CheatDetector 0.5.2.1

[ImportBot]

Originally Posted by kcuestag*:

 

UPDATE TO VERSION 0.4.1

 

0.4.1

• FIX: Increased minimum kills needed for weapons with more headshots than kills back to 130.

Download: myrcon.net/.../cheatdetector-0521

Awesome, thanks!

[ImportBot]

Originally Posted by Singh400*:

 

Excellent, thank you all, I've compiled things neatly and pushed it through to our bug tracker and notified the team leads about what's going on, it seems to be a game code bug.

Great stuff! Cheers for that.

 

Now what we have it virtually confirmed that more headshots than kills is a game code bug when will CheatDetector be updated to exclude the above as cheating criteria?

[ImportBot]

Originally Posted by grizzlybeer*:

 

Great stuff! Cheers for that.

 

Now what we have it virtually confirmed that more headshots than kills is a game code bug when will CheatDetector be updated to exclude the above as cheating criteria?

the latest update requires 130 kills with a weapon to be analyzed, that should be ok i think. i dont see a problem with banning players those players. their stats should normalize until they reach 130 kills.

 

however i could offer several modifications if there is a need for it:

1. threat hs>kills as suspicous (not impossible) and only threat them as impossible if they exeed the average by certain value (around 270%)

 

would threat m40a5 (avg 50% hs/kills) with 130 kills and 174 as suspicious (but not impossible, 257% above avg)

 

2. add a setting to control how this is handled

[ImportBot]

Originally Posted by Singh400*:

 

the latest update requires 130 kills with a weapon to be analyzed, that should be ok i think. i dont see a problem with banning players those players. their stats should normalize until they reach 130 kills.

 

however i could offer several modifications if there is a need for it:

1. threat hs>kills as suspicous (not impossible) and only threat them as impossible if they exeed the average by certain value (around 270%)

 

would threat m40a5 (avg 50% hs/kills) with 130 kills and 174 as suspicious (but not impossible, 257% above avg)

 

2. add a setting to control how this is handled

I respectfully disagree. DarkLord has confirmed that having more headshots than kills is a game code bug. So, that shouldn't be considered as a cheating factor. You have to remember a lot of admins are insta-banning with your plugin and a lot of innocent players could be getting caught in the mix up.

[ImportBot]

Originally Posted by grizzlybeer*:

 

I respectfully disagree. DarkLord has confirmed that having more headshots than kills is a game code bug. So, that shouldn't be considered as a cheating factor. You have to remember a lot of admins are insta-banning with your plugin and a lot of innocent players could be getting caught in the mix up.

guess you are right in the end.

 

i will post a small fix to change it back like in previous versions

this will change the detection for sniper rifles (headshots per kill) to alternative 2. (some post earlier)

 

i will also reflect this change in the CheatDetector Service and see if some bans can be lifted

[ImportBot]

Originally Posted by kcuestag*:

 

guess you are right in the end.

 

i will post a small fix to change it back like in previous versions

this will change the detection for sniper rifles (headshots per kill) to alternative 2. (some post earlier)

 

i will also reflect this change in the CheatDetector Service and see if some bans can be lifted

Looking forward to this fix, we've had a lot of innocent players banned since the last update.

[ImportBot]

Originally Posted by grizzlybeer*:

 

PLUGIN UPDATED TO VERSION 0.4.2

 

0.4.2

-NEW: Settings: Enable Strict Mode (When enabled: 1 IMPOSSIBLE or 4 suspicious stats is considered cheating (=kick/ban). When set to NO: 1 IMPOSSIBLE and 1 suspicous or 5 suspicious stats is considered cheating.)

-FIX: More headshots than kills are only considered IMPOSSIBLE (or suspicious) if they exceed the average by 320% (270%). See myrcon.net/...cheatdetector-0521#entry31673.

 

Download: First Post here: myrcon.net/...cheatdetector-0521#entry31434

 

Looking forward to this fix, we've had a lot of innocent players banned since the last update.

im sorry to hear that. hopefully you find the new setting useful. if you (or anyone else) would like more detailed settings to tweak the detection, just let me know.

 

if you have some kind of list of the innocent banned player, can you send it to me? so i can unban them too.

 

Attached Files:

• CheatDetector 0.4.2.zip

[ImportBot]

Originally Posted by kcuestag*:

 

This is one of the players that complained: http://battlelog.battlefield.com/bf3/user/pakitoespain/

 

Reason: SUSPECTED CHEATING: pakitoespain SVD hs/k: 200%

 

 

I know he isn't a hacker, he's a good old friend, so I unbanned him right away after seeing it was caused by the hs/k being over 100%, which is indeed a game code bug according to the Battlelog dev who posted here.

 

I'll update my plugin to latest version, thank you for your work!

[ImportBot]

Originally Posted by grizzlybeer*:

 

This is one of the players that complained: http://battlelog.battlefield.com/bf3/user/pakitoespain/

 

Reason: SUSPECTED CHEATING: pakitoespain SVD hs/k: 200%

 

 

I know he isn't a hacker, he's a good old friend, so I unbanned him right away after seeing it was caused by the hs/k being over 100%, which is indeed a game code bug according to the Battlelog dev who posted here.

 

I'll update my plugin to latest version, thank you for your work!

thx, i dont have this player banned.

 

in the next days i will also take the time to look through some of the bans (~1700) the system issued. i will keep you guys informed with more stats soon (total players, bans, performance, unbans and such)

[ImportBot]

Originally Posted by cottton*:

 

@grizzlybeer

130 kills is imo too low.

Dunno all behind your tool but i can tell ya that ppl are padding up to 200+ kills.

To be sure i would recommend 250 - 300 kills a wpn.

 

We had a ~lot of padders appealing their bans in the past. Since we set our limit x we have no pad-appeals anymore.

just sayin =)

[ImportBot]

Originally Posted by LumPenPacK*:

 

I've removed nearly all my last metabans caused by high HS ratios with pistols since I know there's a bug with BL stats system. I'm thinking about removing all my bans because I suppose this stats bug exists from the beginning of BF3. I always saw players with high HS ratios, especially on pistols but the rest of the stats were normal.

 

That bug is really bad. I've removed lots of bans but not everybody will do that and I know now there're so many wrong bans on people who didn't cheat, even from serious metaban admins.

[ImportBot]

Originally Posted by grizzlybeer*:

 

your opinions and tips are very much apreciated. i will keep that in mind when looking through the bans.

[ImportBot]

Originally Posted by Singh400*:

 

@grizzlybeer

130 kills is imo too low.

Dunno all behind your tool but i can tell ya that ppl are padding up to 200+ kills.

To be sure i would recommend 250 - 300 kills a wpn.

 

We had a ~lot of padders appealing their bans in the past. Since we set our limit x we have no pad-appeals anymore.

just sayin =)

Yes, I'm glad someone else was thinking the same thing. I would double it to at least 300~.

[ImportBot]

Originally Posted by LumPenPacK*:

 

[00:49:43 90] Enforcing ban on BuBuBronek for SUSPICIOUS STATS: BuBuBronek RPK hs/k: 295,24%

 

:/

[ImportBot]

Originally Posted by Ike*:

 

I think you should probably just disable that stat all together, at least until it is fixed.

 

Even when its "possible" (e.g 50%) its still probably not accurate, it just shows up more in lower kill numbers

[ImportBot]

Originally Posted by grizzlybeer*:

 

[00:49:43 90] Enforcing ban on BuBuBronek for SUSPICIOUS STATS: BuBuBronek RPK hs/k: 295,24%

 

:/

weapon with less then 130 kills. fixed in last update

 

I think you should probably just disable that stat all together, at least until it is fixed.

 

Even when its "possible" (e.g 50%) its still probably not accurate, it just shows up more in lower kill numbers

hs/k gave me a solid detection in the past. the recent appeals i had where for bans by version 0.4, which requiered less then 130 kills when more hs than kills where detected.

i see that this does not work given the bug that happened at some time.

 

with 130 kills (i will probably increase this after viewing the bans) the detection should be solid.

[ImportBot]

Originally Posted by LumPenPacK*:

 

weapon with less then 130 kills. fixed in last update.

yes, I know, I've already updated my CD plugin but this was an enforced ban from another metabans admin and I don't think everybody will now remove his wrong bans.

 

It's also not your fault that BL has some broken stats. The khsr was always a very reliable indicator if somebody was cheating or not even if the newer hacks don't have this high khsr any longer....probably because the hack coders know a high khsr is too easy to detect as cheating. Maybe this is also the reason why we see all those wrong bans now on people who didn't cheat.

[ImportBot]

Originally Posted by LumPenPacK*:

 

I'm not sure if 130 kills minimum can really "fix" this problem.

 

Take a look at this guy:

 

http://bf3stats.com/stats_pc/TheLiquidFace

 

L96

 

Kills: 263 HS: 269 ---> 102.28%

[ImportBot]

Originally Posted by grizzlybeer*:

 

I'm not sure if 130 kills minimum can really "fix" this problem.

 

Take a look at this guy:

 

http://bf3stats.com/stats_pc/TheLiquidFace

 

L96

 

Kills: 263 HS: 269 ---> 102.28%

in the most recent version this player will not be banned 102% does not exeed the average enough. probably one the player i will unban (havent looked into him)

 

yes, I know, I've already updated my CD plugin but this was an enforced ban from another metabans admin and I don't think everybody will now remove his wrong bans.

 

It's also not your fault that BL has some broken stats. The khsr was always a very reliable indicator if somebody was cheating or not even if the newer hacks don't have this high khsr any longer....probably because the hack coders know a high khsr is too easy to detect as cheating. Maybe this is also the reason why we see all those wrong bans now on people who didn't cheat.

requiring more minimum kills will do the trick im sure :-)

i just want to view all the bans first.

 

running the plugin with "Enable Strict Mode" set to NO (default) will also require at least 1 IMPOSSIBLE and 1 suspicious stats.

 

working on aupcoming history check i can say that the same issues apply there (hs>kills, possibly buggy entries) but i can filter them there.

but for the stats check i can only do this manually (different data source)

 

looking at the history here http://bf3stats.com/stats_pc/jorktow...al.killassists will show you what i mean:

 

looking at the entries at 2012-05-26 19:40, 2012-05-27 12:34 and 2012-05-27 20:36 at first it appears that these are just causeed by cheats and that could be, BUT if you look closer at the times you will notice that 1 of these entries added more time than has actually passed.

meaning you generated 1 hour playtime in just 27 minutes

 

i will drop those entries (and the entries 2 weeks around) when analyzing the history entries

[ImportBot]

Originally Posted by kcuestag*:

 

I'm going through all my bans and removing the ones with hs/k above 100%.

[ImportBot]

Originally Posted by kcuestag*:

 

Ok here it is:

 

http://metabans.com/kcuestag

 

I unbanned 12 players that had over 100% hs/k. Hopefully all other admins do this as well.

[ImportBot]

Originally Posted by Ike*:

 

I'm going through all my bans and removing the ones with hs/k above 100%.

I don't understand why you would only remove the ones above 100%.

 

If the data collection method is flawed, then the entire data set is flawed, not just the ones that make you realise it.

 

The data point clearly doesn't do what it implies it does, hence the over 100% values, so you can't take actions on it based on what you think it implies.

 

Basically, you are saying someone is cheating, because they have more than (random number) 60% of their kills as headshots, but yet, if its over 100% then its fine. Where is this cut-off point? If I have 99.9% hs/k is that cheating? I mean, its possible because its less than 100.

[ImportBot]

Originally Posted by kcuestag*:

 

I don't understand why you would only remove the ones above 100%.

 

If the data collection method is flawed, then the entire data set is flawed, not just the ones that make you realise it.

 

The data point clearly doesn't do what it implies it does, hence the over 100% values, so you can't take actions on it based on what you think it implies.

 

Basically, you are saying someone is cheating, because they have more than (random number) 60% of their kills as headshots, but yet, if its over 100% then its fine. Where is this cut-off point? If I have 99.9% hs/k is that cheating? I mean, its possible because its less than 100.

You have a good point, I thought it was only flawed if above 100%, I guess I'll remove them all.

[ImportBot]

Originally Posted by grizzlybeer*:

 

You have a good point, I thought it was only flawed if above 100%, I guess I'll remove them all.

i still think that with enough minimum kills (130 may in fact not be enough, but 200-300, like cottton implied, should be fine) a number of headshots (may it be more or less than kills) is suspicious or impossible if it exeeds the average by 300%

 

iwill keep you updated

 

btw: 3 weeks holidays just started today for me

[ImportBot]

Originally Posted by kcuestag*:

 

i still think that with enough minimum kills (130 may in fact not be enough, but 200-300, like cottton implied, should be fine) a number of headshots (may it be more or less than kills) is suspicious or impossible if it exeeds the average by 300%

 

iwill keep you updated

 

btw: 3 weeks holidays just started today for me

Then what the hell are you doing here? Get off the forums!

 

I removed all hs/k bans in the mean time, not like there were many, hehe.

[ImportBot]

Originally Posted by Ike*:

 

i still think that with enough minimum kills (130 may in fact not be enough, but 200-300, like cottton implied, should be fine) a number of headshots (may it be more or less than kills) is suspicious or impossible if it exeeds the average by 300%

If I have 2 kills, and both of them are headshots, my hs/k may be reported as 150%, if one of the kills counted for 2 headshots and the other for 1 (in whatever way the system works)

 

On the other hand, if I have 130 kills and it shows up as 60%, then officially I have had 78 headshots. (I dont know the % value you use, and am using 60% as an example)

 

While a person cheating may have this value, I may have actually only got 39 headshots, all of which have counted for 2 headshots.

 

Meaning I have only actually got a 30% value, which is "not cheating".

 

As far as I am concerned, one innocent player banned is one too many. I would let 100 people cheat before I banned an innocent player.

 

That said, thats my personal opinion, its your plugin, and not my servers using it. You can do whatever you see fit, just trying to give some of my perspective.

[ImportBot]

Originally Posted by cottton*:

 

@grizzlybeer

keep it simple. You have no control what data you get, but you have control of what your tool does.

So, it ~doesnt matter what happen to the source (where you get the wpn data from). But it matters what you do with it.

 

In short: data-check before player-check =)

 

if($khsr > 100){$khsr = 100} # there is nothing else you can do. you wont be able to "calc" the real khsr.

 

if($khsr > $x and $kpm > $y){...} # always get the kpm compared to the khsr. in best case you calc if the kpm is possible on that khsr.

 

btw - we have a API. If the traffic is not that huge we can give you our expected-kpm. (we would share almost everything except the sourcecode for the expected kpm - that was too much work )

[ImportBot]

Originally Posted by grizzlybeer*:

 

ok. viewed all 1700 bans.

i will unban 184 (this may not be the number you will see on metabans, because not all are banned there. some just local)

 

here is the list if you want to compare it with your lists or unban these players: http://pastebin.com/ea76U0h6

 

almost all were caused by "more headshots than kills". most with less than 130 kills.

 

so the solution is:

1. increase minimum kills to ~200

2. do what cottton said: if(hs>kills){hs=kills}

 

expect an update sometime soon (way more testing this time :-))

 

im sorry for version 0.4 (and possibly 0.4.1). the changes (less min kills if hs>kills and considering hs>kills as cheating not taking the avg into account) to the detection caused most of the false bans.

 

pls make sure you are using version 0.4.2

 

@grizzlybeer

keep it simple. You have no control what data you get, but you have control of what your tool does.

So, it ~doesnt matter what happen to the source (where you get the wpn data from). But it matters what you do with it.

 

In short: data-check before player-check =)

 

if($khsr > 100){$khsr = 100} # there is nothing else you can do. you wont be able to "calc" the real khsr.

 

if($khsr > $x and $kpm > $y){...} # always get the kpm compared to the khsr. in best case you calc if the kpm is possible on that khsr.

 

btw - we have a API. If the traffic is not that huge we can give you our expected-kpm. (we would share almost everything except the sourcecode for the expected kpm - that was too much work )

yes i was wondering how you calc those. tried everything but i couldnt get it right. thx :-)

[ImportBot]

Originally Posted by kcuestag*:

 

Sounds good, thank you for working on this even on your holiday.

[ImportBot]

Originally Posted by cottton*:

 

Caution - if you change the actual data.

I recommend to set the kshr to 100 instead of changing the actual data (hs to kills).

Only change "your" data and show the 3rd party as it is.

In that case the kshr is yours (you calculated it), and the 3rd party (of curse) the kills and hs.

 

yes i was wondering how you calc those. tried everything but i couldnt get it right. thx :-)

Took ~3 month and 2 ppl to get it working. Its still not perfect.

 

We could try it out. I mean i guess every layer with your tool would send a request á player so we have to test it first with a few (traffic ect).

Just pm me any time if you wanna start some testing.