Aimbot and Triggerbot Detection

[divVerent]

I added hitplot writing to Nexuiz. With this, you can after about 100 frags find out whether someone is using an aimbot/triggerbot or not.

Here are some typical hit plots:


A human. Not very skilled though


The bots that come with Nexuiz. Very human-like (they are coded to be that way), but the pattern looks more regular than a human one. Humans get distracted e.g. by specific properties of the player model which causes some bias in the hit plot, as seen in the human one. Bots only see hitboxes, and tend to make a regular pattern.


An aimbot that focuses on head shots. No further comment


A trigger bot. A trigger bot is a bot that does not change the aim, but controls the fire button whenever the crosshair is above an enemy. These are hard to see when watching a demo, but in the hit plot there's just too many hits near the edge of the hitbox. Why is that? Because when you move the mouse over a player, the trigger bot shoots as soon as you first touch the player - which is near the border of the hitbox.

[alpha]

Triggerbot very insignificantly differs from human IMO.

Still, tell us how to use it? I have 3 whole nexuiz clans I would like to check with this.

[divVerent]

Very significantly does. Look at the red blobs at the left and the right.

[C.Brutail]

Omg, you're insane
You still have too much free time on your hands (fortunately )

[GreEn`mArine]

daaaaammmnn

[Bundy]

Oh no, Alpha will catch us all

[12:35][12:30] <alpha``> hurrrrrrrrrrrrr
[12:31] <alpha``> looks like i'll be catching some botters real soon
[12:31] <alpha``> http://forums.alientrap.local/viewtopic.php?t=4745
[12:32] * Spalter is now known as spalter
[12:33] <vociferous> ohh now that is awesome
[12:33] <all-bundy> nice
[12:33] <all-bundy> start with my demos please
[12:33] <all-bundy>
[12:34] <vociferous> finally we can catch kojn =)
[12:34] <morphed_> alpha``: it should have bigger exposure, maybe giant text before you enter the game "we can now detect botters" that way you will catch more ppl
[12:35] <all-bundy> the colors are beautiful and stuff .. but i dont understand it -.-
[12:35] <morphed_> why do you want to undestand it ? :>
[12:35] <alpha``> vociferous exact.y
[12:35] <vociferous> hehe
[12:35] <alpha``> kojn, green marine, bundy, revenger
[12:35] <all-bundy> cause it looks interesting
[12:36] <alpha``> I can't wait
[12:36] <all-bundy> hrhr, alpha`` will be so disapointed to proof
[12:36] <all-bundy> that we dont cheat

[GreEn`mArine]

so, how does it work in practice? I assume it will be included in 2.5.2? Then it's a serverside-setting that generates these images per player? Which intervals? Per Map?

[alpha]

div, i'm still waiting for a manual on how to use it.

Also, you devs should set up a level of sufficient hitplot evidence that would justify punishment.

[divVerent]

It unfortunately cannot analyze demos - it requires you to do g_hitplots 1 on a svn server.

It then writes .plot files for each player, which can be matched to eventlog entries as they use the same match ID and player ID as eventlog uses.

The analyzer tool generating these pictures is hitplot2tga.c in Nexuiz svn. Usage: hitplot2tga plotfile.plot imagefile.tga.

As you see in the post above, I updated the tool a little, to show blue pixels for each exact hit location, and to do better clipping of values > 1.

You need to record about 100 shots of the same player before you can even think about using this method of analysis. A single match often does not contain enough data. The name of the player is recorded at the top of the .plot files, so you can use that to collect data per player.

Unfortunately, demos cannot be analyzed by it. But once player input is moved to csqc, this feature can be added. Until now it has to be enabled on a server.

Also, for now a human eye is required to see if someone is botting. Maybe some automated testing can be added in the future (especially for triggerbots).

[divVerent]

I managed to change my trigger bot to evade this method of analysis.

It however also made the trigger bot much less effective. A noob with this changed bot will still be below average. For a good player, I'd expect it to increase frags by about 20%.

Another method of catching trigger bots - especially if a noob is using them - is watching how they aim. If they slowly swipe the crosshair over a player, and once it hits the player, it shoots, I think you should know what's going on

The best way to use a trigger bot is using a bad and ineffective one like I made for testing, and be NOT aware of using it[1]! But if you're a noob, you won't beat kojn with that... actually, I am sure I wouldn't beat kojn with ANY trigger bot, but if kojn wants, we could try that out

[1]: if you are aware of a cheat you're using, you will change your behaviour accordingly. Aimbotters tend to not even roughly aim, but just hit fire and let the bot do the rest. Trigger botters tend to slowly swipe the crosshair over enemies - which looks the more obvious, the weaker the trigger bot is. Wallhackers tend to track enemies that went behind a wall with the crosshair. It may be easy to throw off ways to detect a cheat (like this method), but doing so has to reduce accuracy of the cheat. So, such methods being available cannot prevent cheating, but they do reduce the effectiveness of cheating.