IW4M-Admin/Plugins/Stats/Cheat
RaidMax 943808562f fix error code page for things over than 404s
allow request token when not logged in
2020-01-14 18:56:23 -06:00
..
Detection.cs update values for snap and offset 2019-10-07 10:26:07 -05:00
DetectionPenaltyResult.cs update some anticheat code 2019-09-27 15:53:52 -05:00
README.md
Strain.cs
Thresholds.cs fix error code page for things over than 404s 2020-01-14 18:56:23 -06:00

IW4MAdmin Anticheat

Initial document draft | 8.30.19

IW4MAdmin anticheat for IW4x uses data from in-game logs to track player locations, hit locations, and view angles to validate against a known set of play styles. Every hit event ocurring in the game (Damage or Kill from a gun) is captured by IW4MAdmin and analyzed. Session analysis occurs against all hit events since the player connected to the server. Lifetime analysis occurs against all hit events ever occuring for a given player.

Detection Types

Bone

Compares the number of times a particular bone is hit against the number of hits on all other bones. Many rudimentary aimbots lock onto a particular player bone position such as Head, or Upper Torso. This detection method has the highest chance of a false positive, as non-cheaters can play abnormally (e.g. going for headshots, or using a weapon that unconciously changes their playstyle)

Hit Location Reference

Number Hit Location
2 Head
3 Neck
4 Upper Torso
5 Lower Torso
6 Upper Right Arm
7 Upper Left Arm
8 Lower Right Arm
9 Lower Left Arm
10 Right Hand
11 Left Hand
12 Upper Right Leg
13 Upper Left Leg
14 Lower Right Leg
15 Right Foot
16 Left Foot

Chest

Identical to Bone detection, except it specifically compares the ratio of Upper Torso / Lower Torso hit counts. It can be thought of as a focused bone detection type, as most aimbots don't aim to unusual bones such as a foot. As with Bone detection, this is prone to false positives.

Offset

Compares the player angles from three snapshots. The first snapshot being the snapshot immediately before the server registered the hit for a player. The second snapshot is the "frame" the server registered the kill. The final snapshot is the snapshot immediately after the server registers the hit. The algorithm is:

let a = first snapshot angles
let b = second snapshot angles
let c = third snapshot angles
offset = ((a - b) + (c - b)) - (a-c)

This detection method is very effective at detecting silent aimbots. Silent aimbots "fake" a shot by changing a player's view angles for a single snapshot. From a spectator's position, the single snapshot view angle altering is not visible. The larger "FOV" (field of view) from a silent aimbot, the more absolute difference between the three snapshots. Over time if the average distance between these two viewangles is higher than expected, a detection is triggered. False positives are very rare with this detection. However, extreme client lag can trigger a false positive in special situations.

Strain

Analyzes the frequency, viewangle distance, and player distance between hits. The algorithm is:

let v = view angle distance between two hits  
let t = delta time (time between hits)  
let d = distance between the attacker and victim
let s = decay over time
strain = ((v / t) * d)^s

A high value indicates fast target switching at large distance intervals, which if done naturally, requires an extreme level of mechanical effort. "Rage" aimbots commonly switch to targets as fast as possible in any direction.

Recoil

Compares the average of the last few snapshots of player angles to 0. Client sided no recoil prevents the view angle "Z" axis from changing. If the average view angle "Z" axis value remains 0, the detection is triggered. As of 8.30.19, there are no known false positives for this detection. Several weapons which do not have any recoil from the game are excluded in this detection.

Format

Detection penalty reasons are as follow: <detectionType>-<location/value>@<hitCount> Example:

  • detectionType = Strain
  • value = 1.39
  • hitCount = 136

Result: Strain-1.39@136 This reason is only visible to logged in privileged users.