Fighting cheaters in online games is nothing but cat and mouse, and at the moment, many anti-cheat technologies are built on determining whether the system itself, on which the game is running, has been compromised. On a PC, these can be so-called “kernel-level drivers” that monitor the system memory for modifications that can make some changes to the normal functioning of the game. On consoles, protection is carried out at the system level, which does not allow unsigned code to start at all (at least until this system is hacked).
But lately, there are more and more cheating systems that can bypass even this protection in many first-person shooters. They consist of video capture cards, devices for “emulating data input” and software for image recognition based on the principle of machine learning, and run on a separate computer, deftly dodging the anti-cheat hunters created by developers of console and PC games, which forces them to search alternative ways.
This is a still from a video that demonstrates how computer vision software (running on an external computer) detects an enemy and calculates how much you need to move the mouse to target that enemy.
And this is the same video, but after a few frames. The device has just generated the necessary movements, and the cheat user automatically aimed at the enemy and fired at him
Another demonstration of how computer vision detects multiple targets inside the central “kill zone”
Recognition works quickly even in the presence of different obstacles and in different lighting conditions
Input emulators like the Titan Two are an important component of such computer vision cheat systems.
How it works
The set of tools used in these cheat systems is pretty simple. First, an external video capture card is used, which records what is happening in the game in real time, and immediately sends it to another computer. These frames are then passed through an object recognition algorithm like You Only Look Once (YOLO), trained to look in the image (or at least in its small central part next to the scope) for humanoid enemies.
After the enemy is detected, this cheating system easily calculates how much and in which direction you need to move the mouse so that this enemy (or even some specific part of his body – for example, his head) is exactly at gunpoint.But put them together and you have a powerful cheat system that does not need to make any changes to the software or hardware on which the game is running. In a sense, it’s like 3D printing a gun out of plastic, or assembling explosives from chemicals that you can safely buy in stores.
“Why create a bomb that can destroy the world?” The cheat developer rhetorically asked in a Discord conversation with one of our journalists. “But we created it anyway.”
Cheat systems from external tools with emulation of data input are not a new idea. But recently, this technology has attracted a lot of attention thanks to a promotional video from the creators of the cheat, which we will call CVCheat in this article (we will not indicate its real name, as well as a link to it). Many CVCheat promo videos were removed from Youtube by Activision’s copyright requirement about 4 days ago, but the most recent can be viewed on Twitter, but without any identifying information.
Ladies and gentlemen, I present you the next generation of cheating now available on console, and has been for a while but lately its been becoming more popular and more of a trend, consoles are no longer a safe space to play your games legit anymore pic .twitter.com / iEQzPVFf1h
– Anti-Cheat Police Department 🕵️ (@AntiCheatPD) July 5, 2021
The current version of CVCheat has several major automation features, including a “trigger bot” that detects if there is an enemy in sight and automatically fires a shot. In addition, among the functions there is automatic recoil compensation, which makes aiming easier by simply moving the mouse to the desired position after each shot (optical image recognition is used here – it detects the type of weapon and selects the appropriate recoil compensation settings).
User interface layout for a computer vision reader. This menu can be used to customize shots to the legs, torso or head
According to the developers of the cheat, users can play with thresholds and speeds to customize the cheat.
The size of the central “kill zone” can be changed – this will affect the speed of detection and how far the capture of the autoaim system can reach
Access to more advanced versions of the cheat opens after a monetary “donation” to its developers
But the developers of CVCheat are already preparing a new version of it and promise to bring the functionality of their cheat to a new level – according to them, it will be an “undetectable and unstoppable system with full auto-aim and auto-shots” based on computer vision, working on “any game” on PC, Xbox and PlayStation.However, a few days ago this barter option for receiving a cheat disappeared from the CVCheat website, but is still present in the Discord channel of its developers.
According to the administrator of the CVCheat Discord channel (I will call it LordofCV, so as not to reveal the name of the cheat), their product was not created in order to upset the competitive balance in online shooters. On the contrary, it is intended to “give console players a shot at games that are already teeming with hackers. Xbox players simply don’t have any chance … our script would never have been created without a request from the users! “
According to LordofCV, the upcoming version of CVCheat is able to identify the enemy on the screen and shoot at him in about 10 ms, and also works well in games with a frame rate of up to 240 fps. At the moment, the recognition algorithm “requires some adjustment” by the user, but as a target recognition threshold, you can simply set “shoot at everything that moves”.
But the algorithm still works best when the target is a large and easily recognizable shape on the screen, rather than a distant clump of pixels.
“When you get used to it, it will work great at close to medium range, but it will also work well at long range with a sniper scope,” says LordofCV.
According to LordofCV, he helped create the idea for the CVCheat toolkit and also helps with the community, while another programmer is responsible for all the scripting and donation work. Also, according to them, about 200 people are currently using the cheat.
Find and eliminate
In a conversation with journalists, LordofCV expressed absolute confidence that their cheating method is completely undetectable, because “we do not manipulate any game files at all … users use it at their own risk, but anti-cheat systems cannot catch it.”
However, at least one person involved in the protection of online games from cheaters disagrees with this statement.
“In general, the direction associated with emulation of input is not new, and the Vanguard team is well aware of it,” says Philip Koskinas, head of Valorant’s anti-cheat system. “Cheaters are constantly looking for new places to hide, but kernel drivers are not the only tool in our arsenal.”
In particular, Koskinas mentioned the 12-month ban that Riot issued in February to former Beşiktaş Esport member Yasin “Nisay” Goku. In short, the banned message states that Nisay was banned after “an automated system created by the Valorant anti-cheat team flagged his account for using a cheat that reads what was happening on the user’s screen and then using the external device emulates the corrective movement of the mouse. “So we really don’t want to cover this topic too much. ” LordofCV himself, even bragging about the “undetectability” of their system, hinted that observer players and / or analytical software can still detect superhuman speed with which their cheat aims and fires.
“The Killkam will be a problem … well, this is suspicious,” they say. “Only humans can move that fast, and this software makes everything even faster.”
But LordofCV still believes that it will be difficult for players to tell the difference between an emulated correction and the game of a real progamer – at least at first glance.
“I have witnessed situations where players played so well that they were even banned,” they say. “Most games can get you banned for no reason.”
However, it is clear that cheats that use emulation of input and computer vision will be another stage in the ongoing battle between cheaters and those who want to stop them. AI technologies continue to evolve, so it becomes easier to hide the use of such cheats, and, accordingly, it is more difficult for anti-cheat systems to detect them. The game of cat and mouse continues ..