Multimodal game bot detection using user behavioral characteristics

Abstract

As the online service industry has continued to grow, illegal activities in the online world have drastically increased and become more diverse. Most illegal activities occur continuously because cyber assets, such as game items and cyber money in online games, can be monetized into real currency. The aim of this study is to detect game bots in a massively multiplayer online role playing game (MMORPG). We observed the behavioral characteristics of game bots and found that they execute repetitive tasks associated with gold farming and real money trading. We propose a game bot detection method based on user behavioral characteristics. The method of this paper was applied to real data provided by a major MMORPG company. Detection accuracy rate increased to 96.06 % on the banned account list.

Efficient Deep Learning Bot Detection in Games Using Time Windows and Long Short-Term Memory (LSTM)

“…Several examples of studies that use machine learning algorithms for detecting bots in MMORPGs exist with promising results in their accuracy scores [2], [4], [5], [13], [14]. For example, a study experimented with different classification architectures in order to ultimately discover a random forest technique that demonstrated the highest degree of accuracy [4]. The results from different works depend on an emphasis of precision over recall.…”

“…The results from different works depend on an emphasis of precision over recall. For example, a study has demonstrated high precision scores in detecting bots in StarCraft [4]. While high precision is important for avoiding false identification of humans as bots, this can often come at the cost of being able to detect a majority of the bot population to improve the gaming experience (i.e., high recall).…”

“…While high precision is important for avoiding false identification of humans as bots, this can often come at the cost of being able to detect a majority of the bot population to improve the gaming experience (i.e., high recall). Low recall scores obtained in some studies [4], [13] present an opportunity for improvement of these bot detection methods.…”

“…The use of statistical methods that compare player behavior to statistically common behaviors exhibited by bots has been found to be a successful strategy for bot detection [2], [3]. Self-similarity, the tendency 30 for bots to repeat the same behavior, has also been noted as a useful means of distinguishing bots from human players [4], [5].…”

“…Most of the existing bot detection methods use a large number of features. For example, when testing the number of features, a study tested predictive models using 114, 62, and 6 features and found that as the number of features used decreases, the model performance suffered, but still retained impressive results even using only 6 features [4]. Another study was able to use 56 features to detect bots with good results inside of StarCraft: Brood War [13].…”