Skip to content
This repository has been archived by the owner on Apr 29, 2023. It is now read-only.

wrong anti aim and aimbot detection #158

Open
bs321 opened this issue Oct 5, 2020 · 75 comments
Open

wrong anti aim and aimbot detection #158

bs321 opened this issue Oct 5, 2020 · 75 comments
Labels
bug Something isn't working

Comments

@bs321
Copy link

bs321 commented Oct 5, 2020

Like the title suggests, anti aim detection and aimbot detection is extremely wrong. Getting false positives when someone is ducking in game using ctrl key or using +duck in console in warmup.
Just noticed that I am getting false positives with aimbot as well after the last update.

@bs321 bs321 changed the title wrong anti aim detection wrong anti aim and aimbot detection Oct 5, 2020
@BeepIsla
Copy link
Owner

BeepIsla commented Oct 6, 2020

Aimbot has not been touched if I remember correctly

@BeepIsla
Copy link
Owner

BeepIsla commented Oct 6, 2020

@BlackYuzia any idea about the AntiAim part?

@BlackYuzia
Copy link
Contributor

@BeepIsla

About aimbot part - we just change death event to hurt.

hurt event has been call more times.
Maybe because that we can get more detects from clear players.

About AA part.
@bs321 can you share demo with me?

Also. Give a screenshot with current results (detects and verdict)

@BlackYuzia
Copy link
Contributor

BlackYuzia commented Oct 6, 2020

About warmup false-positive.
If that true I can return check of warmup for AS detect.

@BeepIsla
Copy link
Owner

BeepIsla commented Oct 6, 2020

Right I forgot about the death event to hurt event change, the hurt event simply triggers a lot more often, the normal aimbot detection already had 1-5 or so detections but it wasn't enough to convict, the hurt event will have a lot more of those since it triggers more often. Might need to adjust the default config.

@BlackYuzia
Copy link
Contributor

BlackYuzia commented Oct 6, 2020

@BeepIsla oh yep.
When I test aimbot detector on legit cheaters they get smaller aimbot detects like clear players. Lol.

If you ask why - they use triggerbot a lot of time.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

//False Positive for Anti Aim

@BlackYuzia @BeepIsla http://www.mediafire.com/file/6v50cq43pvz70t0/3.dem/file

https://imgur.com/a/olCm5jE

run the demo through the bot with steam id 76561198266718297

have a look at the game as well then through the csgo client

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

Please fix the aimbot detection as well. It's extremely poor right now. False positives everywhere.

@BeepIsla
Copy link
Owner

BeepIsla commented Oct 6, 2020

Fix it yourself or wait. Don't just complain.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

It's pointless for me to fix it. Your bot is falsely convicting clean players out there.

@BeepIsla
Copy link
Owner

BeepIsla commented Oct 6, 2020

Overwatch cases are viewed by many people before the suspect is banned. A couple of false verdicts are likely not going to have any impact.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

Sorry if I am being impatient but there's a possibility of someone getting convicted because same demo went to multiple people using OW bot, right?

@BeepIsla
Copy link
Owner

BeepIsla commented Oct 6, 2020

Of course its possible, but false bans are always possible with Overwatch. Its Valve's job to make sure there are no false bans and to ban people automating Overwatch or just making the bot not work.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

Right, thank you for clarifying.

@BlackYuzia
Copy link
Contributor

BlackYuzia commented Oct 6, 2020

//False Positive for Anti Aim

Okey, I has been remove check AA if warmup period is enable.
So, I get only 32 (65 - 33) false-positive results.

image

But I think this didn't full successfully fix. Kinda TYPO fix or something like this.

@bs321 so, if you duck AA detector was think what you are use AA?

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

@BlackYuzia I see. It's not only for warmup. Anytime you duck in game, it's detected as AA.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

If you see the demo, you will see that person was ducking on box on Mirage looking at A main, so that palace guy cannot see him. That's being detected as AA on the bot. @BlackYuzia

@BlackYuzia
Copy link
Contributor

BlackYuzia commented Oct 6, 2020

@bs321 76561198266718297 didn't use cheats/hacks?

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

@BlackYuzia Are you serious? Of course not. Please go through the demo.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

Play a game and like sit somewhere holding ctrl (duck) and then run the game through the bot. @BlackYuzia

@BlackYuzia
Copy link
Contributor

Yep, AA detector is suck big black dick.

I try fix that. Also I try create new method for detect AA.
I check what really detect my AA and sometimes he find some EPIC shit. 👀

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

@BlackYuzia Yes. Thank you brother. Too many false positives. Aimbot too, not sure what's wrong since last update but something's isn't right there as well.
You know how when someone is using AA their legs switch/flick, right? Hope that can help.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

or something like someone is looking somewhere else, but shoots at target?

@BlackYuzia
Copy link
Contributor

BlackYuzia commented Oct 6, 2020

or something like someone is looking somewhere else, but shoots at target?

Already. That checks aimbot detector (actualy - check prev and current angles but ... something like this. yep)
But that sometimes can have false-positive.

@BlackYuzia
Copy link
Contributor

BlackYuzia commented Oct 6, 2020

@BeepIsla tickEnd(tick) OnTickEnd(tick) was return current tick or tick+1?

I can't find some ticks.

@BeepIsla
Copy link
Owner

BeepIsla commented Oct 6, 2020

After everything within a tick has been parsed, it will fire with the current tick.

Some ticks just don't trigger this for whatever reason... Possibly because demos are 32 tick and matches are 64 tick?

@BlackYuzia
Copy link
Contributor

For example:

I can't find ticks 45076, 45114, 45284
But I can find 44973, 45285, 45283

@BlackYuzia
Copy link
Contributor

Some ticks just don't trigger this for whatever reason... Possibly because demos are 32 tick and matches are 64 tick?

yep, demo 32 tick

@BlackYuzia
Copy link
Contributor

BlackYuzia commented Oct 6, 2020

@BeepIsla so, I need add one tick (value + 1) or del one tick (val - 1) to get?

@BlackYuzia
Copy link
Contributor

If anybody want would get a demo with Legit AA Cheaters I have one of Demo

  1. 76561199068955562
  2. 76561199068893162

This both 100% use Legit AA.

@BlackYuzia
Copy link
Contributor

The legit AA, I don’t think it will help. Because in OW demo, you’re locked with one single POV and there is no way to check for legit AA. Even if a human were to inspect it, he wouldn’t know. You need to inspect it in 3rd person to give a verdict and since OW demos don’t allow it, I don’t think it is possible.

Yep, but if >50% was use this bot on one guy ... HE WILL BE BANNED 👀

@BlackYuzia
Copy link
Contributor

I am having difficulties finding demos in which there was hvh with anti aim and stuff.

You can check demo from here if you want.
I tested on this demo.

But if you find a new maybe this would be better.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

@BlackYuzia I agree but then you have to make sure that there are absolutely no false positives. Because then it kills the whole point doing it.

@BlackYuzia
Copy link
Contributor

@BlackYuzia I agree but then you have to make sure that there are absolutely no false positives. Because then it kills the whole point doing it.

I need a bit more demos with legit AA.
If you found that share with me pls.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

I am having difficulties finding demos in which there was hvh with anti aim and stuff.

You can check demo from here if you want.
I tested on this demo.

But if you find a new maybe this would be better.

Ok I am away from the rig atm. I will get home and test it.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

@BlackYuzia I agree but then you have to make sure that there are absolutely no false positives. Because then it kills the whole point doing it.

I need a bit more demos with legit AA.
If you found that share with me pls.

Yes let me get back home. Will go through the history of games and help you find it.

@BlackYuzia
Copy link
Contributor

@bs321 some note - I think demos with Rage AA required too 👀

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

@BlackYuzia http://www.mediafire.com/file/h0t7i3vucv9h78v/12.dem/file

Download this demo. Almost everyone is using legit AA namely, 76561197960909053, 76561199079155684, 76561198872552000, 76561198354741622, 76561198064550048 and 76561198180436401. I think this would be a good starting point as you can test these many players in one single game.

@bs321
Copy link
Author

bs321 commented Oct 6, 2020

I will try finding demos with Rage AA, it's difficult to find as older demos have expired and it's rare to find like a rage aa in prime mm with good trust factor.

@KillaBoi
Copy link
Contributor

KillaBoi commented Oct 7, 2020

Would it not be easy to just go into a local server, turn on Rage AA or Legit AA or whatever is required and start recording in GOTV mode and give you that instead? That's how I used to test things back when I had my stuff running... It should work and it should be easier than trying to find demos from MM games because it gives you more flexibility to change around and test whatever is needed

@BlackYuzia
Copy link
Contributor

@KillaBoi right.

@BlackYuzia
Copy link
Contributor

BlackYuzia commented Oct 7, 2020

@bs321 @BeepIsla I try create AnglesChecker for Legit AA too.
But I think this unstable.

In tests what I complete I was getting next results:

- Cheaters with Rage AA - detects <= 500 / 1000
- Cheaters with Legit AA - I can get two type of results

  • 1 type - detects >= 1000 / 1000
  • 2 type - detects <= 700 / 1000
    (idk why and how but somebody cheaters who really use legit AA didn't get 1000 detects or more) | Verdict - need more tests or change method.
    - Cheaters without AA - one guy who have cheaters get >= 1000 / 1000 detects
    - Clear Players - <= 500 / 1000 detects.

I should spend more time for legit AA I think.
Also I need saw more types of rage AA.

Note: But completed tests show what Angles Checker can detect Rage AA the best.

@BlackYuzia
Copy link
Contributor

@bs321 did you test?

@bs321
Copy link
Author

bs321 commented Oct 7, 2020

I did test using 'CSGO-Overwatch-Bot-BlackYuzia-experimental-angles-check'. It's unable to detect legit AA. The one's in which it detects, the final verdict is negative

@BlackYuzia
Copy link
Contributor

Yep. This branch can detect only rage AA.
I didn't say what this branch can detect legit AA 👀

But I think I complete preversion of legit AA detector.
I push it to the same branch a bit later

@BlackYuzia
Copy link
Contributor

BlackYuzia commented Oct 7, 2020

@bs321 if you want you can try download this branch again https://github.com/BlackYuzia/CSGO-Overwatch-Bot/tree/BlackYuzia-experimental-angles-check

I has been update config and add legit AA detector (also rename old rage AA detector)
But we should complete more tests for legit and rage detectors of AA.

BlackYuzia@cc98c3e
BlackYuzia@643e108

@BlackYuzia
Copy link
Contributor

@bs321 call me If you complete some tests.

@BlackYuzia
Copy link
Contributor

BlackYuzia commented Oct 8, 2020

LOL.
If I right we can't detect Legit AA with GoTV (or wih demo).

I check some docs how work legit AA.
In simple - he stop send packets, then he changle angle to -90 ( + 90) then start send packets and set angle back (to prev).
So clients (also gotv too) get second angle like "real" angle. But this angle are fake.

So, we can't get real angle of player.
In Rage AA we can detect when user look at floor.

I hope I did't right.

@BlackYuzia
Copy link
Contributor

@bs321 forgot about tests legit AA detector.
It would not work :c

Only Rage AA Detector. Only Hardcorecode

@KillaBoi
Copy link
Contributor

Rage AA would be fine with most people tbh, we could also pick up when the Legit AA flicks due to a bad connection or high ping because it gives you a ton of detections currently when it happens...

@BlackYuzia
Copy link
Contributor

@KillaBoi @bs321 try to use this branch - https://github.com/BlackYuzia/CSGO-Overwatch-Bot/tree/BlackYuzia-lbydelta-checker

I will check some sources how to really works AA.
In my tests I get 1K+ detects of rage AA also legit AA too. But IDK how and how all of this works 👀

Check pls because I didn't have time for check a lot of demos.

I also get false-posiitve sometimes but verdict didn't say what this user have AA.
Maybe I forgot something ... but now it is all what I want to say.

Feedback me pls if you do some tests with different players, types of AA and etc.

@bs321
Copy link
Author

bs321 commented Oct 10, 2020

@BlackYuzia I will test it, been a lil busy since past couple of days. There's a bigger issue, false positives on aimbot. It's still detecting legit player as using aimbot.

@BlackYuzia
Copy link
Contributor

About aimbot call @BeepIsla
Maybe we should return player_death event.

@BeepIsla BeepIsla added the bug Something isn't working label Nov 9, 2020
@notcreeperdude
Copy link

Just adding to this, using this replay with steam ID 76561199084564494 gives the results below, even though they are looking at the ground and getting kills. Only a 2 aimbot rating.
OW_replay9.zip
image

@BeepIsla
Copy link
Owner

BeepIsla commented Jun 23, 2021

I have adjusted the aimbot detector, it should be much better now. Note however that the default threshold in the config has changed from 1.0 to 5.0. It is also still easily bypassed by simply having a tiny bit of smoothing but whatever.

@BlackYuzia
Copy link
Contributor

@BeepIsla how possible detect legit aimbot?
I mean, did we can do that only with demo?

@BeepIsla
Copy link
Owner

You can detect static movement, aimbots typically move the crosshair in a single straight line which is detectable, some aimbots add a bit of randomness to it though which makes it a lot harder, if not impossible, to detect.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
bug Something isn't working
Projects
None yet
Development

No branches or pull requests

6 participants