Data processing method and apparatus, electronic device, and medium

Abstract

Embodiments of the present application disclose a data processing method and device, electronic equipment and medium, which can be applied to the technical field of data processing. The method comprises: obtaining to-be-detected operation data of a target object, and determining operation state features corresponding to each sampling time according to operation data of T sampling times; the operation state features comprise state information of the target object and state information of an associated object of the target object, and the associated object is an object performing a targeting operation or an attack operation on the target object; operation state features corresponding to each sampling time are combined in chronological order to obtain an operation state sequence of the target object; and a cheating detection result of the target object is determined based on the operation state sequence. The embodiments of the present application can help to improve the accuracy of cheating behavior detection. The embodiments of the present application can also be applied to various scenes such as cloud technology, artificial intelligence, intelligent transportation, auxiliary driving, intelligent household appliances and the like.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a data processing method, apparatus, electronic device and medium. Background Technology

[0002] In some scenarios, users use cheating tools to undermine fairness. For example, in action games, such as first-person shooters, some players use cheats like aimbots and wallhacks to disrupt the game's balance. Therefore, to ensure fairness, it's necessary to detect cheating during gameplay. Currently, methods for detecting cheating typically rely on manually set rules and limited data, leading to low accuracy in detection. Summary of the Invention

[0003] This application provides a data processing method, apparatus, electronic device, and medium that helps improve the accuracy of detecting cheating behavior.

[0004] On one hand, embodiments of this application disclose a data processing method, the method comprising:

[0005] Acquire the operation data to be detected of the target object, wherein the operation data to be detected includes the operation data of the target object at T sampling times, where T is a positive integer;

[0006] Based on the operation data at the T sampling times, the operation state characteristics corresponding to each sampling time are determined; the operation state characteristics include the state information of the target object and the state information of the associated object of the target object, wherein the associated object is the object to which the target object performs a targeting operation or an attack operation;

[0007] The operation state features corresponding to each sampling moment are combined in chronological order to obtain the operation state sequence of the target object.

[0008] The cheating detection result of the target object is determined based on the operation state sequence, and the cheating detection result is used to indicate whether the target object has cheating behavior.

[0009] On one hand, embodiments of this application disclose a data processing apparatus, the apparatus comprising:

[0010] The acquisition unit is used to acquire the operation data to be detected of the target object, wherein the operation data to be detected includes the operation data of the target object at T sampling times, where T is a positive integer;

[0011] The processing unit is configured to determine the operation state characteristics corresponding to each sampling time based on the operation data of the T sampling times; the operation state characteristics include the state information of the target object and the state information of the associated object of the target object, wherein the associated object is the object to which the target object performs a targeting operation or an attack operation;

[0012] The processing unit is also used to combine the operation state features corresponding to each sampling time in chronological order to obtain the operation state sequence of the target object;

[0013] The processing unit is further configured to determine the cheating detection result of the target object based on the operation state sequence, and the cheating detection result is used to indicate whether the target object has cheating behavior.

[0014] On one hand, embodiments of this application provide an electronic device, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to perform the following steps:

[0015] Acquire the operation data to be detected of the target object, wherein the operation data to be detected includes the operation data of the target object at T sampling times, where T is a positive integer;

[0016] Based on the operation data at the T sampling times, the operation state characteristics corresponding to each sampling time are determined; the operation state characteristics include the state information of the target object and the state information of the associated object of the target object, wherein the associated object is the object to which the target object performs a targeting operation or an attack operation;

[0017] The operation state features corresponding to each sampling moment are combined in chronological order to obtain the operation state sequence of the target object.

[0018] The cheating detection result of the target object is determined based on the operation state sequence, and the cheating detection result is used to indicate whether the target object has cheating behavior.

[0019] On one hand, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, perform the following steps:

[0020] Acquire the operation data to be detected of the target object, wherein the operation data to be detected includes the operation data of the target object at T sampling times, where T is a positive integer;

[0021] Based on the operation data at the T sampling times, the operation state characteristics corresponding to each sampling time are determined; the operation state characteristics include the state information of the target object and the state information of the associated object of the target object, wherein the associated object is the object to which the target object performs a targeting operation or an attack operation;

[0022] The operation state features corresponding to each sampling moment are combined in chronological order to obtain the operation state sequence of the target object.

[0023] The cheating detection result of the target object is determined based on the operation state sequence, and the cheating detection result is used to indicate whether the target object has cheating behavior.

[0024] On one hand, embodiments of this application provide a computer program product or computer program that includes computer instructions that, when executed by a processor, can implement the method provided in the above-mentioned aspect.

[0025] By employing the embodiments of this application, it is possible to acquire the operation data to be detected of the target object, such as the operation data of the target object in a game. Based on this operation data, the operation state characteristics corresponding to each sampling moment can be determined. These operation state characteristics can include the state information of the target object and the object the target object is aiming at or attacking, thus obtaining operation state characteristics with rich information. An operation state sequence is then generated based on the operation characteristics at each sampling moment, and the cheating detection result is determined based on the operation state sequence to determine whether the target object is engaging in cheating behavior. Therefore, the operation state sequence can be determined through the state information of the target object and associated objects. In particular, the state information of the object the target object is aiming at or attacking (i.e., associated objects) is introduced at each sampling moment to construct an operation state sequence for cheating detection, rather than being limited to focusing on some state information of the target object itself. This is equivalent to introducing richer data for cheating detection, which helps improve the accuracy of cheating detection. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a schematic diagram of the structure of a data processing system provided in an embodiment of this application;

[0028] Figure 2This is a schematic diagram of a game scene provided in an embodiment of this application;

[0029] Figure 3 This is a schematic diagram of a game scene provided in an embodiment of this application;

[0030] Figure 4 This is a flowchart illustrating a data processing method provided in an embodiment of this application;

[0031] Figure 5 This is a schematic diagram illustrating the effect of an operation state sequence provided in an embodiment of this application;

[0032] Figure 6 This is a flowchart illustrating a data processing method provided in an embodiment of this application;

[0033] Figure 7 This is a schematic diagram of the structure of an attention mechanism network layer provided in an embodiment of this application;

[0034] Figure 8 This is a schematic diagram illustrating the effect of an attention distribution map provided in an embodiment of this application;

[0035] Figure 9 This is a schematic diagram of a cheating detection process provided in an embodiment of this application;

[0036] Figure 10 This is a schematic diagram of a cheating detection process provided in an embodiment of this application;

[0037] Figure 11 This is a schematic diagram of the structure of a data processing device provided in an embodiment of this application;

[0038] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0039] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0040] This application proposes a data processing scheme that can acquire the operation data of a target object to be detected, such as the operation data of the target object in a game. Based on this operation data, the scheme determines the operation state features corresponding to each sampling moment. These operation state features can include the state information of the target object and the object the target object is aiming at or attacking, thus obtaining operation state features with rich information. An operation state sequence is then generated based on the operation features at each sampling moment, and a cheating detection result is determined based on the operation state sequence to determine whether the target object is engaging in cheating behavior. This approach determines the operation state sequence through the state information of the target object and associated objects. In particular, it introduces the state information of the object the target object is aiming at or attacking (i.e., associated objects) at each sampling moment to construct an operation state sequence for cheating detection, rather than focusing solely on the state information of the target object itself. This is equivalent to introducing richer data for cheating detection, which helps improve the accuracy of cheating detection.

[0041] In one possible implementation, embodiments of this application can be applied to a data processing system. See also... Figure 1 , Figure 1 This is a schematic diagram of the structure of a data processing system provided in an embodiment of this application. For example... Figure 1As shown, the data processing system may include a client and a server. The client provides game services to users, allowing them to play games. The client can collect player action data during gameplay and send it to the server for cheat detection. The server receives the action data from the client and uses the aforementioned data processing scheme to perform cheat detection based on the received data, obtaining a cheat detection result. Optionally, after obtaining the cheat detection result, if the result indicates that a player has cheated, a penalty mechanism can be triggered to ensure game fairness. Optionally, if the cheat detection result indicates that a player has cheated, a notification message can be sent to each player's client in the game, indicating that a player has cheated and has been punished, thus improving the user experience for all players. In one embodiment, a security policy can be determined based on the cheat detection result. This security policy optimizes the anti-cheating function of the game application to ensure game security. For example, by counting the number of players cheating in a game application, and then determining corresponding security strategies based on the range in which the number of cheaters falls, the security of the game can be improved and the occurrence of cheating can be reduced. It is understood that this application constructs a state sequence for cheat detection operations based on the state information of the target object and the objects it is targeting or attacking at each sampling moment, rather than focusing solely on the state information of the target object itself. This introduces richer data for cheat detection, helping to improve the accuracy of cheat detection and thus ensuring the fairness of the game.

[0042] In one possible implementation, different types of cheating tools and functions may exist for different games. Therefore, this application embodiment can adaptively acquire appropriate operational state characteristics for different types of cheating tools to more accurately detect cheating behavior, improve the generalization ability of cheat detection, and reduce long-term operating costs. Optionally, this application embodiment mainly uses cheat detection in first-person shooter games as an example for introduction.

[0043] For example, see Figure 2 , Figure 2 This is a schematic diagram illustrating the effect of a game scene provided in this application. For example... Figure 2 As shown, this game interface is the one with the cheat tool enabled. There are obstacles in front of the target object, but by enabling the cheat function, the game object behind the obstacles can be located. For example... Figure 2The box indicated by 201 in the image is used to locate the position of the player corresponding to player name 4. The game interface may also include cheat function options that the game object can select after installing cheat tools, such as... Figure 2 The option 202 in the diagram may include cheat function options 1-8. These options could include whether to enable the aiming frame, whether to display enemy player information, whether to display enemy player health, whether to display the game's mission antenna, whether to enable auto-aim, the key bindings for auto-aim, the auto-aim target, the auto-aim range, the auto-aim speed, etc. There are no restrictions on these options. It is understood that if the target has enabled the auto-aim cheat function, the target being aimed at will be automatically moved to the attack target. Therefore, during subsequent cheat detection, the target's aiming information can be obtained as status information.

[0044] For example, please see Figure 3 , Figure 3 This is a schematic diagram illustrating the effect of a game scene provided in an embodiment of this application. For example... Figure 3 As shown, this game interface depicts a scenario where attackers target players through obstructions. It can be seen that... Figure 3 Game objects behind occlusions (such as...) Figure 3 As shown in 301, the target object was attacked. It is understandable that if the target object aims at and attacks game objects when there are obstructions, it means that the target object is very likely to have enabled a cheat function that allows it to see through objects. Therefore, when performing cheat detection, information on whether there are obstructions between the target object and related objects can be obtained as a status information.

[0045] In one possible implementation, this application can be applied to the field of artificial intelligence (AI). AI is the theory, methods, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine capable of reacting in a manner similar to human intelligence. AI also studies the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities.

[0046] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0047] In one possible implementation, this application can also be applied to the field of cloud technology. Cloud technology refers to a hosting technology that unifies a series of resources such as hardware, software, and networks within a wide area network or local area network to realize the computation, storage, processing, and sharing of data.

[0048] Cloud technology is a collective term for network technologies, information technologies, integration technologies, management platform technologies, and application technologies applied to the cloud computing business model. It can form resource pools, providing flexible and convenient on-demand access. Cloud computing technology will become a crucial support. Backend services of technical network systems require substantial computing and storage resources, such as video websites, image websites, and many portal websites. With the rapid development and application of the internet industry, every item may have its own identification mark in the future, requiring transmission to backend systems for logical processing. Data at different levels will be processed separately, and various industry data will all require robust system support, which can only be achieved through cloud computing.

[0049] It should be noted that this application may display prompt interfaces, pop-ups, or output voice prompts before and during the collection of user data. These prompt interfaces, pop-ups, or voice prompts are used to inform the user that their data is being collected. This ensures that the application only begins the steps for collecting user data after receiving confirmation from the user regarding the prompt interface or pop-up; otherwise (i.e., without user confirmation), the steps for collecting user data end, meaning no user data is collected. In other words, all user data collected in this application is collected with the user's consent and authorization, and the collection, use, and processing of related user data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0050] The technical solution of this application can be applied to electronic devices, such as the data processing device mentioned above. This electronic device can be a terminal, a server, or other devices used for data processing; this application does not limit the scope. Optionally, the server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Terminals include, but are not limited to, mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle terminals, aircraft, smart speakers, and other smart home appliances.

[0051] It is understood that the above scenarios are merely examples and do not constitute a limitation on the application scenarios of the technical solutions provided in the embodiments of this application. The technical solutions of this application can also be applied to other scenarios. For example, as those skilled in the art will know, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0052] Based on the above description, this application proposes a data processing method. Please refer to [link to relevant documentation]. Figure 4 , Figure 4 This is a schematic flowchart illustrating a data processing method provided in an embodiment of this application. The method can be executed by the aforementioned electronic device. The data processing method may include the following steps.

[0053] S401. Obtain the operation data to be detected for the target object.

[0054] The target object can be a game player (also known as a player, etc.) participating in the game.

[0055] In one possible implementation, cheat detection can be performed on each game object participating in the game, that is, each game object participating in the game can be taken as the target object, and the corresponding operation data to be detected can be obtained respectively.

[0056] In one possible implementation, cheat detection can be performed only on a subset of the game objects participating in the game. That is, the subset of game objects participating in the game can be taken as the target objects, and the corresponding operation data to be detected can be obtained separately.

[0057] In one embodiment, the target object can be determined by: acquiring the game result features of each game object participating in the game; if a game result feature that satisfies the cheating detection condition is detected, then the game object corresponding to the game result feature that satisfies the cheating detection condition is determined as the target object. The game result feature can characterize the performance of the game object in the game. For example, the game result feature may include one or more of the following: the player's ranking (e.g., 1st place, 2nd place, etc.), the number of players defeated by the player, the damage output by the player, the damage received by the player, a score calculated from information such as the number of players defeated and the damage output or received, etc., etc., without limitation.

[0058] The cheat detection conditions refer to the conditions that the game result characteristics must meet when performing cheat detection on game objects. For example, the cheat detection conditions can be one or more of the following: the player's ranking is higher than or equal to the target ranking (such as 1st place, 3rd place, etc.), the number of players defeated by the player is greater than or equal to the player number threshold, the damage output by the player is greater than or equal to the output damage threshold, the damage received by the player is less than or equal to the received damage threshold, the score is greater than or equal to the score threshold, etc., without restrictions. It is understandable that, due to the influence of cheating tools, cheating players usually obtain better game result characteristics after the game ends. In other words, players with better game result characteristics have a higher probability of cheating. Therefore, after the game ends, game objects with better game result characteristics can be automatically detected to identify game objects whose game result characteristics meet the cheat detection conditions as target objects. This eliminates the need to perform cheat detection on every game object, reduces the amount of computation, and improves the efficiency of game settlement.

[0059] In one embodiment, the target object can also be determined by receiving a cheat detection request for the target game object and identifying the target game object indicated by the cheat detection request as the target object. The cheat detection request can be a request to perform cheat detection on the target game object, and this request can be sent by the client of any game object participating in the game. It is understood that during the game, if any game object suspects that another game object may be cheating, it can initiate a cheat detection request for that potentially cheating game object during or after the game, thereby performing cheat detection on the game object indicated by the cheat detection request (i.e., the target game object). This provides a cheat feedback function to game players, improving the gaming experience.

[0060] The operation data to be detected refers to the operation data used for cheat detection. This data includes the target object's operation data at T sampling times, where T is a positive integer. In a game, all of the target object's actions and changes in viewpoint generate corresponding game protocols, essentially log data of the target object's actions and viewpoint changes. Based on these game protocols, the target object's operation data at T sampling times can be obtained—the operation data to be detected.

[0061] In one embodiment, the aforementioned T sampling times can be determined based on a certain sampling interval. This sampling interval refers to the time difference between adjacent sampling times. For example, in a 10-minute game session, if the sampling interval is determined to be 0.5 seconds, then a sampling time can be determined every 0.5 seconds, resulting in 1200 sampling times. Operation data can be acquired at each sampling time, and the operation data to be detected can then be determined from the operation data of these 1200 sampling times. It is understood that this sampling interval can be flexibly set according to the actual application scenario, and is not limited here.

[0062] In one possible implementation, the timing for acquiring the operation data to be detected can be either after the game ends or during the game; there is no restriction here.

[0063] In some scenarios, it's possible to detect whether players engaged in cheating during a game after the match ends. For example, if any player suspects another player of cheating after the match, a cheat detection request can be initiated against that suspected player. Upon receiving the request, the system can acquire all or part of the suspected player's action data from the game as detection data, thus determining whether the suspected player actually cheated throughout the match. Alternatively, after the match ends, the system can acquire the match result characteristics of each player. If any match result characteristic meets cheat detection criteria, the player corresponding to that characteristic is identified as the target, and their action data from the game is acquired as detection data.

[0064] In some scenarios, cheating behavior can also be detected during gameplay. For example, each time a player defeats another player, the moment of the defeat and the player's actions within a certain timeframe prior to the defeat can be collected as detection data. This is because cheating often aims to defeat other players, making the moment of defeat crucial for detection. Alternatively, player action data can be collected at regular intervals for cheat detection. Furthermore, if any player suspects another player is cheating, a cheat detection request can be initiated during the game. Upon receiving this request, the action data of the suspected cheater before the request can be collected (all or part of the sampled data) to detect whether the suspected cheater engaged in cheating throughout the entire game.

[0065] In one embodiment, the operation data to be detected can be operation data from a portion of the sampled moments in a game match. Therefore, obtaining the operation data to be detected for the target object can specifically include the following steps: ① Obtaining the initial operation data of the target object. The initial operation data includes the operation data of the target object at M sampled moments, where M is a positive integer, and M is greater than or equal to T. For example, this initial operation data can be the operation data from all sampled moments in the game match.

[0066] ② If at least R consecutive unaimed sampling moments are detected out of M sampling moments, the operation data of at least R consecutive unaimed sampling moments in the M sampling moments are filtered to obtain the operation data to be detected. Here, an unaimed sampling moment is a sampling moment when the target object does not perform an aiming operation or attack operation. R can be a positive integer, and R can refer to the minimum number of consecutive unaimed sampling moments that can be filtered. This number of consecutive moments refers to the number of consecutive unaimed sampling moments. The value of R can be set according to the actual application scenario and is not restricted here. For example, if R is set to 30, and the number of consecutive unaimed sampling moments in the initial operation data is 10, then since 5 < 30, the operation data of this consecutive unaimed sampling moment segment will not be filtered; if the number of consecutive unaimed sampling moments in the initial operation data is 60, then since 60 > 30, the operation data of this consecutive unaimed sampling moment segment can be filtered. The filtering process can be to remove the operation data of at least R consecutive unaimed sampling moments from the initial operation data. It is understandable that filtering out sampling times without aiming is necessary because, in shooting games, the commonly used wallhack or auto-aiming tools are related to the player's aiming or attacking actions. Therefore, the operational data during sampling times without aiming has little impact on determining whether a player is cheating. Thus, when performing cheat detection, operational data during periods of no aiming beyond a certain time frame can be filtered out to reduce the amount of data that needs to be processed and improve detection efficiency.

[0067] S402. Determine the operation state characteristics corresponding to each sampling time based on the operation data of T sampling times.

[0068] The operation state features include the state information of the target object and the state information of its associated objects. Associated objects are those on which the target object performs aiming or attacking operations. The associated objects can differ at different sampling times. The target object's state information can describe its state, and the associated object's state information can describe its state. Therefore, the operation data at each sampling time records all operations and field-of-view changes of the game object. Based on the operation data to be detected, the operation state features corresponding to each sampling time can be determined. This allows for the extraction of various types of operation state features that may be related to cheating behavior from the operation data to be detected, thus enabling the detection of cheating behavior by game players based on these rich operation state features.

[0069] In one embodiment, if a target object may aim at or attack multiple objects in a single sampling moment, then the game object closest to the target object among the multiple objects that the target object may aim at or attack can be used as the associated object of the target object.

[0070] In one embodiment, the target object's state information includes one or more of the following: the target object's position information, the target object's action information, the target object's aiming target information, and the target object's health information. The target object's position information can be its position in the game's reference coordinate system, or it can be its position change information relative to the previous sampling time; there is no limitation here. The target object's speed information can be its corresponding movement speed. It is understood that since defeating a player is more difficult at higher movement speeds, if a target object defeats an associated object at high speed, the likelihood of cheating is higher. The target object's action information can be information about the actions performed by the target object, such as the target object attacking, the target object aiming, the number of attacks, etc., there is no limitation here. The target object's aiming target information can refer to the information of the target object being aimed at, such as information about other game objects being aimed at, or information about the object being aimed at, etc., there is no limitation here. The target object's health information can refer to the maximum damage the target object can withstand in the game.

[0071] Understandably, the target object's state information in the defined operational state characteristics may differ depending on the type of shooting game or the different cheating tools involved. For example, in some types of games, there may be cheating tools that increase a player's economic value, allowing them to quickly accumulate wealth and acquire more powerful equipment. Therefore, for this type of game and cheating tool, the target object's economic value can be included as one type of target object state information in the operational state characteristics. This allows for flexible setting of the required state information in the operational state characteristics to adapt to different cheating detection scenarios, thereby improving the accuracy and flexibility of cheating detection.

[0072] In one embodiment, the state information of the associated object of the target object includes one or more of the following: whether the associated object is occluded, the position information of the associated object, the speed information of the associated object, the action information of the associated object, and the health information of the associated object. Whether the associated object is occluded can indicate whether there is an obstruction between the associated object and the target object. It is understood that an obstruction prevents the target object from seeing the game object behind it. Normally, the target object will not and find it difficult to aim at other players through obstructions. Therefore, if the target object aims at a game object obscured by an obstruction, it indicates a higher probability that the target object is cheating. The position information of the associated object can be the position of the associated object in the game's reference coordinate system, or it can be the position change information of the target object relative to the previous sampling time. Alternatively, the position information of the associated object can be the distance between the associated object and the target object, or it can be the distance change information between the associated object and the target object relative to the previous sampling time, etc., without limitation. The speed information of the associated object can be the movement speed of the associated object. It is understood that since it is more difficult to defeat a player with a higher movement speed, if the target object defeats the associated object when the associated object is moving quickly, the probability of the target object cheating is higher. The action information of the associated object can be information about the actions performed by the associated object, such as the associated object attacking, the associated object aiming, the associated object running, the number of times the associated object has attacked, etc., and there are no restrictions here. The health information of the associated object can refer to the maximum amount of damage that the associated object can withstand in a game.

[0073] In one embodiment, if the target object is not a game object that is being targeted or attacked, then the state information of the associated object of the target object in the state feature data at the corresponding sampling time is empty.

[0074] S403. Arrange the operation state features corresponding to each sampling time in chronological order and combine them to obtain the operation state sequence of the target object.

[0075] The operation state sequence can be a sequence representing the operation state features of the target object at T sampling times. It is understandable that the operation state sequence obtained from normal gameplay data and the operation state sequence obtained from cheating data are distinguishable in terms of data performance, thus allowing for subsequent cheat detection based on the operation state sequence.

[0076] In one embodiment, the operation state sequence can be represented as a feature matrix. The feature matrix can have a dimension of T*D, where D represents the feature dimension of the operation state feature corresponding to each sampling moment. That is, each row in the feature matrix corresponds to a sampling moment, and each row of data represents the operation state feature at that sampling moment. The operation state features include D types of state information, and each column of data corresponds to one type of state information among the operation state features. It can be understood that each value in the feature matrix represents the state value indicated by the sampling moment corresponding to its row and the type of state information corresponding to its column. This state value can be a specific numerical value (such as the magnitude of speed, the magnitude of health, the number of attacks, etc.), or it can be an identifier indicating different options for a type of state information (such as an identifier indicating whether it is occluded, an identifier indicating different action types, etc.), without limitation. In one embodiment, the order of the state information corresponding to each column of data in the operation state sequence can be arbitrarily combined. For example, the state information of the target object can be placed before the state information of the associated object, or the state information of the associated object can be placed before the state information of the target object, or they can be arranged according to the importance of each state information. The specific arrangement can be set according to the actual application scenario, without limitation.

[0077] For example, the state value indicated by the location information of the target object can be the position coordinate value of the target object; the state value indicated by the speed information of the target object can be the specific speed value of the target object; the action information of the target object can be an action type identifier corresponding to each action type, such as 0 for aiming, 1 for attacking, 2 for running, and so on, to represent multiple types of actions; the state value indicated by the state information of whether the associated object of the target object is occluded can be an identifier corresponding to occluded and unoccluded, such as 1 for occluded and 0 for unoccluded; the aiming target information of the target object can be the identifier of the target that the target object is aiming at.

[0078] This section describes the method for generating operation state sequences with illustrations. For example, please refer to [link to documentation / reference]. Figure 5 , Figure 5 This is a schematic diagram illustrating the effect of an operation state sequence provided in an embodiment of this application. For example... Figure 5 As shown, it is possible to obtain the target object's operation data during the game (such as...). Figure 5 As shown in 501), an operation state sequence can then be generated based on the acquired operation data (such as...). Figure 5 As shown in 502), it can be seen that in Figure 5In this sequence, each row of data corresponds to the state information at a sampling time, thus arranging the state information at each sampling time in chronological order. Each column of data corresponds to a type of state information, such as feature 1, feature 2, feature n, etc., without restriction. For any data in the operation state sequence, it can represent the state value indicated by the target object at the sampling time corresponding to its row and the corresponding type of state information in its column. For example, for... Figure 5 The data shown in 503 can represent the state value indicated by the target object for the state information of type "feature 2" at the first sampling time.

[0079] S404. Determine the cheating detection result of the target object based on the operation state sequence.

[0080] The cheating detection result is used to indicate whether the target object has cheated.

[0081] In one possible implementation, determining the cheating detection result of the target object based on the operation state sequence can be achieved by: determining classification features based on the operation state sequence, and determining the cheating detection result based on the classification features. The classification features can be features used to ultimately classify whether cheating behavior exists. Determining the classification features based on the operation state sequence can be achieved by extracting classification features through some neural networks, and determining the cheating detection result based on the classification features can be achieved through a fully connected layer.

[0082] In one possible implementation, this application embodiment can invoke a cheat detection model to process the operation state sequence to obtain the cheat detection result of the target object. Specifically, the cheat detection model can be invoked to process the operation state sequence to obtain the cheating probability of the target object, and then the cheat detection result of the target object can be determined based on the cheating probability of the target object. The cheating probability can refer to the probability that the target object has cheating behavior. The cheat detection model can be some neural network model, such as CNN (Convolutional Network Model), TextCNN (a convolutional network model), ATT-CNN (a convolutional network model), etc., without limitation. Thus, the cheat detection model can quickly identify whether there is cheating behavior in the operation data corresponding to the operation state sequence. This is because the operation state sequence obtained from normal game operation data is different from the operation state sequence obtained from cheating. After training, the cheat detection model can mine some potential features of cheating operation state sequences based on the operation state sequence. Compared with judging whether cheating is detected based on some numerical values ​​of statistical operation data and artificially set rules, this application can perform cheat detection based on the operation state sequence with higher accuracy.

[0083] By employing the embodiments of this application, it is possible to acquire the operation data to be detected of the target object, such as the operation data of the target object in a game. Based on this operation data, the operation state characteristics corresponding to each sampling moment can be determined. These operation state characteristics can include the state information of the target object and the object the target object is aiming at or attacking, thus obtaining operation state characteristics with rich information. An operation state sequence is then generated based on the operation characteristics at each sampling moment, and the cheating detection result is determined based on the operation state sequence to determine whether the target object is engaging in cheating behavior. Therefore, the operation state sequence can be determined through the state information of the target object and associated objects. In particular, the state information of the object the target object is aiming at or attacking (i.e., associated objects) is introduced at each sampling moment to construct an operation state sequence for cheating detection, rather than being limited to focusing on some state information of the target object itself. This is equivalent to introducing richer data for cheating detection, which helps improve the accuracy of cheating detection.

[0084] Please see Figure 6 , Figure 6 This is a schematic flowchart illustrating a data processing method provided in an embodiment of this application. The method can be executed by the aforementioned electronic device. The data processing method may include the following steps.

[0085] S601. Obtain the operation data to be detected for the target object.

[0086] S602. Determine the operation state characteristics corresponding to each sampling time based on the operation data of T sampling times.

[0087] S603. Arrange the operation state features corresponding to each sampling time in chronological order and combine them to obtain the operation state sequence of the target object.

[0088] The descriptions of steps S601-S603 can be found in the descriptions of steps S401-S403 above, and will not be repeated here.

[0089] S604. Call the convolutional network layer of the cheating detection model to process the operation state sequence and obtain the convolutional features corresponding to the operation state sequence.

[0090] The convolutional feature can be a feature obtained through a convolutional network layer.

[0091] In one possible implementation, to perform cheating detection more accurately, the cheating detection model may include neural convolutional network layers and attention mechanism network layers, thereby enabling the cheating detection model to better uncover the deep features of the operation state sequence and improve the accuracy of cheating detection.

[0092] For example, the cheating detection model can be an ATT-CNN model containing neural convolutional network layers and attention mechanism network layers. The ATT-CNN model is usually used for text detection scenarios. Here, the ATT-CNN model is used in the cheating detection model, which is equivalent to the transfer of the model's application scenario. In order to be more suitable for this cheating application scenario, the ATT-CNN model is adaptively adjusted to obtain a cheating detection model suitable for the cheating detection scenario.

[0093] In one possible implementation, the ATT-CNN model achieves better detection performance compared to other models because it incorporates neural convolutional network layers and attention mechanism network layers. For example, please refer to the table below (Table 1), which compares the detection performance of several models.

[0094]

[0095] Table 1

[0096] Accuracy represents the proportion of samples predicted as positive that are actually positive. Coverage represents the proportion of samples that are actually positive that are predicted as positive. Therefore, the cheating detection model incorporating neural convolutional network layers and attention mechanism network layers achieves better detection results.

[0097] In one embodiment, the convolutional network layer can perform convolution processing on the input operation state sequence. In another embodiment, the convolutional network layer may include multiple convolutional layers, each using a different kernel size. The convolutional feature can be the feature obtained after processing the operation state sequence through the convolutional network layer. Optionally, this application can use one-dimensional convolution in the time dimension to improve the correlation between consecutive sampling times in the operation state sequence; the kernel size of the one-dimensional convolution determines the time span of the correlation between consecutive times in the sequence. Optionally, the Text-CNN (a neural network model) model, typically used for text detection, can employ convolutional layers with kernel sizes of 2, 3, or 4. To adapt to cheating detection scenarios, considering the duration of defeating an opponent in typical action games (such as FPS games), statistical analysis has shown that a combination of convolutional layers with kernel sizes covering 2, 4, 6, and 8 seconds of action data is generally preferred. This combination covers the majority of the duration of defeating an opponent, allowing for the association of action state features within that duration and effectively detecting cheating behavior over a continuous timeframe. This is because, assuming a person continuously aims at a game object through an obstruction for a long period, a relatively large convolutional kernel increases the likelihood of cheating. For example, the probability of a target cheating is higher when aiming at an object through an obstruction at 8-second intervals than when aiming at the object through an obstruction at 4 consecutive 2-second intervals. Therefore, setting a convolutional layer with an 8-second kernel size enhances the correlation between action data over a longer time span, contributing to improved accuracy in cheating detection.

[0098] Understandably, in shooting games (such as FPS games), a player's actions are not isolated; the changes in the sequence of operational states have a certain continuity. For example, a player might continuously move to approach a target, move the crosshair to aim, and launch an attack—a complete operation spanning several seconds. If a sequence-independent fully connected neural network is used for training, the operational data between each sampling moment lacks interaction. The fully connected neural network only operates on the operational state features of a single sampling moment, failing to capture the correlation between each sampling moment in the operational state sequence. Therefore, using convolutional network layers here enhances the connection between operational data at different sampling moments, helping to improve the accuracy of cheat detection.

[0099] S605. Call the attention mechanism network layer of the cheating detection model to process the convolutional features and obtain classification features.

[0100] The classification feature can be a feature used to ultimately classify whether cheating behavior exists. Optionally, the classification feature can be a feature output by the attention mechanism network layer, or a feature processed based on the feature output by the attention mechanism network layer; there is no restriction here.

[0101] This attention mechanism, also known as the attention mechanism, enables neural networks to focus on a subset (or feature) of their input, allocating limited information processing resources to important parts. In other words, by processing convolutional features through attention mechanism network layers, attention can be focused on specific important sampling moments, such as the sampling moment when a target defeats a player. This is because aimless player movement in a game has little effect on cheat detection, but whether the data at the moment of defeating a player is abnormal is crucial. Although convolutional network layers enhance sequence correlation by setting kernel sizes of 2, 4, 6, and 8, they cannot highlight the connection between key sampling moments (also called defeat frames, keyframes, etc.) that are far apart in time. Key sampling moments refer to the sampling moments when a target defeats other game objects. All keyframes themselves can serve as evidence for judging cheating. For example, a key sampling moment could be the sampling moment of defeating an enemy player in close combat, the sampling moment of continuously aiming to defeat an enemy player, or the sampling moment of instantly defeating an enemy player; there are no restrictions here. Therefore, the attention mechanism allows the model to weight these keyframes, enabling it to better extract key features and improve the accuracy of cheat detection.

[0102] In one embodiment, the Attention mechanism learns to focus attention on specific important features when deriving contextual sequence features from the input feature sequence (e.g., convolutional features in this case). When predicting whether cheating occurs in an operational state sequence, the Attention mechanism can determine which sampling moments should receive more attention than other sampling moments in the operational state sequence. The attention output features of the Attention mechanism are obtained by weighted summing and combining the scores of data from different sampling moments in the operational state sequence. For example, the attention output features of the Attention mechanism can be obtained using the following formula (as shown in Equation 1).

[0103] g i =∑ j＝i α i，j ·x j Formula 1

[0104] Where i represents any row of the input features in the Attention mechanism network layer, and j represents any column of the input features in the Attention mechanism network layer. α is called the Attention weight, then α i，j Let represent the weight corresponding to the data in the i-th row and j-th column. Where α ≥ 0 and ∑ j α i，j =1. x j This represents the weighted average of the j-th column of the input features of the Attention mechanism network layer. i This represents the i-th row of data representing the attention output features.

[0105] In one embodiment, α i，j Normalization can be achieved using softmax (a normalization method), and α can be calculated using the following formula (e.g., formula 2). i，j .

[0106]

[0107] Among them, score(x) i x j () represents the score corresponding to the data in the i-th row and j-th column, x i x represents the weighted average of the i-th row of input features of the Attention mechanism network layer. j This represents the weighted average of the j-th column of the input features of the Attention mechanism network layer.

[0108] In one embodiment, there are many ways to calculate the score (also known as vector matching degree) in the Attention mechanism. For example, the additive Attention score can be used, that is, the score can be calculated using the following formula (such as Formula 3).

[0109] score(x i x j ) = v a T tanh(W1x i +W2x j ]) Formula 3

[0110] Here, W1 and W2 can be the trainable weights in the Attention mechanism. a T It is a parameter that can be obtained through training. From this, we can obtain x. i With x j The score between the two scores (x) i x j ).

[0111] It is understandable that the attention output features processed by the Attention mechanism network layer are obtained by weighted summation of all sampling moments based on correlation. The operation state sequence containing global information represents the player's behavior at different sampling moments. Among them, obvious cheating behavior occurs at certain moments, and the Attention mechanism network layer is suitable for capturing these cheating moments.

[0112] In one embodiment, this application embodiment can also introduce the time position feature at each sampling moment, which is equivalent to introducing the positional relationship between each row of data in the input features of the Attention mechanism network layer and other rows of data. For example, the corresponding time position feature is concatenated onto each row of data in the input features (such as convolutional features) of the Attention mechanism network layer. This is because the above-mentioned Attention structure alone cannot capture the order of the operation state sequence. If the operation state sequence is shuffled row by row, the result of Attention is still the same. Position Embedding (position vector) numbers each position and corresponds to a vector. For example, by combining the time position vector with each row of data in the convolutional features, or by combining the time position vector with each row of data in the operation state sequence, certain positional information is introduced into the input features of the Attention mechanism network layer, so that Attention can distinguish data at different positions. For example, the following formula (such as Formula 4) can be used to construct the time position vector.

[0113]

[0114] Where q represents any column of data from which the time-position vector needs to be constructed, and p represents any row of data from which the time-position vector needs to be constructed. Optionally, the features from which the time-position vector needs to be constructed can be convolutional features or operation state sequences, etc. PE 2q (P) represents the value corresponding to the even-numbered column in the time-position vector, PE 2q (P) represents the value corresponding to the odd-numbered columns in the time-position vector. d pos The total number of columns of features needed to construct the time position vector.

[0115] In one embodiment, Position Embedding itself is information about absolute position, but relative position is also important in sequence data. Since sin(α+β) = sinαcosβ + cosαsinβ and cos(α+β) = cosαcosβ - sinαsinβ, this shows that the vector at position p+k can be represented as a linear transformation of the vector at position p, thereby obtaining relative position information.

[0116] This section introduces the attention mechanism network layer with illustrations. For example, please refer to [link to example]. Figure 7 , Figure 7 This is a schematic diagram of the structure of an attention mechanism network layer provided in an embodiment of this application. Figure 7 As shown, this attention network layer can include an input layer to characterize features of the input attention mechanism, such as the convolutional features mentioned above. The attention network layer can also include hidden layers, which can include some trained model parameters, such as... Figure 7 The values ​​w1 and w2 in the model can be used for further normalization and weighted summation based on softmax (a normalization method) to obtain the attention output features.

[0117] In one embodiment, after deriving the attention distribution map of the attention mechanism network layer, the weights corresponding to each sampling time in the attention mechanism network layer can be seen, and the sampling time corresponding to the high weight can be the key sampling time. For example, please refer to... Figure 8 , Figure 8 This is a schematic diagram illustrating the effect of an attention distribution map provided in an embodiment of this application. It can be seen that the high weights in the attention mechanism network layer can be respectively as follows: Figure 8 The weights shown in 801, 802, and 803 correspond to the moments of aiming down sights, engaging in firefights, and aiming from a position, respectively. Therefore, when performing cheat detection, the attention mechanism network layer of the cheat detection model can assign higher weights to keyframes. This is because keyframes have greater reference value in the cheat detection process, thereby improving the accuracy of cheat detection.

[0118] S606. Determine the cheating probability of the target object based on the classification features, and determine the cheating detection result of the target object based on the cheating probability.

[0119] As mentioned above, the probability of cheating can refer to the probability that the target object engages in cheating behavior.

[0120] Optionally, if the cheating probability is greater than or equal to a threshold, the cheating detection probability can be determined to indicate that the target object has cheating behavior; if the cheating probability is less than a threshold, the cheating detection probability can be determined to indicate that the target object has not cheating behavior.

[0121] In one possible implementation, before calling the cheat detection model to process the operation state sequence, the cheat detection model can be trained first, so that it can accurately predict whether cheating behavior exists in the game based on operation data over a period of time. Training the cheat detection model can then include the following steps:

[0122] ① Obtain sample operation data and determine the corresponding sample operation state sequence based on the sample operation data. The sample operation data has sample labels, which indicate whether cheating behavior exists in the sample operation data. This sample operation data can be data used to train a cheat detection model, or it can be operation data of game objects obtained from actual game scenarios. The method for determining the corresponding sample operation state sequence based on the sample operation data can refer to the method described above for determining the corresponding operation state sequence based on the operation data to be detected, and will not be repeated here. In one embodiment, if the sample operation data contains cheating operations, the sample label can be 1; if the sample operation data does not contain cheating operations, the sample label can be 0.

[0123] ② The convolutional network layer of the initial cheat detection model is invoked to process the sample operation state sequence, obtaining the sample convolutional features corresponding to the sample operation state sequence. This initial cheat detection model refers to the untrained cheat detection model; the model parameters between the convolutional network layers of the initial cheat detection network and the convolutional network layers of the actual cheat detection network can be different. The method for generating sample convolutional features can refer to the relevant description of obtaining convolutional features above, and will not be repeated here.

[0124] ③ The attention mechanism network layer of the initial cheating detection model is invoked to process the convolutional features of the samples, obtaining sample classification features, and the sample cheating probability is determined based on the sample classification features. The model parameters of the attention mechanism network layer of the initial cheating detection network and the convolutional attention mechanism network layer of the cheating detection network can be different. The method for generating sample convolutional features can refer to the relevant description of obtaining convolutional features above, and will not be repeated here. The method for generating sample cheating probabilities can refer to the relevant description of obtaining cheating probabilities above, and will not be repeated here.

[0125] ④ Based on the sample cheating probability and sample labels, the model parameters of the initial cheating detection model are corrected to obtain the cheating detection model. In one embodiment, correcting the model parameters of the initial cheating detection model can be done by calculating the loss value based on the sample cheating probability and sample labels, and then correcting the model parameters of the initial cheating detection model based on the loss value, so that the loss value gradually decreases until convergence. During the process of correcting the model parameters of the initial cheating detection model, the sample cheating probability and the numerical value indicating the existence of cheating behavior indicated by the sample labels should gradually become closer. This allows the trained cheating detection model to accurately predict whether cheating behavior exists through the operation state sequence. Furthermore, since an attention mechanism is introduced into this cheating detection model, the accuracy of cheating detection is greatly improved.

[0126] This section illustrates the entire cheat detection process with illustrations. For example, please see... Figure 9 , Figure 9This is a schematic diagram of a cheating detection process provided in an embodiment of this application. First, the operation data to be detected (such as...) can be obtained. Figure 9 As shown in 901), for example, operation data 1 corresponding to sampling time 1, operation data 2 corresponding to sampling time 2, operation data n corresponding to sampling time n, etc. can be obtained. That is, the operation data to be detected can include operation data from n consecutive sampling times. Furthermore, a state feature sequence (such as...) can be generated based on the obtained operation data to be detected. Figure 9 As shown in 902 (in the diagram), the operation state features corresponding to each sampling time point include multiple types of state information, such as state information 1, state information 2, state information 3, state information 4, etc. Therefore, the multiple types of state information corresponding to each sampling time point can be combined into a state feature sequence. This state feature sequence can then be processed through a convolutional network layer (such as...). Figure 9 The convolutional features are obtained by processing (as shown in 903), and then input into the attention mechanism network layer (such as...). Figure 9 As shown in 904), the classification features are obtained, and finally the cheating detection result is obtained (e.g., ...). Figure 9 (As shown in 905). Thus, cheating detection of the target object can be achieved through the operation state sequence with rich state information. Furthermore, since convolutional network layers and attention mechanism network layers are used in the cheating detection process, the deep features of the operation state sequence can be better mined, thereby improving the accuracy of cheating detection.

[0127] In one possible implementation, the aforementioned classification features can also be features obtained by processing the attention output features based on the attention mechanism network layer. Optionally, the attention output features output by the attention mechanism network layer can be concatenated with statistical features to obtain classification features. These statistical features can be characteristics of the target object in the game, such as the total number of players defeated by the target object, the number of times the target object performs attack operations, the number of times the target object dies, etc., without limitation. This is because some statistical features can also indicate whether the target object is cheating. For example, if the number of attacks by the target object in the game is close to the total number of players defeated, it indicates that the target object defeated players with each attack, which is highly difficult in the game, thus increasing the likelihood of cheating. Optionally, the attention output features output by the attention mechanism network layer can also be pooled, and the pooled attention output features can be concatenated with statistical features to obtain classification features.

[0128] This section illustrates the entire cheat detection process with illustrations. For example, please see... Figure 10 , Figure 10This is a schematic diagram of a cheating detection process provided in an embodiment of this application. First, the operation data to be detected (such as...) can be obtained. Figure 10 As shown in 1001), a state feature sequence (such as...) can be generated based on the acquired data of the operation to be detected. Figure 10 (As shown in 1002). Then the state feature sequence can be passed through a convolutional network layer (such as...). Figure 10 The convolutional features are obtained by processing (as shown in 1003), and then input into the attention mechanism network layer (such as...). Figure 10 As shown in 1004), the attention output features are obtained, and the attention output features are then subjected to weighted pooling (e.g., ...). Figure 10 (As shown in 1005). Furthermore, statistical features (such as...) can be obtained based on the data of the operation to be detected. Figure 10 As shown in 1006), the statistical features are concatenated with the attention output features after pooling (e.g., Figure 10 As shown in 1007 (in the example), the final classification features are obtained. This allows the classification to be processed by a classifier (such as...). Figure 10 As shown in Figure 1008, the classification features are processed to obtain the final cheating detection result. Thus, cheating detection of the target object can be achieved through operation state sequences with rich state information. Furthermore, since convolutional network layers and attention mechanism network layers are used in cheating detection, the deep features of the operation state sequences can be better mined, improving the accuracy of cheating detection.

[0129] By employing the embodiments of this application, it is possible to acquire the operation data to be detected of the target object, such as the operation data of the target object in a game. Based on this operation data, the operation state characteristics corresponding to each sampling moment can be determined. These operation state characteristics can include the state information of the target object and the object the target object is aiming at or attacking, thus obtaining operation state characteristics with rich information. An operation state sequence is then generated based on the operation characteristics at each sampling moment, and the cheating detection result is determined based on the operation state sequence to determine whether the target object is engaging in cheating behavior. Therefore, the operation state sequence can be determined through the state information of the target object and associated objects. In particular, the state information of the object the target object is aiming at or attacking (i.e., associated objects) is introduced at each sampling moment to construct an operation state sequence for cheating detection, rather than being limited to focusing on some state information of the target object itself. This is equivalent to introducing richer data for cheating detection, which helps improve the accuracy of cheating detection.

[0130] Please see Figure 11 , Figure 11 This is a schematic diagram of a data processing apparatus provided in an embodiment of this application. Optionally, this data processing apparatus can be disposed in the aforementioned electronic device. For example... Figure 11As shown, the data processing apparatus described in this embodiment may include:

[0131] Acquisition unit 1101 is used to acquire the operation data to be detected of the target object, wherein the operation data to be detected includes the operation data of the target object at T sampling times, where T is a positive integer;

[0132] Processing unit 1102 is configured to determine the operation state features corresponding to each sampling time based on the operation data of the T sampling times; the operation state features include the state information of the target object and the state information of the associated object of the target object, wherein the associated object is the object to which the target object performs the aiming operation or attack operation;

[0133] The processing unit 1102 is also used to combine the operation state features corresponding to each sampling time in chronological order to obtain the operation state sequence of the target object;

[0134] The processing unit 1102 is further configured to determine the cheating detection result of the target object based on the operation state sequence, wherein the cheating detection result is used to indicate whether the target object has cheating behavior.

[0135] In one implementation, the processing unit 1102 is specifically used for:

[0136] The operation state sequence is processed by the convolutional network layer of the cheating detection model to obtain the convolutional features corresponding to the operation state sequence;

[0137] The attention mechanism network layer of the cheating detection model is invoked to process the convolutional features to obtain classification features;

[0138] The cheating probability of the target object is determined based on the classification features, and the cheating detection result of the target object is determined based on the cheating probability.

[0139] In one implementation, the processing unit 1102 is further configured to:

[0140] Acquire sample operation data and determine the corresponding sample operation state sequence based on the sample operation data. The sample operation data has a sample label, which is used to indicate whether there is cheating behavior in the sample operation data.

[0141] The convolutional network layer of the initial cheating detection model is invoked to process the sample operation state sequence to obtain the sample convolutional features corresponding to the sample operation state sequence.

[0142] The attention mechanism network layer of the initial cheating detection model is invoked to process the convolutional features of the samples to obtain sample classification features, and the probability of cheating of the samples is determined based on the sample classification features.

[0143] The cheat detection model is obtained by correcting the model parameters of the initial cheat detection model based on the sample cheat probability and the sample label.

[0144] In one implementation, the state information of the target object includes one or more of the following: the position information of the target object, the speed information of the target object, the action information of the target object, the aiming target information of the target object, and the health information of the target object.

[0145] In one implementation, the state information of the associated objects of the target object includes one or more of the following: whether the associated object is occluded, the position information of the associated object, the speed information of the associated object, the action information of the associated object, and the life value information of the associated object.

[0146] In one implementation, the processing unit 1102 is specifically used for:

[0147] Acquire the initial operation data of the target object, which includes the operation data of the target object at M sampling times, where M is a positive integer and M is greater than or equal to T;

[0148] If at least R consecutive unaimed sampling moments are detected among the M sampling moments, the operation data of the at least R consecutive unaimed sampling moments among the M sampling moments are filtered to obtain the operation data to be detected; R is a positive integer, and the unaimed sampling moment is the sampling moment when the target object does not perform an aiming operation or an attack operation.

[0149] In one implementation, the processing unit 1102 is further configured to:

[0150] The game results features of each game object participating in the game are obtained. If a game result feature that satisfies the cheating detection condition is detected, the game object corresponding to the game result feature that satisfies the cheating detection condition is determined as the target object.

[0151] Alternatively, receive a cheat detection request for a target game object, and identify the target game object indicated by the cheat detection request as the target object.

[0152] Please see Figure 12 , Figure 12This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device described in this embodiment includes: a processor 1201 and a memory 1202. Optionally, the electronic device may also include a network interface or a power supply module, etc. The processor 1201 and the memory 1202 can exchange data.

[0153] The processor 1201 described above can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0154] The aforementioned network interface may include input devices and / or output devices. For example, the input device may be a control panel, microphone, receiver, etc., and the output device may be a display screen, transmitter, etc., which will not be listed here.

[0155] The aforementioned memory 1202 may include read-only memory and random access memory, and provides program instructions and data to the processor 1201. A portion of the memory 1202 may also include non-volatile random access memory. The processor 1201 executes the program instructions when it calls them:

[0156] Acquire the operation data to be detected of the target object, wherein the operation data to be detected includes the operation data of the target object at T sampling times, where T is a positive integer;

[0157] Based on the operation data at the T sampling times, the operation state characteristics corresponding to each sampling time are determined; the operation state characteristics include the state information of the target object and the state information of the associated object of the target object, wherein the associated object is the object to which the target object performs a targeting operation or an attack operation;

[0158] The operation state features corresponding to each sampling moment are combined in chronological order to obtain the operation state sequence of the target object.

[0159] The cheating detection result of the target object is determined based on the operation state sequence, and the cheating detection result is used to indicate whether the target object has cheating behavior.

[0160] In one implementation, the processor 1201 is specifically used for:

[0161] The operation state sequence is processed by the convolutional network layer of the cheating detection model to obtain the convolutional features corresponding to the operation state sequence;

[0162] The attention mechanism network layer of the cheating detection model is invoked to process the convolutional features to obtain classification features;

[0163] The cheating probability of the target object is determined based on the classification features, and the cheating detection result of the target object is determined based on the cheating probability.

[0164] In one implementation, the processor 1201 is further configured to:

[0165] Acquire sample operation data and determine the corresponding sample operation state sequence based on the sample operation data. The sample operation data has a sample label, which is used to indicate whether there is cheating behavior in the sample operation data.

[0166] The convolutional network layer of the initial cheating detection model is invoked to process the sample operation state sequence to obtain the sample convolutional features corresponding to the sample operation state sequence.

[0167] The attention mechanism network layer of the initial cheating detection model is invoked to process the convolutional features of the samples to obtain sample classification features, and the probability of cheating of the samples is determined based on the sample classification features.

[0168] The cheat detection model is obtained by correcting the model parameters of the initial cheat detection model based on the sample cheat probability and the sample label.

[0169] In one implementation, the state information of the target object includes one or more of the following: the position information of the target object, the speed information of the target object, the action information of the target object, the aiming target information of the target object, and the health information of the target object.

[0170] In one implementation, the state information of the associated objects of the target object includes one or more of the following: whether the associated object is occluded, the position information of the associated object, the speed information of the associated object, the action information of the associated object, and the life value information of the associated object.

[0171] In one implementation, the processor 1201 is specifically used for:

[0172] Acquire the initial operation data of the target object, which includes the operation data of the target object at M sampling times, where M is a positive integer and M is greater than or equal to T;

[0173] If at least R consecutive unaimed sampling moments are detected among the M sampling moments, the operation data of the at least R consecutive unaimed sampling moments among the M sampling moments are filtered to obtain the operation data to be detected; R is a positive integer, and the unaimed sampling moment is the sampling moment when the target object does not perform an aiming operation or an attack operation.

[0174] In one implementation, the processor 1201 is further configured to:

[0175] The game results features of each game object participating in the game are obtained. If a game result feature that satisfies the cheating detection condition is detected, the game object corresponding to the game result feature that satisfies the cheating detection condition is determined as the target object.

[0176] Alternatively, receive a cheat detection request for a target game object, and identify the target game object indicated by the cheat detection request as the target object.

[0177] Optionally, when the program instructions are executed by the processor, other steps of the method in the above embodiments can also be implemented, which will not be described in detail here.

[0178] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform the above-described method, such as the method performed by the above-described electronic device, which will not be elaborated here.

[0179] Optionally, the storage medium involved in this application, such as a computer-readable storage medium, may be non-volatile or volatile.

[0180] Optionally, the computer-readable storage medium may primarily include a stored program area and a stored data area. The stored program area may store the operating system, at least one application program required for a given function, etc.; the stored data area may store data created based on the use of blockchain nodes, etc. Here, the blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and cryptographic algorithms. A blockchain is essentially a decentralized database, a chain of data blocks linked using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain may include a blockchain underlying platform, a platform product service layer, and an application service layer, etc.

[0181] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0182] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0183] This application also provides a computer program product or computer program that includes computer instructions that, when executed by a processor, can implement some or all of the steps in the methods described above. For example, the computer instructions are stored in a computer-readable storage medium. The processor of a computer device (i.e., the aforementioned electronic device) reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps executed in the embodiments of the methods described above. For example, the computer device can be a terminal or a server.

[0184] The data processing method, apparatus, electronic device, and storage medium provided in the embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A data processing method, characterized in that, The method includes: The target object's operation data to be detected is obtained. The operation data to be detected includes the target object's operation data at T sampling times, where T is a positive integer. The operation data to be detected is obtained based on the game protocol generated during the game. The operation data at each sampling time records all operations and field of view changes of the target object. Based on the operation data at the T sampling times, the operation state characteristics corresponding to each sampling time are determined; the operation state characteristics include the state information of the target object and the state information of the associated object of the target object, wherein the associated object is the object to which the target object performs a targeting operation or an attack operation; The operation state features corresponding to each sampling moment are combined in chronological order to obtain the operation state sequence of the target object. The operation state sequence is processed by the convolutional network layer of the cheating detection model to obtain the convolutional features corresponding to the operation state sequence; the convolutional network includes multiple convolutional layers with convolutional kernels of different sizes, and the convolutional layers with convolutional kernels of different sizes are used to associate the operation state features of different time spans in the operation state sequence. The attention mechanism network layer of the cheat detection model is invoked to process the convolutional features to obtain classification features; the attention mechanism network layer is used to assign higher weights to the operation state features at key sampling moments in the operation state sequence, where the key sampling moment refers to the sampling moment when the target object defeats other game objects; The cheating probability of the target object is determined based on the classification features, and the cheating detection result of the target object is determined based on the cheating probability. The cheating detection result is used to indicate whether the target object has cheating behavior.

2. The method according to claim 1, characterized in that, The method further includes: Acquire sample operation data and determine the corresponding sample operation state sequence based on the sample operation data. The sample operation data has a sample label, which is used to indicate whether there is cheating behavior in the sample operation data. The convolutional network layer of the initial cheating detection model is invoked to process the sample operation state sequence to obtain the sample convolutional features corresponding to the sample operation state sequence. The attention mechanism network layer of the initial cheating detection model is invoked to process the convolutional features of the samples to obtain sample classification features, and the probability of cheating of the samples is determined based on the sample classification features. The cheat detection model is obtained by correcting the model parameters of the initial cheat detection model based on the sample cheat probability and the sample label.

3. The method according to claim 1, characterized in that, The target object's status information includes one or more of the following: the target object's position information, the target object's speed information, the target object's action information, the target object's aiming information, and the target object's health information.

4. The method according to claim 1, characterized in that, The status information of the associated objects of the target object includes one or more of the following: whether the associated object is occluded, the position information of the associated object, the speed information of the associated object, the action information of the associated object, and the health information of the associated object.

5. The method according to claim 1, characterized in that, The acquisition of the target object's detection operation data includes: Acquire the initial operation data of the target object, which includes the operation data of the target object at M sampling times, where M is a positive integer and M is greater than or equal to T; If at least R consecutive unaimed sampling moments are detected among the M sampling moments, the operation data of the at least R consecutive unaimed sampling moments among the M sampling moments are filtered to obtain the operation data to be detected; R is a positive integer, and the unaimed sampling moment is the sampling moment when the target object does not perform an aiming operation or an attack operation.

6. The method according to claim 1, characterized in that, The method further includes: Obtain the match result characteristics of each game object participating in the game match. If a match result characteristic that satisfies the cheating detection condition is detected, then the game object corresponding to the match result characteristic that satisfies the cheating detection condition is determined as the target object; or, Receive a cheat detection request for a target game object, and identify the target game object indicated by the cheat detection request as the target object.

7. A data processing apparatus, characterized in that, The device includes: The acquisition unit is used to acquire the operation data to be detected of the target object. The operation data to be detected includes the operation data of the target object at T sampling times, where T is a positive integer. The operation data to be detected is obtained based on the game protocol generated during the game. The operation data at each sampling time records all operations and field of view changes of the target object. The processing unit is configured to determine the operation state characteristics corresponding to each sampling time based on the operation data of the T sampling times; the operation state characteristics include the state information of the target object and the state information of the associated object of the target object, wherein the associated object is the object to which the target object performs a targeting operation or an attack operation; The processing unit is also used to combine the operation state features corresponding to each sampling time in chronological order to obtain the operation state sequence of the target object; The processing unit is further configured to invoke the convolutional network layer of the cheat detection model to process the operation state sequence, thereby obtaining convolutional features corresponding to the operation state sequence; the convolutional network includes multiple convolutional layers with convolutional kernels of different sizes, which are used to associate operation state features at different time spans in the operation state sequence; the attention mechanism network layer of the cheat detection model is invoked to process the convolutional features, thereby obtaining classification features; the attention mechanism network layer is used to assign higher weights to the operation state features at key sampling moments in the operation state sequence, where the key sampling moment refers to the sampling moment when the target object defeats other game objects; the cheating probability of the target object is determined based on the classification features, and the cheating detection result of the target object is determined based on the cheating probability, wherein the cheating detection result is used to indicate whether the target object has engaged in cheating behavior.

8. An electronic device, characterized in that, The system includes a processor and a memory, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is configured to invoke the program instructions to execute the method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method as described in any one of claims 1-6.

Images