FACEIT collabs with Google to combat AI cheats

Through a post on X, FACEIT's Director of Product Integrity, revealed the platform's collaboration with Google to strengthen their defenses against AI cheats. Amir Asdhad states that FACEIT's anti-cheat developers, data scientists and other staff spent a couple of days at Google's London office to reinforce their approach towards combating AI cheats.

AI cheats are exponentially harder to detect than traditional cheats. A normal cheating software usually manipulates entries from the game's server/memory dumps to gain access to restricted data. Upon restructuring this data into readable, critical information like player positions, the cheating software executes client side code to give the cheater an unfair advantage. However, AI cheats work in an entirely different way, almost never touching any of the game files directly, in fact AI cheats often run on an entirely different machine, making detection a big challenge.

These AI cheats use technologies like capture cards to read the player's screen and detect enemies much faster than the human eye. Upon detection, the AI cheat physically emulates mouse movements to flick the crosshair on to the visible enemy in an almost human-like sequence. Arshad reveals FACEIT's approaches towards detecting AI cheats:

• Training models on past matches where the FACEIT anti-cheat confirmed players cheated, learning to recognize similar patterns elsewhere.

• Modelling what normal play looks like to flag behavior that sits outside it.

FACEIT's only advantage in this cat-mouse chase is the sheer volume of data they posses. The Director of Product Integrity shares that FACEIT processes millions of matches every month, including years of high-level gameplay which creates a solid foundation to train models.

We're running new behavioral analysis in the next months, building an automated pipeline of confirmed-cheating data for our models to learn from, and improving how we capture more supporting data for the context of kills.
Amir Arshad, Director of Product Integrity, FACEIT

He concludes his post by pointing out the long road ahead before any of the aforementioned measures can be deployed, mentioning that FACEIT would work on behavioral analysis in the upcoming months to build an automated pipeline to catch supporting data for model training.