Quote from: pixie85 on July 15, 2018, 06:33:17 PM

I think that it would be a great idea as long as you wouldn't be able to cheat the bot by spamming reports with your army of accounts to nuke people.
<>

Quote from: LoyceV on July 15, 2018, 06:56:28 PM

<...>
I can think of several ways to sabotage this:
-A spammer can post the same post as a real user, hoping that user gets banned too
-A spammer can easily randomize his post enough, so that a bot won't detect it
-A bad guy can trick an innocent user to post a homograph attack. The user won't even know he did it.

The idea that I had in mind is that when you report a spam post, Its down to the moderator/admin to include the text from the post in the spam list that the cop-bot searches for. That action performed by the moderator/admin is manual and down to their criteria of what a spam post is. Posts should get added to the spam list on a post by post basis. That is, if a profile is categorized as a spammer, only those posts that are individually considered spam by that user should be added to the spam list (not all automatically per se). This should in theory nullify deliberate abuse.

If a spammer decides to copy a post from a real user in order to get him banned, it should not have a wrongful effect if all is properly implemented. The above paragraph vouches for adding spam posts to the contrast spam list on a post by post basis, so if the spammer copies a proper users post to get him banned, it should not work (if the post itself is not spam).

The randomizing part is the potential issue. The idea is to go after spam that is of a similar nature. If a spammer makes up a different post each time, having each post not enough similarities to any post in the spam list that is being constructed, the he will be in the clear from automatic detection, being it down to manual reporting again for such cases.

The general idea is to start to create automatisms that help admin/mods and people that report spam, by kicking in some logic that should evolve and get more complex overtime. Unless something already exists on the market that can easily fit in with the forums software, creating something that evolves from within by studying the spam cases and extracting common factors and patters, could help ease the burden to some extent.

I wonder if the MindlessElectron account is related to this type of approach somehow