Reversing Anti-Cheat's Detection-Generation Cycle With Configurable Hallucinations

David Durst, Carly Taylor

Anti-cheat and cheat developers are locked in a cat-and-mouse cycle of detection and generation. Anti-cheat developers struggle to detect cheating players' behavior. Cheat developers easily evade the detections by generating different behavior.

Anti-cheat developers may reverse this cycle with hallucinations. Hallucinations are ingame entities only visible to cheating players. Cheat developers must detect them, or cheating players will react and self-identify. Hallucinations' key property is configurable ingame behavior. Anti-cheat developers can respond to the detections by generating different behavior. Hallucinations' configurability reduces the anti-cheat problem to the bot design problem. As game developers create bots that better imitate humans, they can reuse these behaviors to make hallucinations harder to detect.

In this paper, we define hallucinations; discuss prior, non-configurable implementations; explain our implementation; and provide evidence that popular cheating techniques treat our hallucinations as real players.

This whitepaper was originally presented at GDC in 2023.

Read the whitepaper

The Software License and Service Agreement will be updated. Please follow this link [https://www.activision.com/legal/ap-eula] in order to see these changes.

Privacy Policy Update
We’ve updated our Privacy Policy. You can view the revised policy here. By continuing to use Activision’s websites, products or services, you acknowledge this revised Privacy Policy.