Gaming bot detection

View on Springer. Save to Library Save. Create Alert Alert. Share This Paper. Background Citations. Methods Citations.

Results Citations. Figures, Tables, and Topics from this paper. Massively multiplayer online role-playing game Gold farming Multimodal interaction Virtual economy Online service provider Agriculture.

Citation Type. Has PDF. Publication Type. More Filters. A time series classification approach to game bot detection. While Chimpeon is started, everything it performs is written to a log so you can conveniently see what happened and when it happened Chimpeon is one of the safest ways to automate a game because it doesn't alter game files, memory or CPU processes. It also includes many mechanisms to help avoid detection and remain ban-free.

If you have a disability or injury, Chimpeon can ease the physical demands of playing a game by performing certain actions for you while also allowing you to continue playing yourself. It is code signed and is guaranteed to be free of third-party addons and applications, viruses and malware.

Each licence enables you to run a single instance of Chimpeon without a time restriction. Payment is made to Fly Software Limited who develop and distribute Chimpeon. If you need help with Chimpeon or you are experiencing a problem, please refer to Support before submitting. Full contact details can be found at flysoftware. Notify me about Chimpeon offers and news.

Chimpeon Overview. Features Chimpeon is packed with features that provide almost limitless game automation possibilities. Trigger Happy Chimpeon automates games by responding to Triggers. Learn More. Super Fast Pixel Detection Detecting a pixel at a fixed or variable position in the game window has never been faster. Action Hero Chimpeon can be configured to perform any of the following tasks when a condition is satisified: press a key; perform a mouse action; play a sound; send a text string; stop Chimpeon; start again.

Streaming Ready Chimpeon works alongside popular streaming solutions so games running remotely on another computer can be automated. Control Freak Status hot keys can be configured to automatically or manually turn automation processes Triggers on and off so only the ones you need at specific times are available and used. Captain's Log While Chimpeon is started, everything it performs is written to a log so you can conveniently see what happened and when it happened Undercover Operative Chimpeon is one of the safest ways to automate a game because it doesn't alter game files, memory or CPU processes.

A game bot is an automated program that plays a given game on behalf of a human player. Game bots can earn much more game money and items than human users because the former can play without requiring a break. Game bots also disturb human users because they consistently consume game resources.

For instance, game bots defeat all monsters quite rapidly and harvest items, such as farm produce and ore, before human users have an opportunity to harvest them. Accordingly, game bots cause complaints from human users and damage the reputation of the online game service provider. Several studies for detecting game bots have been proposed in academia and industry. These studies can be classified into three categories: client-side, network-side, and server-side.

Most game companies have adopted client-side detection methods that analyze game bot signatures as the primary measure against game bots. This method is similar to antivirus programs that detect computer viruses Mohaisen and Alrawi For this reason, many countermeasures that are based on this approach, such as commercial anti-bot programs, are not currently preferred.

Network-side detection methods, such as network traffic monitoring or network protocol change analysis, can cause network overload and lag in game play, a significant annoyance in the online gaming experience. To overcome these limitations of the client-side and network-side detection methods, many online game service providers employ server-side detection methods.

Server-side detection methods are based on data mining techniques that analyze log data from game servers. Most game servers generate event logs whenever users perform actions such as hunting, harvesting, and chatting. Hence, these in-game logs facilitate data analysis as a possible method for detecting game bots. Online game companies analyze user behaviors or packets at the server-side, and then online game service providers can selectively block those game bot users that they want to ban without deploying additional programs on the client-side.

For that, most online game service providers prefer server-side detection methods. In addition, some online game companies introduced big data analysis system approaches that make use of data-driven profiling and detection Lee et al. Such approaches can analyze over TB of logs generated by game servers and do not cause any side-effects, such as performance degradation or conflict with other programs. The literature is rich of various works on the problem of game bot detection that we review in the following.

Table 1 summarizes and compares various server-side detection schemes. We present key server-side detection methods classified into six analysis categories: action frequency, social activity, gold farming group, sequence, similarity, and moving path. Action frequency analysis uses the fact that the frequencies of particular actions by game bots are much higher than that of human users.

To this end, Chen and Hong studied the dynamics of certain actions performed by users. They showed that idle and active times in a game are representative of users and discriminative of users and bots. To detect game bots, Park et al. Chung et al.

Zhang et al. While this approach provides high accuracy, it is limited in several ways. First, they only focus on observations of short time window, thus they are easy to evade. Social activity analysis uses the characteristics of the social network to differentiate between human users and game bots.

Varvello and Voelker proposed a game bot detection method emphasizing on the social connections of players in a social graph.

Our previous study chose chat logs that reflect user communication patterns and proposed a chatting pattern analysis framework Kang et al. Oh et al. Our other previous work found that the goal of game bot parties is different from that of human users parties, and proposed a party log-based detection method Kang et al. This approach is however limited to detecting misbehavior in party play and cannot detect misbehavior in single play games. Gold farming group analysis uses the virtual economy in online games and traces abnormal trade networks formed by gold farmers, merchants, bankers, and buyers.

To characterize each player, Itsuki et al. Seo and Kim analyzed gold farming group connection patterns using routing and source location information. Kwon et al. This work, while distantly related, is not concerned with the detection of bots, but with understanding the unique roles each bot plays in the virtual underground ecosystem given a valid detection. Sequence analysis uses iterated sequence datasets from login to logout. Ahmad et al. Platzer used the combat sequence each avatar produces.

Lee et al. While such technique has been shown to work in the past, such feature lacks context, and might be easily manipulated by bot settings. Similarity analysis uses the fact that game bots have a strong regular pattern because they play to earn in-game money. Game bots repeatedly do the same series of actions, therefore their action sequences have high self-similarity.

Their scheme requires a lot of data of certain behavior for establishing self-similarity. Moving path analysis uses the fact that game bots have pre-scheduled moving paths, whereas human users have various moving patterns.

Thawonmas et al. They presented user clusters based on transition probabilities. To identify game bots and human users, van Kesteren et al.

Mitterhofer et al. Pao et al. They employed a Markov chain model to describe the behavior of the target trajectory. However, their feature also can be evaded and noised by adaptive bots that integrate human-like moving behavior. In this paper, we propose a game bot detection framework. We adopted some features discovered in the prior literature in confirmed in our analysis, as well as some new features discovered in this study.

We combine those features in a single framework to achieve better accuracy and enable robust detection. An additional contribution of this work is also the exploration of characteristics of the misclassified users and bots, highlighting plausible explanations that are in line with users and bots features, as well as the game operations.

Before elaborating on the framework and workflow of our method, we first highlight the dataset and ethnical guidelines used for obtaining and analyzing it. To perform this study, we rely on a real-world dataset obtained from the operation of Aion, a popular game.

Our Aion dataset contains all in-game action logs for 88 days, between April 9th and July 5th of During this period, there were 49, characters that played more than 3 h.

Among these players, characters were game bots, identified and labeled by the game company. The banned list was provided by the game company to serve as the ground truth, and each banned user has been vetted and verified by human labor and active monitoring.

In order to perform this study we follow best practices in ensuring users privacy and complying with ethical guidelines. First, the privacy of users in the data is ensured by anonymizing all personal identifiable information. One of such parties was our research group, and for research purpose only. Our proposed framework for game bot detection is shown in Fig. We posed the problem of identifying game bots as a binary classification problem.

At a high-level, our method starts with a data collection phase, followed by a data exploration phase including feature extraction , a machine learning phase, and a validation phase. In the following we highlight each of those phases.

In the data collection phase, we gathered a dataset that combines in-game logs and chat contents. We then performed data exploration in order to comprehend the characteristics of the dataset using data preprocessing, feature extraction, feature representation, exploration, and selection for best discriminating between bots and normal users. In the feature representation procedure, we followed standard methods for unifying data and reducing its dimensionality.

For example, we quantized each network measure into three clusters with low, medium, and high values using the k-means clustering algorithm. In the feature exploration phase, we selected the components of the data vectors and pre-pocessed them. For example, we determined seven activities as social interactions and quantified the diversity of social interactions by the Shannon diversity entropy.

In the feature selection phase, we selected significant features with the best-first search, greedy-stepwise search, and information gain ranking filter to avoid overfitting and reduce the features thus improving the performance.

In the machine learning phase, we choose algorithms e. We further build models using the data fed and establish baselines by computing various performance metrics. In the evaluation phase, we summarize the performance of each classifier with the banned account list provided by the game company as a ground truth, by providing various performance measures, such as the accuracy, precision, recall, and F-measure. As indicated in Table 2 , we classified the features we used in our work into personal and social features.

Given that the aim of game bots is to earn unfair profits, there is a gap between the values of the personal features of game bots and those of human users. The personal features can be also categorized into player information and actions. The player information features include login frequency, play time, game money, and number of IP address.

The frequency and ratio of these actions reflects the behavioral characteristics of game bots and human users. For example, game bots sit more frequently than human users to recover health and mana points. Moreover, a player can acquire PK points by defeating players of opposing factions. PK points can be used to purchase various items from vendors. The high ranking player can feel a sense of accomplishment. On the other hand, it is seen that game bots are not interested in rank.

In addition, there is gap between the values of the social features of game bots and those of human users because game bots do not attempt to social as humans.

The social features can be categorized into group activities, social interaction diversity, and network measures. The features of group activities include the average duration of party play and number of guild activities. Party play is a group play formed by two or more players in order to undertake quests or missions together.

The goals of party play commonly are to complete difficult quests by collaboration and enjoy socialization. Interestingly, some game bots perform party play, but the goal of party play of the game bots is different from that of human users. Their aim is to acquire game money and items faster and more efficiently. Hence, there are the behavioral differences between game bots and human users. The social interaction diversity feature indicates the entropy of party play, friendship, trade, whisper, mail, shop, and guild actions.