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.
Pages: 1
was based on open-sourced ATD, but now the detection methods are completely different from its predecessor-ATD. More numerical values are used to determine Trolls&&Art.
{//including:
Degree of Random placements =Complete=
Block placements speed =Complete=
User movements speed =Optimizing=
Bot?Kills users when they enter to check whether they are Bots =Bug found=
Straight Line? =Bug found=
How Curve does the user draws a line? Implementing...
Better way to determine whether a user is filling something. Pending
}
But then there's still a number of false detection cases. Hence, I need your help ? cuz Data Collection does not 100% help.
====HOW YOU COULD HELP===
STEP1. Find a World Called <EE TROLL CATCHER> in the game lobby {NOTE: THE WORLD is not 24hrs Online}
STEP2. Start Recording Your screen.
STEP3a. Find the loophole of EETC and TROLL!!!!!!* or...
STEP3b. Draw and Build Stuffs normally, see if there's any false detection.
STEP4. Stop Recording... Upload Your Awesome Video to Youtube or another places.
STEP5. Please Link the Video to here Thanksss
=== COMPLETE CHALLENGE ===
*It is not easy to complete step 3a, try your best. I'll credit you if you do a great job!
Offline
False detection when users draw fast... (i believe trollers are a bit faster, unless they are laggy and get detected before they can even ruin anything)
Offline
Today, I was able to slowly turn someone's special art thing into a lime green mush. It appears that the detection is based on how quickly someone is drawing; I think that there are more variable to consider such as what color they are drawing, and have less tolerance for colors that look like they don't match and less tolerance for coins. Being able to get rid of a vandal's edits but preserving the underlying artwork (instead of deleting it) would be a nice addition. There are many cases that trollers do not move and can spread a lot of blocks quickly so the movement algorithm might need to be tuned.
Drawing a straight line can circumvent the curve feature, unfortunately. But overall it looks to be an ambitious project--I hope that it fairs well!
Offline
Today, I was able to slowly turn someone's special art thing into a lime green mush. It appears that the detection is based on how quickly someone is drawing; I think that there are more variable to consider such as what color they are drawing, and have less tolerance for colors that look like they don't match and less tolerance for coins. Being able to get rid of a vandal's edits but preserving the underlying artwork (instead of deleting it) would be a nice addition. There are many cases that trollers do not move and can spread a lot of blocks quickly so the movement algorithm might need to be tuned.
Drawing a straight line can circumvent the curve feature, unfortunately. But overall it looks to be an ambitious project--I hope that it fairs well!
Hmm, what about counting variety of blocks overlapped, when a player draws over others work, +1 for every kind of blocks he overlapped.
Offline
Hexagon wrote:Today, I was able to slowly turn someone's special art thing into a lime green mush. It appears that the detection is based on how quickly someone is drawing; I think that there are more variable to consider such as what color they are drawing, and have less tolerance for colors that look like they don't match and less tolerance for coins. Being able to get rid of a vandal's edits but preserving the underlying artwork (instead of deleting it) would be a nice addition. There are many cases that trollers do not move and can spread a lot of blocks quickly so the movement algorithm might need to be tuned.
Drawing a straight line can circumvent the curve feature, unfortunately. But overall it looks to be an ambitious project--I hope that it fairs well!
Hmm, what about counting variety of blocks overlapped, when a player draws over others work, +1 for every kind of blocks he overlapped.
What if you were changing the color of your own work though? The problem with bots like this, is that you can only make them search for people causing problems, there's no way to make them search for those not causing problems, and keep them safe from being detected.
Trolling is a form of building, just like making art or minis is a form of building. Everything involved in trolling, is involved in making art or minis as well. You can't simply create a bot that will differentiate the two, when they are essentially the same.
There will always be issues and overlaps where the bot will detect someone who is not causing harm. The only way to fix this would be to remove some things from the bot, which in turn would cause the opposite problem; people who are causing harm wouldn't be detected.
Offline
goeyfun wrote:Hexagon wrote:Today, I was able to slowly turn someone's special art thing into a lime green mush. It appears that the detection is based on how quickly someone is drawing; I think that there are more variable to consider such as what color they are drawing, and have less tolerance for colors that look like they don't match and less tolerance for coins. Being able to get rid of a vandal's edits but preserving the underlying artwork (instead of deleting it) would be a nice addition. There are many cases that trollers do not move and can spread a lot of blocks quickly so the movement algorithm might need to be tuned.
Drawing a straight line can circumvent the curve feature, unfortunately. But overall it looks to be an ambitious project--I hope that it fairs well!
Hmm, what about counting variety of blocks overlapped, when a player draws over others work, +1 for every kind of blocks he overlapped.
What if you were changing the color of your own work though? The problem with bots like this, is that you can only make them search for people causing problems, there's no way to make them search for those not causing problems, and keep them safe from being detected.
Trolling is a form of building, just like making art or minis is a form of building. Everything involved in trolling, is involved in making art or minis as well. You can't simply create a bot that will differentiate the two, when they are essentially the same.
There will always be issues and overlaps where the bot will detect someone who is not causing harm. The only way to fix this would be to remove some things from the bot, which in turn would cause the opposite problem; people who are causing harm wouldn't be detected.
First, blocks overlapped == blocks variety overlapped && Not build by themselves.
Second, out of topic, but you would be able to distinguish two accurately by machine self-learning abilities, but certainly i don't have the relevant equipment and also knowledge. Ask those in the IBM.
hence, what i'm doing to avoid false positive detection is by distinguish whether the user is doing "art", or even .trust them to bypass the bot
Offline
If humans can make the difference between art and trolling then bots can do the same.
Yeah, it will not work right....
Everybody edits, but some edit more than others
Offline
Well, I don't actually suppose that one may simply write "that one bot that will prevent trolling". I assume so, as you will have for any of your detections either some heuristics or at least some thresholds. In fact though, you could easily leverage them to at least create a utility which might come in handy at some point of development. You see: You could define several "rules" whose thresholds may be changed in a configuration file. By doing so you would enable the users of your bot to try out and experiment what effects different violation levels of the rules you implemented in your bot would have which effects on the gameplay in general. Thus your users could determine on their own how far they'd like to take it with troll prevention. Just for clarification, here's one stupid example for demonstration:
Let's assume you had a rule which always checks on how many blocks a player places in a specific time span. If the value surpasses a configurable threshold it would throw a violation which your bot could then handle accordingly.
The configuration might look like this:
# The number of blocks a player may set in x seconds:
rule.brick.place.threshold=50
By combining multiple rules now (e.g. how distant those blocks are placed from each other - useful for making a distinct detection of whether a player is building those "troll lines" we all know) you could effectively narrow down the set of possible trolling interactions. Only if defined (maybe event configurable) criterias would be fulfilled you would then assume a player to be a troll. In case you know the Minecraft Plugin "NoCheatPlus" - this is effectively what it does in order to prevent hacking: giving its users the ability to sort out whatever they don't like on their servers.
Hopefully you got a rough understanding of what I'd suppose.
Best regards,
EuadeLuxe.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
C++, C#, Java, PHP, JavaScript and Python programmer
Enjoys developing all day
Don't believe it? Then verify me here, https://keybase.io/verify, and here: https://www.gommehd.net/members/blackypaw.32820/ (open the Informations tab and you'll find another PGP signature)
-----BEGIN PGP SIGNATURE-----
Version: Keybase OpenPGP v2.0.46
Comment: https://keybase.io/crypto
wsBcBAABCgAGBQJWCo8vAAoJEG4qS86NYi+iL1kH/Aq8bRm7tjZAqeBE4YBR0bJx
ao40xBbFsQ1KjZUnUgQosFP66H0JqlGmrJnyseyBLM+xccrkXapXN/T6csKTTzMy
dFV6i55+Jn/r1Y45+7yZHSYYZeG+OsFRin+Yvj0R69U6fuGLzDhfd9DFFlj0ORUw
qScy5/A0v+F9mLi+4a/fp9io4oF2kHiqzFNGCy4TueKgU5N+Um8UvhkNonLvRyvD
kOpjym8co4VLEDLfTVJgQAUhmmwW1NHvyaDDj9hjW4ERmhuX7ITwtrVdyAG6VhJB
DEN9SjkSTbGnvWepCv2xnqKmvXwyCXrds4d6Tt/wOz+794Optk8paYT/IO2RM2w=
=nW2J
-----END PGP SIGNATURE-----
Offline
Well, I don't actually suppose that one may simply write "that one bot that will prevent trolling". I assume so, as you will have for any of your detections either some heuristics or at least some thresholds. In fact though, you could easily leverage them to at least create a utility which might come in handy at some point of development. You see: You could define several "rules" whose thresholds may be changed in a configuration file. By doing so you would enable the users of your bot to try out and experiment what effects different violation levels of the rules you implemented in your bot would have which effects on the gameplay in general. Thus your users could determine on their own how far they'd like to take it with troll prevention. Just for clarification, here's one stupid example for demonstration:
Let's assume you had a rule which always checks on how many blocks a player places in a specific time span. If the value surpasses a configurable threshold it would throw a violation which your bot could then handle accordingly.
The configuration might look like this:
# The number of blocks a player may set in x seconds: rule.brick.place.threshold=50
By combining multiple rules now (e.g. how distant those blocks are placed from each other - useful for making a distinct detection of whether a player is building those "troll lines" we all know) you could effectively narrow down the set of possible trolling interactions. Only if defined (maybe event configurable) criterias would be fulfilled you would then assume a player to be a troll. In case you know the Minecraft Plugin "NoCheatPlus" - this is effectively what it does in order to prevent hacking: giving its users the ability to sort out whatever they don't like on their servers.
Hopefully you got a rough understanding of what I'd suppose.
Best regards,
EuadeLuxe.
ohh i get it, I will add a tab like the chrome's flag page later on
Offline
i really like this bot!
Offline
Well, I don't know whether you actually got my point. In one sentence: "Work by connecting several small evidences in a user-defined manner into a cohesive detection network." I personally can't image what chrome's flag page has to do with that.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
C++, C#, Java, PHP, JavaScript and Python programmer
Enjoys developing all day
Don't believe it? Then verify me here, https://keybase.io/verify, and here: https://www.gommehd.net/members/blackypaw.32820/ (open the Informations tab and you'll find another PGP signature)
-----BEGIN PGP SIGNATURE-----
Version: Keybase OpenPGP v2.0.46
Comment: https://keybase.io/crypto
wsBcBAABCgAGBQJWCo8vAAoJEG4qS86NYi+iL1kH/Aq8bRm7tjZAqeBE4YBR0bJx
ao40xBbFsQ1KjZUnUgQosFP66H0JqlGmrJnyseyBLM+xccrkXapXN/T6csKTTzMy
dFV6i55+Jn/r1Y45+7yZHSYYZeG+OsFRin+Yvj0R69U6fuGLzDhfd9DFFlj0ORUw
qScy5/A0v+F9mLi+4a/fp9io4oF2kHiqzFNGCy4TueKgU5N+Um8UvhkNonLvRyvD
kOpjym8co4VLEDLfTVJgQAUhmmwW1NHvyaDDj9hjW4ERmhuX7ITwtrVdyAG6VhJB
DEN9SjkSTbGnvWepCv2xnqKmvXwyCXrds4d6Tt/wOz+794Optk8paYT/IO2RM2w=
=nW2J
-----END PGP SIGNATURE-----
Offline
Okay, I like to say that this results in false positives a large number of times, but I think it's just got really good at detecting swastikas.
However, testing this today was something I couldn't do because I was trusted from the one time I didn't draw a swastika and as it turns out, begging two admins to untrust me and making about twenty swastikas for the sheer purpose of proving my former statement (in addition to seeing how long my edit was taken away) doesn't work.
tl;dr: get better admins who don't ban people for dumb stuff and actually pay attention, bot works pretty good, please reset the time after a day (I think I was up to 1920 seconds -- that's half an hour), still does a bunch of false positives.
proc's discorb stylish themes for forums/the game
꧁꧂L O V E & C O R N꧁꧂ ᘛ⁐̤ᕐᐷ
danke bluecloud thank u raphe [this section of my sig is dedicated to everything i've loved that's ever died]
?
Offline
Okay, I like to say that this results in false positives a large number of times, but I think it's just got really good at detecting swastikas.
However, testing this today was something I couldn't do because I was trusted from the one time I didn't draw a swastika and as it turns out, begging two admins to untrust me and making about twenty swastikas for the sheer purpose of proving my former statement (in addition to seeing how long my edit was taken away) doesn't work.
tl;dr: get better admins who don't ban people for dumb stuff and actually pay attention, bot works pretty good, please reset the time after a day (I think I was up to 1920 seconds -- that's half an hour), still does a bunch of false positives.
Yeahh sry, the bug listed above aren't fixed due to my laziness :3 ~~ which greatly affect the detection =.= anyway lol i did not aim for detecting swastikas but thx for pointing it out ... which soon add a d!..k detection cuz kids love to draw them(maybe their favorite)?
Offline
It detects me so many time and i don't do anything but drawing conners of world.
Everybody edits, but some edit more than others
Offline
Offline
If humans can make the difference between art and trolling then bots can do the same.
Yeah, it will not work right....
Well, there's a few straightforward telltale signs... but there's always going to be a gr(a|e)y area. That's what I don't like about this bot-- sadly, it can never be perfected. Computers don't mold their definitions with what they see... (rather, not in this level of programming)
good luck
but seriously, don't go overly crazy with this; there's always going to be a loophole... e.g., every time I join your world I erase peoples' work.
or do I?
Your bot doesn't know.
as a testament to my laziness,
behold my lack of conformity
YOU SHALL NOT STEAL MY PASSWORD, HACKERS!!!
but really though
I don't know how to get back into my old account
and I don't care. (yes this is hz5)
Offline
Zumza wrote:If humans can make the difference between art and trolling then bots can do the same.
Yeah, it will not work right....Well, there's a few straightforward telltale signs... but there's always going to be a gr(a|e)y area. That's what I don't like about this bot-- sadly, it can never be perfected. Computers don't mold their definitions with what they see... (rather, not in this level of programming)
good luck
but seriously, don't go overly crazy with this; there's always going to be a loophole... e.g., every time I join your world I erase peoples' work.
or do I?
Your bot doesn't know.
I notice he doesn't use a strategy he could use to fix people erasing:
Detect people that erase blocks that aren't theirs
(he could put block[x,y]... after X,Y and ID variables and put following code below that, to avoid mistakes)
if ((block[x,y].placer != names[m.GetInt(4)]) && BlockID==0)
{
//Detection code
}
Offline
If humans can make the difference between art and trolling then bots can do the same.
Yeah, it will not work right....
Welcome in the world of signal and image processing: You are wrong. Not technically. But still.
Offline
Pages: 1
[ Started around 1732422552.8807 - Generated in 0.102 seconds, 12 queries executed - Memory usage: 1.72 MiB (Peak: 1.96 MiB) ]