CN112642161B

CN112642161B - Cheating detection and model training method and equipment for shooting game and storage medium

Info

Publication number: CN112642161B
Application number: CN202011480964.2A
Authority: CN
Inventors: 管炜; 朱飞
Original assignee: Perfect World Zhengqi Shanghai Multimedia Technology Co ltd
Current assignee: Perfect World Zhengqi Shanghai Multimedia Technology Co ltd
Priority date: 2020-12-15
Filing date: 2020-12-15
Publication date: 2024-06-18
Anticipated expiration: 2040-12-15
Also published as: CN112642161A; WO2022127277A1

Abstract

The embodiment of the application provides cheating detection and model training methods and equipment for shooting games and a storage medium. After obtaining video data of the shooting game, the server detects aiming paths of virtual shooting operation of a user in the shooting game based on the video data, classifies the aiming paths according to the neural network model, and obtains cheating detection results of the virtual shooting operation. The implementation mode is easy to deploy on the server side, an anti-cheating program does not need to be deployed on the game terminal, and furthermore, the cheating detection strategy is not easy to be bypassed by the cheating user in a targeted manner. Meanwhile, the virtual shooting operation can be accurately identified based on the aiming path according to strong learning and calculating capabilities of the neural network model, so that the accuracy and reliability of the cheating detection method are greatly improved.

Description

Cheating detection and model training method and equipment for shooting game and storage medium

Technical Field

The application relates to the technical field of internet, in particular to cheating detection and model training method and equipment for shooting games and a storage medium.

Background

In the First-person shooter game (FPS), some users have the act of cheating through a plug-in. For example, the enemy location is quickly located by the plug-in, the enemy is quickly targeted by the plug-in, and so on. These cheating actions undermine the fairness and interest of the game, resulting in a large number of user churn.

At present, some FPS game anti-cheating detection methods based on clients exist, and the method mainly comprises the steps of adding a kernel or an upper anti-cheating component to a game client to prevent an external program from accessing a game, collecting external feature codes and reporting the external feature codes so as to realize the blocking of cheating users.

However, in the existing anti-cheating technology, the functions of preventing cheating and detecting the plug-in are required to be deployed on the client, so that a plug-in developer can easily debug the anti-cheating component on the local machine, and a targeted bypass strategy is formulated. Further, the user's cheating behavior is not easily detected. Therefore, a new solution is to be proposed.

Disclosure of Invention

Aspects of the present application provide a cheating detection, model training method, apparatus, and storage medium for a shooting game, for improving detection accuracy for cheating in the shooting game.

The embodiment of the application provides a cheating detection method for shooting games, which is applicable to a server and comprises the following steps: acquiring video data of a shooting game; acquiring continuous multi-frame images of a user before virtual shooting operation in the shooting game from the video data; detecting aiming coordinates of the user from the continuous multi-frame images respectively to obtain a plurality of aiming coordinates; generating aiming paths corresponding to the virtual shooting operation according to the plurality of aiming coordinates; inputting the aiming path into a neural network model to obtain a cheating detection result corresponding to the virtual shooting operation.

The embodiment of the application also provides a training method of the cheating detection model of the shooting game, which comprises the following steps: acquiring a plurality of sample data, the plurality of sample data comprising: aiming path of virtual shooting operation of non-cheating user, and aiming path of virtual shooting operation of cheating user; the true value of the sample data corresponding to the non-cheating user is marked as non-cheating, and the true value of the sample data corresponding to the cheating user is marked as cheating; inputting the plurality of sample data into a neural network model to obtain respective cheating prediction results of the plurality of sample data; calculating a loss function of the neural network model according to the true values of the sample data and the cheating prediction results of the sample data; and optimizing model parameters of the neural network model based on the loss function until the loss function converges to a specified range.

The embodiment of the application also provides a server, which comprises: a memory and a processor; the memory is used for storing one or more computer instructions; the processor is configured to execute the one or more computer instructions to: the steps in the method provided by the embodiment of the application are executed.

The embodiment of the application also provides a computer readable storage medium storing a computer program, and the computer program can realize the steps in the method provided by the application embodiment when being executed.

In the cheating detection method for the shooting game provided by the embodiment of the application, after the server acquires video data of the shooting game, the aiming path of the virtual shooting operation of the user in the shooting game is detected based on the video data, and the aiming path is classified according to the neural network model, so that the cheating detection result of the virtual shooting operation is obtained. The implementation mode is easy to deploy on the server side, an anti-cheating program does not need to be deployed on the game terminal, and furthermore, the cheating detection strategy is not easy to be bypassed by the cheating user in a targeted manner. Meanwhile, the virtual shooting operation can be accurately identified based on the aiming path according to strong learning and calculating capabilities of the neural network model, so that the accuracy and reliability of the cheating detection method are greatly improved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute a limitation on the application. In the drawings:

FIG. 1 is a flow chart of a cheating detection method for a shooting game according to an exemplary embodiment of the present application;

FIGS. 2a, 2b, and 2c are schematic illustrations of aiming paths for a virtual shooting operation by a non-cheating user;

FIG. 2d, FIG. 2e, and FIG. 2f are schematic diagrams of aiming paths for a virtual shooting operation by a cheating user;

FIG. 3 is a schematic diagram of a neural network model according to an exemplary embodiment of the present application;

FIG. 4 is a flowchart of a training method of a cheating detection model of a shooting game according to an exemplary embodiment of the present application;

fig. 5 is a schematic structural diagram of a server according to an exemplary embodiment of the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application will be clearly and completely described below with reference to specific embodiments of the present application and corresponding drawings. It will be apparent that the described embodiments are only some, but not all, embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

For example, in CSGO (Counter-Strike: global Offensive) games, anti-cheating techniques are mainly provided by CSGO game official stem (game platform) and third party fight platforms (e.g., faceit,5E, etc. platforms). The anti-cheating technical scheme of the platform is mainly concentrated on a client, and through adding a kernel or an upper anti-cheating component on the client, the access of a plug-in program to CSGO games is prevented, and plug-in feature codes are collected and reported to realize blocking to cheating users in CSGO games.

However, in the existing anti-cheating technology, the functions of preventing cheating and detecting the plug-in are required to be deployed on the client, so that a plug-in developer can easily debug the anti-cheating component on the local machine, and a targeted bypass strategy is formulated. For example, a plug-in developer may customize the bypass policy by restricting the network, prohibiting anti-cheating reporting; or the bypass strategy can be customized by modifying the anti-cheating protocol to report only normal behavior; or patches can be added to the anti-cheating code to make it no longer work, thereby customizing the bypass policy. Based on the above, the anti-cheating technique of the game deployed at the client cannot accurately detect the cheating behavior of the user.

Meanwhile, when developing the anti-cheating technology, a developer needs to consider the stability and performance of the anti-cheating technology on a mass user machine and the compatibility of the anti-cheating technology with some third party software, so that the strength of the anti-cheating technology is sacrificed to a certain extent. However, the plug-in developer does not need to consider the above factors, and more cooperation operations of the plug-in user can be obtained, for example, the plug-in user can unload the anti-cheating program first and then enable the plug-in program, so as to avoid the detected cheating behavior. The above factors also result in the inability of client-based anti-cheating techniques to accurately and reliably detect the user's cheating behavior.

In view of the foregoing technical problems, in some embodiments of the present application, a solution is provided, and in the following, the technical solutions provided by the embodiments of the present application will be described in detail with reference to the accompanying drawings.

Fig. 1 is a flow chart of a cheating detection method for a shooting game according to an exemplary embodiment of the present application, as shown in fig. 1, when the method is executed on a server side, the method mainly includes the following steps:

step 101, obtaining video data of shooting games.

Step 102, acquiring continuous multi-frame images of a user before virtual shooting operation in the shooting game from the video data.

And 103, detecting aiming coordinates of the user from the continuous multi-frame images respectively to obtain a plurality of aiming coordinates.

And 104, generating an aiming path corresponding to the virtual shooting operation according to the plurality of aiming coordinates.

And 105, inputting the aiming path into a neural network model to obtain a cheating detection result corresponding to the virtual shooting operation.

The embodiment can be executed by a server independent of the game terminal to avoid the cheating detection process from being interfered by the cheating user. The server may be implemented as a conventional server, a cloud host, a virtual center, or other devices, which is not limited in this embodiment. The server device mainly includes a processor, a hard disk, a memory, a system bus, and the like, which are similar to a general computer architecture and will not be described again.

In a non-step 101, the shooting game may include a plurality of FPS games. Video data of the shooting game can be obtained through recording a screen during the game process or can be obtained based on a frame synchronization technology. The video data is acquired based on a frame synchronization technology, and is a video data acquisition mode which is not perceived by a user (game player), so that a bypass strategy is not easily specified in a targeted manner. An embodiment of acquiring video data based on the frame synchronization technique will be exemplarily described below.

In some FPS games, a game frame is typically built using frame synchronization techniques. The frame synchronization refers to decomposing data of the FPS game in the same second into N status frames, where each status frame includes complete status information of the game at this time, such as character position, visual field orientation, equipment, blood volume, and other game status information. When the game server accesses a plurality of game clients, the game clients can synchronize the status frames of the game in real time in order to ensure that each user sees the same game world. That is, the game server may send N status frames corresponding to each second to N game clients, so as to ensure that the game server and all clients joining the game synchronize the status frames of the game in real time.

Taking CSGO for example, CSGO typically generates a video at the end of each game. The video data may be imported CSGO into the game for playback to play back the game from various perspectives. The video data may be referred to as Demo, which may be regarded as a collection of game status frames of the game.

When the frame synchronization is performed on the game, real-time game data can be sampled according to the set state frame sampling frequency. For example, CSGO games may be sampled at a set frame sampling frequency to obtain a status frame. The sampling frequency of the status frame may be 32 tics (32 frames/second), 64 tics (64 frames/second), 128 tics (128 frames/second), or other sampling frequencies, which is not limited in this embodiment.

The collection of the status frames of each game obtained by the game server may be referred to as video data corresponding to the game. In a game, the user may perform one or more virtual shooting operations. In general, before performing a virtual shooting operation, in order to increase the hit rate of the shooting operation, a user may control the virtual shooting prop to adjust the angle and position in the game world, so as to "aim" at a target object in the game world.

It should be appreciated that the aiming operation of the user in the game may be analyzed on a microscopic time scale when the game process is sampled at a higher frame sampling frequency. On a microscopic time scale, the aiming operation flow (e.g., aiming curve, mouse movement curve, etc.) for each user should be smooth, inertial, free of abrupt changes, as shown in fig. 2a, 2b, and 2 c. When a user uses the cheating software of the plug-in type in a game, the software of the plug-in type usually needs to forcedly correct the input of the user so as to realize the high shooting hit rate effects of the lock head, the one-shot kill and the like of the virtual shooting prop. Thus, the aiming operation flow of the cheating user in the game often presents characteristics of jump, inertia-free steering and the like which do not conform to the human operation rule on a microscopic time scale, as shown in fig. 2d, 2e and 2 f.

Based on the above analysis, in the present embodiment, by analyzing the operation flow of the user on a microscopic time scale, it is possible to detect whether or not the virtual shooting operation of the user has a cheating behavior.

After obtaining the video data of the FPS in step 102 and step 103, the virtual shooting operation of the user may be detected based on the video data first, and a key frame image for the user to perform the virtual shooting operation may be determined. After the key frame images are determined, successive K frame images preceding the key frame images may be acquired from the video data as images for analyzing the aiming path. Wherein, K is a positive integer, and the value of K can be set according to actual requirements. In some alternative embodiments, K may be 6, 8, 10 or other alternative values, which are not limiting.

Next, the user's targeting coordinates may be detected from the successive multiframe images, respectively. Wherein one aiming coordinate is detected for each frame of image and a plurality of aiming coordinates are detected for successive frames of images. After a plurality of aiming coordinates are obtained, an aiming path corresponding to the virtual shooting operation can be generated according to the plurality of aiming coordinates.

In step 104, the aiming path is used to describe a moving path formed by controlling the virtual shooting prop to shoot and aim before the user performs the virtual shooting operation, that is, an operation flow of the user performing the virtual shooting operation.

In the FPS game, a game player performs shooting aiming through the subjective view angle of the virtual game character, and the subjective view angle of the virtual game character is the aiming view angle. Thus, in some embodiments, visual angle changes of the virtual game character may be detected based on the video data, thereby detecting the aiming operation flow of the user.

For each virtual shooting operation of the user, when the aiming path is detected, a plurality of frame images before the virtual shooting operation, namely a plurality of frame images corresponding to the aiming process of the user, can be obtained from video data. The multi-frame image typically contains visual angle state information for a user to operate a virtual shooting prop (e.g., virtual gun) of the virtual game character to aim at. And detecting the visual angle based on the multi-frame image, and calculating a change curve of the detected visual angle to determine the aiming path of the virtual shooting operation.

It should be appreciated that as the perspective of the game character changes during targeting, the point of fall of the game character's line of sight also changes. Thus, in other embodiments, the coordinates of the drop point of the line of sight of the game character may be detected and the change in the drop point coordinates calculated to yield the aiming path for the virtual shooting operation. After the targeting path for the virtual firing operation is obtained, the targeting path may be input into a pre-trained neural network model.

In step 105, the neural network model may be trained in advance from a large number of sample data. In the process of training the neural network model, the neural network model can learn the characteristics of a normal aiming curve under the condition of non-cheating and the characteristics of the aiming curve under the condition of cheating based on sample data adopted by training, so that the aim of intelligently classifying the input aiming curve can be fulfilled. The optional implementation manner of training the neural network model will be described in the following examples, which are not repeated here.

Based on this, in this embodiment, after the aiming path of the virtual shooting operation is obtained, the aiming path may be input into the neural network model, and the cheating detection result corresponding to the virtual shooting operation may be determined according to the output of the neural network model. Wherein the cheating detection result may include; the virtual shooting operation is a cheating behavior or the virtual shooting operation is a non-cheating behavior.

In some embodiments, if it is detected that multiple virtual shooting operations of a user in the game process are all cheating actions, a warning message may be sent to the user, or the game account of the user may be blocked, or the like, which is not limited in this embodiment.

In this embodiment, after acquiring video data of a shooting game, a server detects an aiming path of a virtual shooting operation of a user in the shooting game based on the video data, and classifies the aiming path according to a neural network model to obtain a cheating detection result of the virtual shooting operation. The implementation mode is easy to deploy on the server side, an anti-cheating program does not need to be deployed on the game terminal, and furthermore, the cheating detection strategy is not easy to be bypassed by the cheating user in a targeted manner. Meanwhile, the virtual shooting operation can be accurately identified based on the aiming path according to strong learning and calculating capabilities of the neural network model, so that the accuracy and reliability of the cheating detection method are greatly improved.

In addition, the cheating detection scheme for non-client deployment provided by the application embodiment can avoid influencing the normal installation and operation of the client on one hand, and can achieve no perception of a user on the other hand, and the accuracy rate of the cheating detection is higher.

The foregoing examples describe embodiments of detecting an aiming path based on a change in viewing angle and embodiments of detecting an aiming path based on a line-of-sight landing, both of which are further exemplified below.

An alternative embodiment of detecting an aiming path will be exemplarily described below taking any one of the consecutive multi-frame images as an example.

Embodiment a: based on the change in viewing angle, an aiming path is detected.

In FPS games, a game player aims at shooting through the subjective view of the virtual game character, and thus the view of the game character is the aim view. Based on this, optionally, for the t-th frame image of the continuous multi-frame images, a target ray may be emitted in the image with the aiming viewpoint of the game character as an end point to determine the aiming direction. The aiming view point may be, but not limited to, a position of an eye of a game character, a position of a muzzle of a virtual prop used by the game character, or a position of a sight on the virtual prop gun.

Next, the angles between the target ray and the horizontal plane and the vertical plane may be calculated, to obtain a first angle offset value x _t and a second angle offset value y _t, and according to the first angle offset value x _t and the second angle offset value y _t, the aiming coordinate (x _t,y_t) of the user in the image is obtained.

For example, when performing cheating analysis based on the Demo of the 64tick of CSGO game, the user's aiming coordinates may be sampled every 1/64 second (15.6 milliseconds) and recorded as two angular offsets of x, y in the game. When the aiming coordinate is sampled, a ray can be emitted from eyes of a character in a game, x represents an included angle between the ray and a horizontal plane, the value range is (-180 degrees, 180 degrees), y represents an included angle between the ray and a vertical plane, and the value range is (-180 degrees, 180 degrees).

Embodiment B: based on the change in the line-of-sight landing point, an aiming path is detected.

In this embodiment, for any frame of image, a target ray is emitted with the aiming viewpoint of the game character as an end point, and based on the image, an intersection point at which the target ray intersects an object in the game world is determined. Next, coordinates of the intersection point in the game world are calculated as aiming coordinates of the user in the image.

After the aiming coordinates corresponding to each of the continuous multi-frame images are acquired, an aiming path corresponding to the virtual shooting operation can be generated based on the aiming coordinates. An exemplary description will be given below in connection with embodiment a.

In some alternative embodiments, a coordinate sequence formed by a plurality of aiming coordinates may be used as the aiming path for the firing operation. For example, the aiming path may be described as { (x ₁,y₁),(x₂,y₂),…,(x_K,y_K) }.

In other alternative embodiments, an aiming vector for each aiming coordinate may be calculated, a plurality of aiming vectors may be derived, and an aiming path for the virtual firing operation may be determined based on the plurality of aiming vectors.

Optionally, for any one of the plurality of aiming coordinates, the aiming coordinate and an adjacent subsequent aiming coordinate may be vectorized to obtain an aiming vector of the aiming coordinate. Continuing with the example of the aiming coordinate (x _t,y_t) in the image of the t frame, the change in the aiming coordinate of the t+1st frame and the aiming coordinate of the t frame can be vectorized and denoted as an aiming vector v _t＝(x_t+1-x_t,y_t+1–y_t of the image of the t frame.

After deriving the respective aiming vectors for the plurality of aiming coordinates, a matrix of the plurality of aiming vectors may be used to describe the aiming path of the virtual shooting operation. Wherein the aiming path can be described as: { v ₁,v₂…v_k }, the dimension of the matrix is m×k, m representing the characteristic dimension of v _t.

In some embodiments, at the computational level, v _t may be described by four values, including: projection of v _t on the x-axis, projection of v _t on the y-axis, modulus of v _t, angle of v _t with respect to the positive x-axis. In this calculation, v _t has 4 feature dimensions, and then the matrix corresponding to the aiming path has 4*K feature dimensions in total. When the value of K is 8, the matrix corresponding to the aiming path has 4*8 =32 feature dimensions in total.

After the aiming path corresponding to the virtual shooting operation is obtained based on the above embodiment, the aiming path may be input into the neural network model. Alternatively, in the present embodiment, the neural network model may be implemented as: MLP (Multi-Layer Perceptron), RNN (Recurrent Neural Network ), transducer model, etc., the present embodiment includes but is not limited to this. An exemplary description will be given below taking an MLP as an example.

The structure of the neural network model may be as shown in fig. 3, which is an MLP fully connected network, and the output layer of the neural network model is a two-dimensional regression function softmax, which is used as a classifier, and may output probability distribution that the aiming path belongs to normal operation and cheating operation.

As shown in fig. 3, the input layer (first layer) and each intermediate hidden layer of the neural network model may include a number of neurons and activation functions Relu, respectively. The input layer of the neural network model may include 400 neurons, and the number of neurons gradually decreases in the hidden layer between the input layer and the output layer. The rule of decreasing number of neurons is related to the drop rate (the proportion of neurons lost in the network layer). Wherein the drop out rate may be set to 0.25 or other alternative values. It should be understood that the structure illustrated in fig. 3 is only used for illustrating the structure of the neural network model, and in practice, the structure of the neural network model may be adjusted according to the detection requirement, and is not limited to the structure illustrated in fig. 3.

After inputting the targeting path into the input layer of the neural network model, in the neural network model, feature calculation can be performed on the targeting path based on model parameters learned in advance. Based on the calculated features, a probability of the virtual shooting operation corresponding to the aiming path being a cheating behavior may be calculated. If the calculated probability is greater than the set probability threshold, the cheating detection result of the virtual shooting operation as the cheating behavior can be output. That is, the trajectory of the virtual shooting operation of the user is regarded as a suspicious trajectory.

The set probability threshold may be set according to actual requirements, for example, may be set to 80%, 85%, 90% or other probability thresholds, which is not limited in this embodiment. For example, in some embodiments, if the probability of a user's one virtual shooting operation being a cheating action is greater than 80% or 90%, the trajectory of the user's current virtual shooting operation may be determined to be a suspicious trajectory.

Wherein, for a user, if the number of virtual shooting operations using plug-ins in the virtual shooting operation in the game of the user is determined to be greater than a set number threshold, the user can be marked as a cheating user. The set number threshold may be set according to actual requirements, for example, may be set to 5, 8, 10, etc., which is not limited in this embodiment. For example, if the trajectory of a virtual shooting operation of a user in a game is detected as a number of suspicious trajectories greater than 5, the user may be marked as a cheating user. For the cheating user, warning processing or account blocking processing can be performed, and details are omitted.

It is worth to describe that the server-based cheating detection method provided by the embodiment of the application can be used independently or matched with the cheating detection method deployed by the client. When the method is matched with the cheating detection method deployed at the client, on one hand, the accuracy of the cheating detection result can be improved, and on the other hand, the cheating detection result deployed at the server and the detection result deployed at the client can be subjected to cross quotation so as to continuously optimize the two detection methods. An exemplary description will be made below.

Optionally, the aiming path of the virtual shooting operation of the user is input into the neural network model, and after the cheating detection result output by the neural network model is obtained, the cheating detection data of the user sent by the client can be further obtained. Wherein, the client runs the cheating detection program.

Next, suspicious cheating features of the user can be obtained from the cheating detection data; wherein the suspected cheating feature may comprise: the user uses a driver with an unknown signature at the client or has an unknown application to initiate access to the gaming system, etc., and the present embodiment is not limited.

Next, a cheating judgment score for the user may be calculated based on the suspicious cheating characteristics detected by the client and the cheating detection results calculated by the server. Alternatively, a weight coefficient may be set in advance for each suspected cheating feature and the cheating detection result. After the suspicious cheating features and the cheating detection results are obtained, weighting calculation can be performed on the suspicious cheating features and the cheating detection results based on the weight coefficients, so that the cheating judgment score is obtained. If the cheating judgment score is greater than the set score threshold, the user can be determined to be a cheating user. The score threshold may be set according to actual requirements, which is not limited in this embodiment.

Continuing with CSGO game as an example, the anti-cheating function module running on the CSGO game client may detect whether the user has accessed CSGO using an unknown signed driver or an unknown program during the course of the game. If the suspicious cheating feature is detected, and the server detects that the user has self-aiming cheating behaviors based on machine learning, the server can determine CSGO that the user has plug-in behaviors. At this point, the program used by the user may be blacklisted. If the user or other users are detected to start the programs in the blacklist on the client, account blocking processing can be carried out on the users.

In this embodiment, the strength of the anti-cheating function of the cheating detection program running on the client (i.e., the strength of the anti-cheating is sacrificed) can be appropriately reduced by the auxiliary detection function of the server. Furthermore, on one hand, the performance of the game operated by the client can be ensured, the stability of the game operated by the client can be enhanced, and the compatibility of the third party software can be improved, on the other hand, the comprehensive judgment of the cheating behavior can be carried out by combining various detection data, the accuracy of the cheating detection result can be improved, and the anti-cheating effect of 1+1>2 is realized.

In the following examples, alternative implementations of training neural network models will be exemplarily described with reference to the accompanying drawings.

Fig. 4 is a flowchart of a training method of a cheating detection model of a shooting game according to an exemplary embodiment of the present application, as shown in fig. 4, where the method includes:

Step 401, acquiring a plurality of sample data, wherein the plurality of sample data includes: aiming path of virtual shooting operation of cheating user and aiming path of virtual shooting operation of non-cheating user; the true value of the sample data corresponding to the non-cheating user is marked as non-cheating, and the true value of the sample data corresponding to the cheating user is marked as cheating.

And step 402, inputting the plurality of sample data into the neural network model to obtain respective cheating prediction results of the plurality of sample data.

Step 403, calculating a loss function of the neural network model according to the true values of the sample data and the cheating prediction results of the sample data.

And step 404, optimizing model parameters of the neural network model based on the loss function until the loss function converges to a specified range.

In this embodiment, video data of shooting games of cheating users and video data of shooting games of non-cheating users may be acquired. Based on video data of the cheating user, a targeting path of virtual shooting operation of the cheating user can be obtained; based on the video data of the non-cheating user, the aiming path of the virtual shooting operation of the cheating user can be obtained. The optional implementation manner of acquiring the aiming path based on the video data may refer to the description of the foregoing embodiment, which is not repeated here.

A plurality of sample data required for training the neural network model, including an aiming path for virtual shooting operations of the cheating user and an aiming path for virtual shooting operations of the non-cheating user; the true value of each sample data is marked according to the type of the user. The user types refer to cheating user types and non-cheating user types. That is, the true value of the aiming path of the virtual shooting operation of the cheating user is marked as non-cheating, and the true value of the aiming path of the virtual shooting operation of the cheating user is marked as cheating.

In the sample labeling mode, the thought of approximate labeling is adopted, and the truth value marks of all aiming paths of cheating users with cheating behaviors are used as the cheating. Although, each virtual shooting operation of the cheating user is not corrected by the plug-in program, and some normal aiming data may be included, the true value of the normal aiming data is marked as cheating and can be used as noise in the training data. During training, a large amount of data of non-cheating users can automatically correct the noise, and obvious cheating characteristics can be maintained. That is, the normal targeting data of the cheating user has a feature that coincides with the data of a large number of non-cheating users, and during training, the normal targeting data of the cheating user can be re-corrected by the targeting data of a large number of non-cheating users, while the data of those cheating features is not corrected. Based on the mode, the performance of the neural network model obtained through training can be improved while the labeling cost is reduced.

After a plurality of sample data are obtained, the plurality of sample data can be input into a neural network model to obtain respective cheating prediction results of the plurality of sample data.

After obtaining the cheating prediction result of each sample data, a loss function of the neural network model can be calculated according to the true value of each sample data and the cheating prediction result of each sample data, model parameters of the neural network model are optimized based on the loss function, and a result model obtained through training is output until the loss function converges to a specified range.

It should be noted that, the execution subjects of each step of the method provided in the above embodiment may be the same device, or the method may also be executed by different devices. For example, the execution subject of steps 401 to 404 may be device a; for another example, the execution subject of steps 401 and 402 may be device a, and the execution subject of step 403 may be device B; etc.

In addition, in some of the flows described in the above embodiments and the drawings, a plurality of operations appearing in a specific order are included, but it should be clearly understood that the operations may be performed out of the order in which they appear herein or performed in parallel, the sequence numbers of the operations, such as 401, 402, etc., are merely used to distinguish between the various operations, and the sequence numbers themselves do not represent any order of execution. In addition, the flows may include more or fewer operations, and the operations may be performed sequentially or in parallel.

It should be noted that, the descriptions of "first" and "second" herein are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, and are not limited to the "first" and the "second" being different types.

Fig. 5 illustrates a schematic structural diagram of a server according to an exemplary embodiment of the present application, which is suitable for the cheating detection method for shooting game provided in the foregoing embodiment. As shown in fig. 5, the server includes: memory 501 and processor 502.

Memory 501 is used to store computer programs and may be configured to store various other data to support operations on the server. Examples of such data include instructions for any application or method operating on a server, contact data, phonebook data, messages, pictures, video, and the like.

The memory 501 may be implemented by any type of volatile or non-volatile memory device or combination thereof, such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disk.

A processor 502 coupled to the memory 501 for executing the computer program in the memory 501 for: acquiring video data of a shooting game; detecting an aiming path of a virtual shooting operation of a user in the shooting game based on the video data; inputting the aiming path into a neural network model to obtain a cheating detection result corresponding to the virtual shooting operation.

Further optionally, the processor 502 is specifically configured to, when detecting, based on the video data, an aiming path of a virtual shooting operation of a user in the shooting game: acquiring continuous multi-frame images before shooting operation from the video data; detecting aiming coordinates of the user from the continuous multi-frame images respectively to obtain a plurality of aiming coordinates; and generating an aiming path corresponding to the virtual shooting operation according to the plurality of aiming coordinates.

Further optionally, the processor 502 is specifically configured to, when detecting the aiming coordinate of the user from the continuous multi-frame images to obtain a plurality of aiming coordinates: aiming at any frame of image in the continuous multi-frame images, in the image, aiming view points of game roles are taken as endpoints, and target rays are emitted; respectively calculating the included angles of the target rays and the horizontal plane and the vertical plane to obtain a first angle offset value and a second angle offset value; and obtaining aiming coordinates of the user in the image according to the first angle offset value and the second angle offset value.

Further optionally, the processor 502 is specifically configured to, when generating the aiming path corresponding to the virtual shooting operation according to the plurality of aiming coordinates: aiming at any one of the plurality of aiming coordinates, carrying out vectorization processing on the aiming coordinate and the adjacent next aiming coordinate to obtain an aiming vector of the aiming coordinate; and acquiring a matrix formed by the aiming vectors of the plurality of aiming coordinates as an aiming path corresponding to the virtual shooting operation.

Further optionally, the processor 502 is specifically configured to, when generating the aiming path corresponding to the virtual shooting operation according to the plurality of aiming coordinates: and taking a coordinate sequence formed by the plurality of aiming coordinates as an aiming path corresponding to the virtual shooting operation.

Further optionally, when the aiming path is input into a neural network model to obtain a cheating detection result corresponding to the virtual shooting operation, the processor 502 is specifically configured to: inputting the targeting path into a neural network model; in the neural network model, performing feature calculation on the aiming path based on model parameters learned in advance; according to the calculated characteristics, calculating the probability of the virtual shooting operation corresponding to the aiming path as the cheating behavior; and if the calculated probability is larger than a set probability threshold, outputting a cheating detection result of which the virtual shooting operation is a cheating behavior.

Further optionally, the processor 502 is further configured to: and if the number of the virtual shooting operations detected as the cheating behavior is larger than the set number threshold in the virtual shooting operations of the user, determining that the user is the cheating user.

Further optionally, the processor 502 is further configured to: acquiring cheating detection data of the user sent by a client, wherein the client runs a cheating detection program; acquiring suspicious cheating characteristics of the user from the cheating detection data; calculating the cheating judgment score of the user based on the suspicious cheating characteristics and the cheating detection result; and if the cheating judgment score is larger than the set score threshold, determining that the user is a cheating user.

Further optionally, the processor 502 is further configured to: acquiring a plurality of sample data, the plurality of sample data comprising: aiming path of virtual shooting operation of cheating user and aiming path of virtual shooting operation of non-cheating user; the true value of the sample data corresponding to the non-cheating user is marked as non-cheating, and the true value of the sample data corresponding to the cheating user is marked as cheating; inputting the plurality of sample data into the neural network model to obtain respective cheating prediction results of the plurality of sample data; calculating a loss function of the neural network model according to the true values of the sample data and the cheating prediction results of the sample data; and optimizing model parameters of the neural network model based on the loss function until the loss function converges to a specified range.

Further, as shown in fig. 5, the server further includes: a communication component 503, a power supply component 504, and the like. Only some of the components are schematically shown in fig. 5, which does not mean that the server only comprises the components shown in fig. 5.

Wherein the communication component 503 is configured to facilitate wired or wireless communication between the device in which the communication component is located and other devices. The device in which the communication component is located may access a wireless network based on a communication standard, such as WiFi,2G, 3G, 4G, or 5G, or a combination thereof. In one exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication component may be implemented based on Near Field Communication (NFC) technology, radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, ultra Wideband (UWB) technology, bluetooth (BT) technology, and other technologies.

Wherein the power supply assembly 504 is configured to provide power to various components of the device in which the power supply assembly is located. The power components may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the devices in which the power components are located.

Accordingly, the present application also provides a computer readable storage medium storing a computer program, where the computer program is executed to implement the steps executable by the server in the above method embodiments.

It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Disks (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises the element.

The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and variations of the present application will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the application are to be included in the scope of the claims of the present application.

Claims

1. The cheating detection method for the shooting game is suitable for a server and is characterized by comprising the following steps:

acquiring video data of a shooting game;

acquiring continuous multi-frame images of a user before virtual shooting operation in the shooting game from the video data;

Detecting aiming coordinates of the user from the continuous multi-frame images respectively to obtain a plurality of aiming coordinates;

Generating aiming paths corresponding to the virtual shooting operation according to the plurality of aiming coordinates; the aiming path is described by a matrix formed by aiming vectors of the aiming coordinates;

inputting the aiming path into a neural network model to obtain a cheating detection result corresponding to the virtual shooting operation;

inputting the aiming path into a neural network model to obtain a cheating detection result corresponding to the virtual shooting operation, wherein the method comprises the following steps: in the neural network model, performing feature calculation on the aiming path based on model parameters learned in advance; according to the calculated characteristics, calculating the probability of the virtual shooting operation corresponding to the aiming path as the cheating behavior; and if the calculated probability is larger than a set probability threshold, outputting a cheating detection result of which the virtual shooting operation is a cheating behavior.

2. The method of claim 1, wherein detecting the user's targeting coordinate from the successive multi-frame images, respectively, results in a plurality of targeting coordinates, comprising:

aiming at any frame of image in the continuous multi-frame images, in the image, aiming view points of game roles are taken as endpoints, and target rays are emitted;

Respectively calculating the included angles of the target rays and the horizontal plane and the vertical plane to obtain a first angle offset value and a second angle offset value;

And obtaining aiming coordinates of the user in the image according to the first angle offset value and the second angle offset value.

3. The method of claim 1, wherein generating an aiming path corresponding to the virtual shooting operation from the plurality of aiming coordinates comprises:

Aiming at any one of the plurality of aiming coordinates, carrying out vectorization processing on the aiming coordinate and the adjacent next aiming coordinate to obtain an aiming vector of the aiming coordinate;

And acquiring a matrix formed by the aiming vectors of the plurality of aiming coordinates as an aiming path corresponding to the virtual shooting operation.

4. The method of claim 1, wherein generating an aiming path corresponding to the virtual shooting operation from the plurality of aiming coordinates comprises:

And taking a coordinate sequence formed by the plurality of aiming coordinates as an aiming path corresponding to the virtual shooting operation.

5. The method as recited in claim 1, further comprising:

And if the number of the virtual shooting operations detected as the cheating behavior is larger than the set number threshold in the virtual shooting operations of the user, determining that the user is the cheating user.

6. The method as recited in claim 1, further comprising:

Acquiring cheating detection data of the user sent by a client, wherein the client runs a cheating detection program;

acquiring suspicious cheating characteristics of the user from the cheating detection data;

calculating the cheating judgment score of the user based on the suspicious cheating characteristics and the cheating detection result;

And if the cheating judgment score is larger than the set score threshold, determining that the user is a cheating user.

7. The method as recited in claim 5, further comprising:

Acquiring a plurality of sample data, the plurality of sample data comprising: aiming path of virtual shooting operation of cheating user is not aiming path of virtual shooting operation of cheating user, and aiming path of virtual shooting operation of cheating user; the true value of the sample data corresponding to the non-cheating user is marked as non-cheating, and the true value of the sample data corresponding to the cheating user is marked as cheating;

Inputting the plurality of sample data into the neural network model to obtain respective cheating prediction results of the plurality of sample data;

calculating a loss function of the neural network model according to the true values of the sample data and the cheating prediction results of the sample data;

And optimizing model parameters of the neural network model based on the loss function until the loss function converges to a specified range.

8. A training method of a cheating detection model of a shooting game is characterized by comprising the following steps:

Acquiring a plurality of sample data, the plurality of sample data comprising: aiming path of virtual shooting operation of non-cheating user, and aiming path of virtual shooting operation of cheating user; the true value of the sample data corresponding to the non-cheating user is marked as non-cheating, and the true value of the sample data corresponding to the cheating user is marked as cheating; the aiming path of any user of the non-cheating user and the cheating user is described by adopting a matrix formed by aiming vectors of a plurality of aiming coordinates of the any user; the plurality of aiming coordinates are detected from continuous multi-frame images of any user before virtual shooting operation in a shooting game;

inputting the plurality of sample data into a neural network model to obtain respective cheating prediction results of the plurality of sample data;

optimizing model parameters of the neural network model based on the loss function until the loss function converges to a specified range; the converged neural network model is used for outputting a cheating detection result of the virtual shooting operation as a cheating behavior according to the input aiming coordinates of the virtual shooting operation.

9. A server, comprising: a memory and a processor;

The memory is used for storing one or more computer instructions;

the processor is configured to execute the one or more computer instructions to: performing the steps of the method of any one of claims 1-8.

10. A computer readable storage medium storing a computer program, characterized in that the computer program is capable of implementing the steps of the method of any one of claims 1-8 when executed.