CapsuleAI is announcing the Alpha programme for Orca. Orca is a machine learning anti-cheat product (initially targeting aimbots) which maximises convenience for developers while minimising the economic burden of anti-cheat work.

Orca automatically and accurately analyses player behaviour to detect cheating, regardless of the mechanism. Orca shields developers from the treadmill of anti-cheat work, suppressing cheating in the long-term, by enabling review to scale with automated accurate reviewing.

Ineffective Anti-Cheat

Cheating is a continual source of pain for developers and players of games (us included), but current anti-cheat measures struggle to stamp out this problem. At its core, anti-cheat work requires either obstruction, which can be circumvented by cheat makers, or manual review of games, which requires a large review workforce.

Obstruction has a fundamental weakness which prohibits economical solutions to anti-cheat: developers are in an arms race with cheaters, creating a treadmill of work to secure games against exploitation.

This is compounded by new developments in cheating which cannot be feasibly tackled by obstructing systems. For example, external cheat devices will avoid any local anti-cheat system, making even locked down systems such as consoles vulnerable to cheating. Furthermore, the rise of computer vision techniques has made it viable to create AI based cheats which interact only with the screen and player input devices, avoiding anti-cheat countermeasures.

This leaves review as perhaps the only effective anti-cheat solution in the long term. But the expense of performing reviews remains, there will always be too much data to manually review.

Orca

CapsuleAI's Orca anti-cheat product is an automated review system utilising machine learning.

In the combined development and testing phase, using labelled and unlabelled real game event data from Assault Cube matches, Orca was able to achieve accuracies of up to 97.4%.

Orca is a web based service that works directly with game event data as input and returns classifications for cheating as output, maximising developer convenience. Orca also provides mechanisms to efficiently create labeled datasets.

We are now looking for partners for our Alpha programme to further validate the system in realistic contexts and ensure it meets the needs of games developers. If you would like to get involved you can get in touch here!