An AI-powered aiming aid for Call of Duty gives PC and console players superhuman abilities. Is this the beginning of the end of competitive online gaming outside of controlled environments?
Activision has a vested interest in keeping its own online games like Call of Duty cheat-free: The microtransactions alone, which require regular use and thus enjoyment of the game, bring hundreds of millions of US dollars into the game publisher’s coffers every quarter – and the trend is upwards.
Too many cheaters are one of the biggest game fun dangers for a gaming community: through software tricks, they obtain superhuman abilities and thus disrupt the game flow. Activision is fighting back with corresponding vehemence, for example with legal letters to cheat developers. Nevertheless, the gaming company has to keep beating new cheat innovations in Call of Duty and its Free-2-Play version Warzone.
Cheaters on PC and console are already a problem without AI
On the PC, numerous cheat developers offer aimbots and other hacks that directly access the game files and thus let players see through walls, for example. Consoles aren’t safe either: external modules connected between controller and console allow scripts that automatically compensate for recoil, for example.
Activision regularly sends cease-and-desist letters to cheat providers, but the cheaters themselves couldn’t care less about them. Just recently, Warzone development studio Raven Software took down over 50.000 cheater accounts from the game. Other online shooters like Rainbow Six: Siege, PUBG, CS:GO, Overwatch, Valorant or Escape from Tarkov are also affected by the cheater problem.
As if these traditional cheaters, who at least have to risk being detected and punished, weren’t enough, a new generation of cheaters is now emerging that takes advantage of AI advances in machine vision. The crux: AI cheaters are currently hardly or no longer identifiable with classic detection methods.
The next generation of cheating
In early July, a cheat hunter drew attention to the "next generation of cheating" on Twitter. An AI algorithm trained on machine vision independently tracks down and attacks enemy players in Call of Duty: Modern Warfare and Call of Duty: Black Ops 3 via image analysis.
The cheat uses machine learning and sends input to your controller whenever it sees a valid target, this is aim assist but more amplified without you even needing to do anything all you have to do is aim in the general area and the machine will do the work for you
– Anti-Cheat Police Department&️ (@AntiCheatPD) July 5, 2021
Barely ten days later the website for the AI cheat tool including the Discord server disappears. All that remains is a message from the cheat developer: he has stopped developing and distributing the software at the request of Activision Publishing.
However, Pandora’s box has been opened: At least 500 times the AI tool "Userviz" (User Vision Pro) has been sold. It is only a matter of time until other providers appear or users replicate the cheat and make it available for free.
Image analysis AI botches with the view from the outside
Userviz relies on an open-source AI image analysis library to analyze the video game image in real time. The AI cheat tool can be trained for any shooter.
- MIXED.de without advertising banners
- Access to more than 9.000 items
- Cancellation possible online at any time
During training, the computer vision model learns to detect and track opponents on its own, reads weapon names off the screen, loads the appropriate anti-recoil script or automatically attacks only the legs, head or body of the opponent if desired. The accuracy and speed of the aiming aid can be infinitely adjusted, as can the size of the image section to be analyzed.
Userviz is different from previous cheater technology because the AI cheat tool does not interfere with game files. Instead, Userviz is installed on a separate PC, to which the game image from the console or gaming PC is transferred via a capture card or via the Internet.
On this separate PC, Userviz analyzes the image and sends the matching input from controller or mouse and keyboard back to the gaming device via an external module that supports image analysis.
The developer of Userviz states that his invention works on all consoles and the PC and is not recognizable for server operators: the AI sends the input signals to the game like a player does. AI cheat could help poor and mediocre players in particular to achieve better results. Outstanding players are expected to benefit less, which could change in the future with better trained algorithms.
AI-Cheats: Not only for shooters
The ability to use AI image analysis for cheats is one of the selling points of the transmission module used by Userviz. Back in April, the developer showed a video demonstrating the use of AI image analysis in NBA 2K21 for consistent scoring from all positions.
As yet, the use of such AI tools is tied to the possession of a powerful graphics card. But with ever more efficient image analysis algorithms and ever better edge AI chips, complete autonomous solutions will appear on the market in a few years at the latest.
With algorithms against algorithms
To assume that AI cheat technology can be contained again would be naive. Instead, it probably also needs AI solutions that identify AI cheaters. You could, for example, examine players’ mouse or stick movements for patterns that are barely perceptible to the human eye.
Thus, the cat-and-mouse game between cheat providers and anti-cheat programs continues with AI and is likely to lead to new tools and rapid improvements. As with deepfakes, the outcome of the race between checker and counterfeiter is uncertain, because the AI playing field is new, and there is a possibility that AI-generated input or output will one day be indistinguishable from the original. By then at the latest, competitive online gaming outside of controlled environments will have a major problem.
Steam operator and game developer Valve has already been using machine learning to detect cheaters in Counter-Strike: Global Offensive since 2018, with VACnet using data from human cheat analysts on Overwatch for AI training and combining it with information about the history of the account under investigation.
Detected cheaters are not banned by Valve: You have to play together with other cheaters. Since cheaters don’t like to play against cheaters either, this approach could at least have an educational effect.