Bayesian filtering for Activity Streams spam

Software
1

In 2002 Paul Graham published a document called “A Plan For Spam”. A Plan for Spam It outlined techniques for using Bayesian filtering on email to separate spam from non-spam.

It’s been extremely successful.

WordPress has a service called Akismet that does Bayesian filtering for blog comments.

For identi.ca, we had a brutal spam problem. We built a spam detection server that did Bayesian filtering on Activity Streams 1.0 objects:

https://gitlab.com/evanp/activityspam

I’d like to update the service for AS2. But I need to find some servers to test with.

Are there any AP servers that enable pluggable filters on S2S incoming activities?

4 Likes
2

-1 concept NACK

I’m very much against individuals or software spamming mainnet, or releasing code to do so

Expect huge push back on this

Perhaps create a testnet with some experiments would work better

3

Pleroma: Message Rewrite Facility - Pleroma Documentation.

And here’s a thread were I was musing about cross-platform pluggable filters: Pluggable activity handlers.

2 Likes
4

Melvin, you realise this is about a tool to filter out spam, and not one to generate spam, right?

5

Thank you! I’d heard about the Pleroma milters. I’ll give em a try.

6

If its not generating spam, that’s OK

The repo activityspam was behind a login challenge, so it was unclear from your original post

Objection retracted.

7

Suggest renaming it from something like “activityspam” to “activityshield” or “activityguard”, something along those lines

Especially while it’s behind a login challenge, it makes it more clear the intention

Or make it link to a public README

1 Like