Demographic profiling from MMOG gameplay
This paper examines profiling basic demographic information (gender and age) from the gameplay of 1040 World of Warcraft (WoW) players. The authors develop two monitoring systems to track the players, one based on in-game observation and the other on a data source provided by the operators of the game. We describe and extract four feature sets, each from different assumptions regarding the type and amount of data available to an adversary: 1) a one-time snapshot of each character, 2) a series of snapshots from which we extract features for character progression, 3) a mapping of players to characters that allows us to extract higher level features over all the characters belonging to a player and 4) a superset of the previous three sets. We show that one can predict gender and age (within 5 years) for 53% of players using machine learning and one can predict gender and age (within 1 year) for over 11% of participants solely based on the features monitored by our systems.
Likarish, P.; Brdiczka, O.; Yee, N. Demographic profiling based on MMOG play. Fourth Hot Topics in Privacy Enhancing Technologies Workshop (HotPETs); 2011 July 27-29; Waterloo, Canada.