Microsoft developed an AI to catch Xbox Live cheaters

Microsoft recently filed for a patent for an AI that detects Xbox cheaters by combing through their gaming history for suspicious ranks or achievements.

The patent application was published by the US Patent and Trademark Office last month. Unlike other efforts by the likes of Valve, this AI wouldn’t just look for cheaters within active games. Instead, it would snoop through the interactions between game and platform, such as “notifications of achievements, game scores reported to the platform, and/or a player rank achieved on the platform based upon game activities.”

Inspecting things like achievements and rank might give a more blunt-force approach to detecting cheaters than just watching them play. As Microsoft points out in the patent, sometimes cheating isn’t made apparent within the game itself:

As such, a platform that hosts third party games may not be able to detect cheating that occurs in third party games, even where achievements in third party games are managed at the platform level.

On the other hand, if you’ve only played four or five hours worth of play in a game and you’ve got an achievement that typically takes ten or more to get, that’s a bit of a red flag. It wouldn’t be caught by in-game cheat detection, but Microsoft’s AI would catch it.

Microsoft aren’t the only ones attempting to catch cheaters with AI: Valve has for years been working on a system that detects cheatbots in games like CS:GO. Comments from a spokesperson on Reddit about the complexities of the system might give a decent explanation for why Microsoft doesn’t rely solely on real-time monitoring to catch cheaters:

The process of parsing, training, and classifying player data places serious demands on hardware, which means you want a machine other than the server doing the work. And because you don’t know ahead of time who might be using this kind of cheat, you’d have to monitor matches as they take place…

From that perspective, it makes sense Microsoft might try to look through profiles and other, less demanding forms of data.