Valve uses 1,700 CPUs to catch CS:GO cheaters with VACnet

Valve are using 1,700 CPUs to catch CS:GO cheaters with VACnet

Valve uses 1,700 CPUs to catch CS:GO cheaters with VACnet

Popular online games fight a continuous battle against those who desire to cheat the system, making anti-cheating tools invaluable to both developers and gamers alike. Players don’t want to be beaten by cheaters and developers want to remove toxicity from their online gaming environment, today cheating has become of the most common complaints levelled at online games, be it CS:GO, Fortnite or PUBG. 

Valve’s John McDonald noticed that in 2016, while the programmer was between projects, that the CS:GO community was constantly talking about cheating. This lead McDonald to start researching deep learning for Valve, eventually creating an addition to their VAC system called VACnet, which works alongside CS:GO’s Overwatch system to detect cheaters. 

While some of you will be despairing at the thought of Valve replacing their human moderated VAC (Valve Anti-Cheat) system with AI, be aware that this is not a replacement for the company’s existing VAC system. VACnet is designed to submit cases to a human moderator for review, with a success/conviction rate of 80-95%, showcasing a high degree of confidence in the system. Player/human submitted cases have a 15-30% conviction rate, making the VACnet AI better at detecting cheaters than the average human.  

Valve’s existing VAC system will work alongside VACnet, which itself is an extension of their Overwatch system, learning more about cheaters as it is used to become more accurate and adapt to newly discovered cheats. Since the introduction of this system cheating in CS:GO has become rarer, creating a more positive online community. 

Valve are using 1,700 CPUs to catch CS:GO cheaters with VACnet

 

To power this new system, Valve uses a server farm with 1,700 CPUs, doubling this number to give them further room for expansion in the future. Valve wants to bring this system to other online games, including non-Valve releases, allowing other developers to benefit from Valve’s VACnet AI.   

Deep Learning and AI have the potential to revolutionise the gaming industry, allowing programmers to create new tools which will make the lives of gamers a lot easier moving forward. Catching cheaters is just one application of Deep Learning, which when combined with human moderators allows for higher cheater conviction rates without any wrongful bans when the AI submits honest players to human moderators. 

You can join the discussion on Valve’s use of their VACnet AI to catch CS:GO cheaters on the OC3D Forums.