Research on the Application of Artificial Intelligence in Online Game

Anti-Cheating

Lehan Li

Institute of Problem Solving, Chongqing University of Posts and Telecommunications,

Chongwen Road,

Nan'an District,

Chongqing, China

Keywords: Anti-Cheating, Game Security, Artificial Intelligence, Online Games.

Abstract: In recent years, the popularity of online games has made anti-cheating systems particularly important.

Although the traditional anti-cheating system has formed a relatively perfect system after years of

development and iteration, with the update of hardware and software and the popularity of AI, cheating

methods are also iterating rapidly. The use of Artificial Intelligence (AI) and Direct Memory Access (DMA)

to cheat has become a popular option for cheaters, while traditional anti-cheating systems have little effect.

This paper reviews the application of artificial intelligence in anti-cheating in online games, analyses the

advantages and disadvantages of existing methods, and looks forward to the future development trend. By

using artificial intelligence technology, game developers and researchers can more efficiently analyse the data

in the game to determine whether players cheat, and can also assist the anti-cheating process by analysing

players' text information. Therefore, combining artificial intelligence with anti-cheating can more effectively

detect and prevent cheating in online games, to maintain the fairness of the game and the good experience of

players.

1 INTRODUCTION

Online games have become increasingly popular in

recent years, and many companies' businesses

revolve entirely around their online games and the

consistent revenue generated accordingly. If there are

a large number of cheaters in a game, it will seriously

affect the ecology of the game and the revenue of the

company, but also cause great damage to the

reputation of the game and even the reputation of the

company.

Anti-cheating system is helpful to provide better

game experience for players. After years of

development and iteration, the traditional anti-

cheating system has formed a relatively perfect

system to deal with cheating in online games.

However, with the update of hardware and software

and the popularity of AI, cheating methods are also

iterated rapidly, and the use of AI and external

equipment DMA cheating has become a general trend

of cheating development, and for these new cheating

means, the effect of traditional anti-cheating systems

is minimal. In this context, game developers and

https://orcid.org/0009-0004-8363-7345

researchers began to experiment with artificial

intelligence technology to detect and prevent cheating.

Around the world, many researchers have begun to

explore the use of artificial intelligence technology to

combat cheating in online games. For example, in

their study in 2013, Alayed and Frangoudes et al used

machine learning technology to analyze the data

transmitted between the client and the server,

calculate the eigenvalue, and judge whether the user

has cheating behavior through the calculated results

(Alayed et al., 2013). In a 2022 study, Kanervisto and

Kinnunen et al designed a deep learning approach to

control a computer mouse to enhance a player's level

of play and investigated its effectiveness against

cheating (Kanervisto et al., 2022). In their 2022 study,

Tao and Xiong et al proposed a workflow called

GAXI, this approach leverages the capabilities of

multi-view data sources while also benefiting from

the clarity and transparency offered by multi-view

black box models, and has good results in game

cheating detection on real-world data sets (Tao et al.,

2022). In his 2017 study, Maguluri proposed a

general automatic player classification method based

106

Li, L.

Research on the Application of Artiﬁcial Intelligence in Online Game Anti-Cheating.

DOI: 10.5220/0013234500004558

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management (MLSCM 2024), pages 106-111

ISBN: 978-989-758-738-2

on artificial intelligence technology and natural

language processing to classify players as

cheaters/victims/neutralists based on the text of their

statements on social platforms to aid the anti-cheating

process (Maguluri, 2017).

These studies show that artificial intelligence

technology has broad application prospects in the

field of anti-cheating. However, with the continuous

evolution of cheating methods, anti-cheating systems

are also facing new challenges. In the future, anti-

cheating systems will need to continuously

incorporate the latest artificial intelligence

technology to deal with increasingly sophisticated

cheating, maintain the fairness of the game and the

good experience of the players.

2 OVERVIEW OF CHEATING

TECHNIQUES

In online games, cheating usually refers to the

behaviour of players to gain an advantage over their

opponents or achieve certain goals by modifying the

game or using other improper means (

Yan, & Randell,

2005

). These cheating behaviours not only destroy the

fairness of the game, but also may seriously affect the

game experience of other players and the reputation

of the game company.

In their research on the classification of cheating

in online games in 2005, Jeff and Brian mentioned the

definition of player cheating: players who violate the

rules set by game operators and gain more advantages

than other players through illegal means are regarded

as cheating (Jeff & Brian, 2005).

2.1 Traditional Cheating Methods

In the game, every object (including characters,

buildings, props, etc.) is a 3D model generated by a

computer, and these models have a memory store,

which includes the location of the model, state and

other information.

Traditional cheats are usually done by reading the

game's memory, such as:

 Perspective cheating: By modifying or reading

the game's memory data, the player can "see"

information that is not normally visible. For example,

by reading the position information of other player

models, it is possible to "see" each other from behind

a wall, which is called "perspective".

 self-aiming cheating: by reading the data in the

memory, obtain the position information of other

players, and then through calculation, automatically

adjust the player's aiming Angle and shooting time,

so that players can easily hit each other.

In the classification study on cheating in online

games in 2005, Jeff and Brian sorted out 15 types of

plug-in categories according to the potential

vulnerabilities, consequences and cheating principles

(Jeff & Brian, 2005), which almost completely

showed the ways and harms of traditional cheating

methods.

2.2 Novel Cheating Methods

With the continuous development of technology, the

continuous update of hardware and software, and the

popularity of AI, cheating methods are also in rapid

iteration. Among them, the use of AI cheating and

DMA cheating has begun to take shape, and the

inability of traditional anti-cheating methods has

caused the proliferation of AI cheating and DMA

cheating in today's online games, which has a serious

impact on the game ecology, company reputation and

revenue.

2.2.1 Using AI to Cheat

AI cheating uses artificial intelligence techniques

(such as machine learning, deep learning, etc.) to

achieve game cheating. Compared with traditional

cheating methods, AI cheating is more intelligent and

hidden. The principle is as follows:

 Image recognition: By capturing the game

screen, AI can identify enemies, items and other

information in the game according to the trained

model. This method does not directly read the game

memory, more difficult to detect by the anti-cheat

system.

 Decision and operation: AI makes decisions

(such as moving, aiming, shooting, etc.) according to

the identified information, and then implements

cheating functions by simulating player operations

(such as simulating keys, mouse operations, etc.).

This mode of operation is closer to real player

operation, reducing the risk of detection.

2.2.2 Using DMA to Cheat

DMA cheating is a way to realize game cheating

through hardware devices. DMA cheating equipment

can obtain or modify game data by directly accessing

game memory, so as to achieve cheating effect. The

principle is as follows:

 Hardware devices: DMA cheating requires

specialized hardware devices, such as PCIe devices

and USB devices. These devices can access computer

Research on the Application of Artiﬁcial Intelligence in Online Game Anti-Cheating

107

memory directly without the intervention of the

operating system.

 Memory access: DMA devices read or modify

game data (such as player position, aiming Angle,

etc.) by directly accessing game memory to achieve

cheating effect. Because of the hardware-level access,

it is difficult to be detected by the anti-cheat system.

 Drivers: In order to realize the interaction

between the DMA device and the game, a special

driver is also needed. These drivers generally need to

bypass the operating system's security mechanisms in

order to be able to access protected memory areas.

3 ANTI-CHEATING METHODS

3.1 Traditional Anti-Cheating Methods

In order to maintain the good ecology of online games,

the anti-cheating system is one of the unattainable

links. The anti-cheating system in online games refers

to a set of technologies and measures aimed at

detecting, preventing and combating game cheating,

the main purpose of which is to maintain the fairness,

stability and game experience of players.

Traditional anti-cheating methods mainly rely on

the monitoring, analysis and processing of game data

and player behaviour to identify and combat cheating,

including the following principles: abnormal data

detection, behaviour pattern recognition, client-side

integrity verification, plug-in program detection,

server-side verification, blacklist system.

In a 2020 study, Lehtonen analysed the common

anti-cheating methods used in online games today,

and rated the anti-cheating methods based on tamper-

proof, ease of implementation, lack of overhead, non-

intrusion, and suitability for various games (Lehtonen,

2020).

3.2 Artificial Intelligence

Anti-Cheating Methods

As cheating methods continue to evolve, traditional

anti-cheating methods may not be effective in some

cases against new types of cheating such as AI or

DMA. Therefore, modern anti-cheating systems often

combine artificial intelligence technology to improve

the accuracy and real-time detection of cheating.

3.2.1 Use Artificial Intelligence Technology

to Efficiently Analyse Data and

Identify Cheating

Alayed et al. proposed a method by setting

characteristic values to data, using artificial

intelligence technology to analyse the data

transmitted between the client and the server,

calculate the characteristic values, and judge whether

the user has cheating behaviour through the

calculated results.

From the experimental data of the author in

Table1, it can be seen that the analysis of data

between client and server based on artificial

intelligence method has a high detection rate for

cheating methods such as Lock, Auto-Switch, Auto-

Miss, Slow-Aim and Auto-Fire.

The anti-cheating method proposed by Alayed et

al. effectively uses artificial intelligence technology

to better and more efficiently analyse the data

between the client and the server to determine

whether players’ cheat. However, the core of this

method is still to analyse the normal transmission of

data in the game, which is not good for the detection

of cheating methods that do not use traditional data

transmission such as AI dual machine /DMA cheating.

3.2.2 Use Artificial Intelligence Technology

to Analyse Coordinate Information

Aimed at Players to Determine

Cheating

Willma proposes an RNN-based deep learning

algorithm (Willman, 2020) that analyses the

coordinates that a player is aiming at to determine if

an Aimbot is being used to cheat.

Table 1: Accuracy results for Part 4 (Alayed et al., 2013).

Cheat Type Best Selected Results

TPR for L

TPR for AF TPR for Normal Frame Size

Classifier

L 96.6% 2.4% 100% 60 SVM-RBF

AS 100% 0% 100% 60 Both

AM 89.7% 0% 100% 60 SVM-RBF

SA 89.7% 0% 100% 60 Both

AF 55.2% 100% 96.9% 60 SVM-L

MLSCM 2024 - International Conference on Modern Logistics and Supply Chain Management

108

The anti-cheating method proposed by Willman is

no longer confined to the analysis of the transmission

data between the client and the server, and opens the

idea of artificial intelligence in anti-cheating methods.

Unfortunately, it is mentioned in the paper that the

target accuracy of the algorithm should be over 90%

in order to truly identify whether a player is cheating

with Aimbot (Willman, 2020), but the author of the

paper only achieved an accuracy of slightly over 50%

due to time and data set limitations. Whether this

method can be used in existing online games still

needs to be studied.

Pinto et al. introduced sophisticated anti-cheating

methods utilizing artificial intelligence that are very

flexible yet substantially dependent on in-game data.

Consequently, each game must incorporate an

adaptive anti-cheating system(Pinto et al., 2021).

Pinto et al., therefore, propose a new cheat

detection method that does not rely on in-game data

(pinto et al., 2021). The approach initially analyzes

the multimodal interactions between the player and

the platform as multivariate time series. Subsequently,

it employs a CNN network to classify these time

series and ascertain if the game player has engaged in

cheating behaviour.

The contribution of the article can be summarized

as follows:

 New techniques for detecting cheating in video

games only based on input data, such as

keystrokes and mouse movements, which are

derived from player behaviour.

 CNN structure to detect cheating in a

supervisory manner.

 Create data collection and processing methods

for human-computer interaction multivariable

time series restatements.

Kanervisto et al. designed an AI method (GAN-

Aimbot - GAN-generated adversarial network) to

control a computer mouse to enhance the player's

level of play, and studied its effectiveness against

cheating. At the same time, an automatic cheat

detection mechanism (VACNet) built using DNN

networks is also introduced, which can automatically

detect cheating players by detecting whether players

cheat only through mouse movement coordinate data

(Kanervisto et al., 2022).

The experimental results in this paper show that

the test results of Aimbot designed based on GAN are

much better than those of ordinary Aimbot, and the

automatic cheating detection system built based on

DNN network has better performance on both

training set and test set.

It can be seen that whether it is cheating or anti-

cheating, artificial intelligence technology can bring

more obvious improvements.

3.2.3 Construct AI Workflow for

Anti-cheating

Tao et al proposed the concept of XAI (eXplainable

AI) and also proposed the workflow of GXAI (Tao et

al., 2022). As shown in Figure 1, this process

combines the power of a multi-view data source with

the clear transparency of a multi-view black box

model.

This study introduces four distinct classifiers and

interpreters that analyse character portraits, behaviour

sequences, client images, and social graphs. It can be

seen from the author's experimental data in the table

2 that XAI has a good effect on anti-cheating

detection in four dimensions.

Tao and his team's approach, as discussed in their

paper, has been applied and utilized in three real-

world applications in NetEase games. These include

generating evidence and reasons, debugging and

testing models, and compressing and comparing

models. User studies have given it highly favourable

feedback. This confirms the practicality of using AI

in combating cheating in online games.

3.2.4 Use Artificial Intelligence Technology

to Analyse the Player's Speech Text to

Distinguish Cheating Players

Maguluri studied various natural language processing

and artificial intelligence methods as well as text

classification tools in his 2017 article, and then

proposed a general automatic player classification

method to distinguish cheating players (Maguluri,

2017).

The main contribution of this paper is the design

of a general automatic player classification method

based on natural language processing and artificial

intelligence methods to classify players as

cheaters/victims/neutrals based on the text of their

statements on social platforms to aid the anti-cheating

Table 2: User studies for evidence and reason generation (Tao et al., 2022).

Job PortraitExplainer BehaviorExplainer ImageExplainer GraphExplainer

Game cheating detection

Game operator teams 95.9% 90.1% 96.4% 81.1%

Game designer teams 83.1% 76% 88.6% 68.3%

Research on the Application of Artiﬁcial Intelligence in Online Game Anti-Cheating

109

process.

In a 2019 paper, Hughes et al. applied natural

language processing (NLP) tools to automate post

types Automatic classification of post types (Hughes

et al., 2019). It also creates a systematic data

processing framework for analysing large

unstructured forum data sets.

The main purpose of the method proposed in this

paper is to classify users through their statements on

social media and analyse whether the users are likely

to commit crimes in real life. This method can also be

used for anti-cheating in online games. By setting the

labels of game cheaters, analysing the players'

speeches in game forums, and classifying cheating

players, it can also assist the anti-cheating process.

Figure 1: Proposed GXAI Workflow (Tao et al.,

2022).

4 CHALLENGES AND

PROSPECTS

In recent years, artificial intelligence technology has

made remarkable achievements in the field of anti-

cheating in online games. However, with the

increasing renovation and sophistication of cheating

methods, AI technology still faces many challenges

and opportunities in future anti-cheating applications.

 More application of artificial intelligence

technology: Artificial intelligence technology has

made breakthroughs in image recognition, natural

language processing and other fields, and is expected

to play a greater role in anti-cheating online games in

the future.

 Introduction of reinforcement learning

technology: Reinforcement learning technology may

become an important means of anti-cheating in online

games in the future. By analysing game data in real

time, reinforcement learning models can

automatically detect potential cheating behaviour and

adjust according to the cheater's strategy to achieve

dynamic anti-cheating. In a paper published in 2022,

Lukas et al used reinforcement learning techniques to

build an anti-cheating system (Lukas et al., 2022) that

could distinguish between robots and normal players

in a game, but did not apply the method to anti-

cheating in online games. Lukas's approach opens up

the idea of using reinforcement learning techniques in

online game anti-cheating, which can be used in the

future to build an online game anti-cheating system

and further strengthen the learning ability of anti-

cheating systems to combat cheating methods that are

difficult to combat with conventional means such as

AI or DMA.

 Integration of cross-field technologies: The

application of artificial intelligence technology in

anti-cheating online games will continue to expand,

involving more related fields. For example, the

integration of big data analysis, image recognition,

natural language processing and other technologies

will help improve the comprehensiveness and

accuracy of cheating detection.

 Personalized anti-cheating strategies: With the

development of artificial intelligence technology,

future anti-cheating systems will be more intelligent

and personalized. Through in-depth analysis of player

behaviour and game data, the anti-cheating system

can provide customized anti-cheating strategies for

different types of players, improving game fairness

and user experience.

5 CONCLUSION

This paper first introduces the difference between

traditional cheating methods and new cheating

methods, and then compares and analyses the

difference between traditional anti-cheating methods

and the new era anti-cheating methods using artificial

intelligence. This paper explains the important role of

artificial intelligence technology in dealing with the

modern and novel cheating methods in online games,

and introduces in detail the different ways of applying

artificial intelligence technology in the field of anti-

cheating and the efforts and results made by

predecessors in this direction.

At the same time, the paper also puts forward the

future prospect of the research content, which can

seek breakthroughs in the application of more

artificial intelligence technology, the introduction of

reinforcement learning technology, the integration of

cross-domain technology and personalized anti-

cheating strategies, improve the performance of

artificial intelligence technology in the field of anti-

cheating in online games, and effectively combat the

MLSCM 2024 - International Conference on Modern Logistics and Supply Chain Management

110

currently difficult to solve AI/DMA cheating

methods. Optimize the game ecosystem and protect

the interests of players and game companies.

REFERENCES

Alayed, H., Frangoudes, F., & Neuman, C., 2013, August.

Behavioral-based cheating detection in online first

person shooters using machine learning techniques.

In 2013 IEEE conference on computational inteligence

in games (CIG) (pp. 1-8). IEEE.

Kanervisto, A., Kinnunen, T., & Hautamäki, V., 2022. Gan-

aimbots: Using machine learning for cheating in first

person shooters. IEEE Transactions on Games, 15(4),

566-579.

Yan, J., & Randell, B., 2005, October. A systematic

classification of cheating in online games.

In Proceedings of 4th ACM SIGCOMM workshop on

Network and system support for games (pp. 1-9).

Willman, M., 2020. Machine Learning to Identify Cheaters

in Online Games. URL http://urn. kb. se/resolve.

Lehtonen, S., 2020. Comparative study of anti-cheat

methods in video games. University of Helsinki,

Faculty of Science.

Pinto, J. P., Pimenta, A., & Novais, P., 2021. Deep learning

and multivariate time series for cheat detection in video

games. Machine Learning, 110(11), 3037-3057.

Tao, J., Xiong, Y., Zhao, S., Wu, R., Shen, X., Lyu, T., ...

& Pan, G., 2022. Explainable ai for cheating detection

and churn prediction in online games. IEEE

Transactions on Games, 15(2), 242-251

Maguluri, N. S. N., 2017. Multi-class classification of

textual data: Detection and mitigation of cheating in

massively multiplayer online role playing games.

Hughes, J., Collier, B., & Hutchings, A., 2019, November.

From playing games to committing crimes: A multi-

technique approach to predicting key actors on an

online gaming forum. In 2019 APWG Symposium on

Electronic Crime Research (eCrime) (pp. 1-12). IEEE.

Lukas, M., Tomicic, I., & Bernik, A., 2022. Anticheat

system based on reinforcement learning agents in

unity. Information, 13(4), 173.

Research on the Application of Artiﬁcial Intelligence in Online Game Anti-Cheating

111