Topic on User talk:GethN7

It's Virus Bot. I'm making it to patrol edits for vandalism or anything else we don't want. I will start by having it notify sysops for suspicious activity and once it gets reliable we can think about using it as a sysop to issue short bans to users who make bad edits.

I see. Please confer with the other sysops for ideas on hows to best implement this bot, because I'd like their opinions before handing out a bot flag.

It's just... I wonder what you mean by suspicious activity. This seems like a good way to start a Thought Police. It makes me nervous to hand out bans automatically -- it's exactly the kind o thing that could drive users away. I have no problem with you using the API to make recommendations, but it's really worrying to see an account requesting sysop privileges on the day after it was created.

It's going to do stuff like compare the byte count and if it's too low it would be flagged as abuse. LulzKiller already the process of adding rights to the bot. I don't expect to get a sysop bot now and I'd find it unwise if those rights were handed out so easily,

We can discuss algorithms the bot can use. It could look for signatures of known vandalism like certain HTML or copy/paste messages or look to see if a user is deleting or adding a large amount or similar content from pages.

The problem is that low byte count is not by any measure a good metric to determine abuse. I make a lot of changes that can weigh in at +/-20 bytes, just as a matter of tweaking layouts, taking care of spelling errors, correcting redlinks caused by incorrect capitalization, or fixing unescaped "="s embedded in quote templates.

Our most recent vandalizations have almost universally inserted text that looks like comment spam into pages while deleting hundreds or thousands of characters. Others have blanked pages outright. That's the kind of thing we should be looking for. The main problem i can see, though, is how to distinguish legitimate large deletions -- say, splitting up a Trivia page that has items for two or more works into individual pages -- from vandalism. And the opposite, detecting spam pages and distinguishing them from legitimate new pages.

There's more then one approach we can take. I'd look for very large deletions and anything that reduces the size to 0 should automatically be flagged. I don't mean ban the user automatically without other flags corroborating the likelihood of vandalism. We can look at info like what user groups the user is a member of or account age so a confirmed user is more trusted then a new user.

We could look for behavior like a single message being copied across multiple pages in a short amount of time.

This should have been my regular account, not the bot. Not having global bot passwords is a real pain and caused this when I had to log out to generate them on other wikis.

Labster not to interject however solid did say it was suspicious activity alerting first and then if it was good enough, then sysop privileges. Solid wants the bot designation/flag today, and the sysop at a later date if the bot was successful. A bit nitpicky but it's important to clarify.

Interject all you want, friend. Anytime someone brings up sysop on a new account in any context makes my mod senses tingle. That's all. Now if the bot was OSS, that would go a long towards making my fears go away.

I could post the source code for the bot if that's what you're asking. I don't expect a sysop account at any point in the near future.