Automatic Player Identification in Dota 2
This addresses the issue of tracking or banning players with unwanted behavior in online games, offering a domain-specific solution for game moderation.
The paper tackled the problem of identifying anonymous players in Dota 2 by developing a machine learning approach that uses mouse movements, in-game statistics, and game strategy as a digital fingerprint, achieving 95% accuracy in predicting if two matches were played by the same player.
Dota 2 is a popular, multiplayer online video game. Like many online games, players are mostly anonymous, being tied only to online accounts which can be readily obtained, sold and shared between multiple people. This makes it difficult to track or ban players who exhibit unwanted behavior online. In this paper, we present a machine learning approach to identify players based a `digital fingerprint' of how they play the game, rather than by account. We use data on mouse movements, in-game statistics and game strategy extracted from match replays and show that for best results, all of these are necessary. We are able to obtain an accuracy of prediction of 95\% for the problem of predicting if two different matches were played by the same player.