to the huge wealth of knowledge pulled from all the services which gather global feedback, and it is entirely possible that you would actually make it less precise.

I guess it all really depends on what your goals are...

[Netopia]

As little spam as possible. That was originally my goal, and I saw what UT had to offer and wanted to give it a spin.

I actually would prefer to not have to do much or any administration to the box. Coming from my own home brew machines (using FOSS of course), it's difficult to go from often monitoring and tweaking to get every last percentage I could, to just sitting back and not really knowing exactly what's going on under the hood or having a way to control it.

If my customer is pleased with the results, then I suppose I need to sit on my hands and behave myself. Actually.... if he likes the results, I'm probably going to get him to buy some modules and I'll start thinking of replacing other units!

Really though... it's only AV/Anti-Spam that I'm looking for.

Joe