Data processing method and related device

Abstract

The embodiment of the invention discloses a data processing method and a related device, and the method comprises the steps: obtaining playback data of a game, the playback data being used for restoring behaviors generated by a virtual character in the game and a game environment in which the virtual character is located; for a target virtual character in the game, generating a pheromone image corresponding to a reference moment of the target virtual character in the game according to the playback data; the pheromone image is used for representing game environment information perceived by the target virtual character at the reference moment; based on the corresponding pheromone image of the target virtual character at the reference moment, predicting a reference decision intention of the target virtual character at the target moment in the game; and comparing the reference decision intention with the actual decision intention at the target moment, and determining whether the target virtual character uses the plug-in or not according to a comparison result. According to the method, the game plug-in can be accurately and effectively detected.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular, to a data processing method and related devices. Background Art

[0002] With the rapid development of computer technology, e-sports games have now flourished. In the field of e-sports games, Multiplayer Online Battle Arena (MOBA) games have attracted more and more players due to their unique team collaboration mode and complex and diverse game mechanisms. However, the problem of game cheats has also emerged accordingly.

[0003] Game cheats refer to third-party software or programs that can modify the normal data and game logic in the game, and are often used to help players cheat in the game. On the one hand, game cheats will disrupt the game balance, causing players to lose interest in the game. On the other hand, they will also occupy a large amount of game system resources, reduce the network speed, and even cause the system to crash. It can be seen that how to accurately and effectively detect game cheats is an urgent problem to be solved at present. Summary of the Invention

[0004] Embodiments of this application provide a data processing method and related devices, which can accurately and effectively detect game cheats.

[0005] In view of this, the first aspect of this application provides a data processing method, and the method includes:

[0006] Obtain the replay data of the game session; the replay data is used to restore the behaviors and the game environment of the virtual characters in the game session;

[0007] For a target virtual character in the game session, generate a pheromone image corresponding to the reference moment of the target virtual character in the game session according to the replay data; the pheromone image is used to represent the game environment information perceived by the target virtual character at the reference moment;

[0008] Based on the pheromone image corresponding to the reference moment of the target virtual character, predict the reference decision intention of the target virtual character at the target moment in the game session; the target moment is after the reference moment;

[0009] Compare the reference decision intention of the target virtual character at the target moment with the actual decision intention of the target virtual character at the target moment, and determine whether the target virtual character uses cheats according to the comparison result.

[0010] The second aspect of this application provides a data processing device, and the device includes:

[0011] A data acquisition module, configured to acquire replay data of a game session; the replay data is used to restore the behaviors and the game environment of virtual characters in the game session;

[0012] A pheromone image generation module, configured to generate, for a target virtual character in the game session, a pheromone image corresponding to a reference moment in the game session according to the replay data; the pheromone image is used to represent the game environment information perceived by the target virtual character at the reference moment;

[0013] A decision intention prediction module, configured to predict a reference decision intention of the target virtual character at a target moment in the game session based on the pheromone image corresponding to the reference moment of the target virtual character; the target moment is after the reference moment;

[0014] An external cheat detection module, configured to compare the reference decision intention of the target virtual character at the target moment with the actual decision intention of the target virtual character at the target moment, and determine whether the target virtual character uses an external cheat according to the comparison result.

[0015] A third aspect of the present application provides a computer device, which includes a processor and a memory:

[0016] The memory is used to store a computer program;

[0017] The processor is configured to execute the steps of the data processing method described in the first aspect above according to the computer program.

[0018] A fourth aspect of the present application provides a computer-readable storage medium, which is used to store a computer program, and the computer program is used to execute the steps of the data processing method described in the first aspect above.

[0019] A fifth aspect of the present application provides a computer program product or a computer program, which includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps of the data processing method described in the first aspect above.

[0020] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0021] An embodiment of the present application provides a data processing method, which innovatively proposes a mechanism for detecting game cheats based on pheromone images. Specifically, in this method, first obtain the replay data of the game session, which is used to restore the behaviors of the virtual characters and the game environment in the game session. Then, for the target virtual character in the game session, generate a pheromone image corresponding to the reference moment of the target virtual character in the game session according to the replay data; the pheromone image is a new concept proposed in the embodiment of the present application, which is used to represent the game environment information perceived by the target virtual character at the reference moment; it should be noted that the game environment information represented by the pheromone image is different from the real game environment information at the reference moment. The game environment information represented by the pheromone image is the information obtained from the perspective of the target virtual character by virtue of the target virtual character's senses, corresponding to the perception of the target virtual character from its own perspective, and the information not covered by the perspective of the target virtual character cannot be accurately represented. Furthermore, based on the pheromone image corresponding to the reference moment of the target virtual character, predict the reference decision intention of the target virtual character at the subsequent target moment; that is, starting from the perspective of the target virtual character, predict the decision intention that should be generated in the currently perceived game environment. Finally, compare the reference decision intention of the target virtual character at the target moment with the actual decision intention of the target virtual character at the target moment (determined according to the replay data), and determine whether the target virtual character uses a cheat according to the comparison result; it should be understood that the above comparison operation of the decision intention aims to compare the actual decision intention generated by the target virtual character in the game session with the decision intention that should be generated by the target virtual character in the currently perceived game environment. If the two are inconsistent, it means that the game environment information on which the target virtual character bases the above actual decision intention in the game session is not the game environment information it perceives. It may have obtained richer and more accurate game environment information through a cheat and generated a decision intention that does not conform to the actually perceived game environment information, thus realizing accurate and effective detection of game cheats. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the application scenario of the data processing method provided by the embodiment of the present application;

[0023] Figure 2 It is a schematic flowchart of the data processing method provided by the embodiment of the present application;

[0024] Figure 3 It is a schematic diagram of the movement path of the imaginary enemy model provided by the embodiment of the present application;

[0025] Figure 4 It is a schematic diagram of the distance attenuation characteristic of the pheromone provided by the embodiment of the present application;

[0026] Figure 5Schematic diagram of the time decay characteristic of the pheromone provided by the embodiment of the present application;

[0027] Figure 6 Schematic diagram of a dangerous pheromone image provided by the embodiment of the present application;

[0028] Figure 7 Schematic diagram of a safe pheromone image provided by the embodiment of the present application;

[0029] Figure 8 Schematic diagram of another dangerous pheromone image provided by the embodiment of the present application;

[0030] Figure 9 Schematic diagram of another safe pheromone image provided by the embodiment of the present application;

[0031] Figure 10 Schematic diagram of a scenario of approaching without a field of view provided by the embodiment of the present application;

[0032] Figure 11 Schematic diagram of the measurement conditions for the approaching segment without a field of view provided by the embodiment of the present application;

[0033] Figure 12 Schematic diagram of the game interface facing the target virtual character provided by the embodiment of the present application;

[0034] Figure 13 Schematic diagram of the operation of the decision intention prediction model provided by the embodiment of the present application;

[0035] Figure 14 Schematic diagram of the implementation architecture of the data processing method provided by the embodiment of the present application;

[0036] Figure 15 Data comparison chart of the recall effect of suspicious segments provided by the embodiment of the present application;

[0037] Figure 16 Schematic diagram of the structure of the data processing device provided by the embodiment of the present application;

[0038] Figure 17 Schematic diagram of the structure of the terminal device provided by the embodiment of the present application;

[0039] Figure 18 Schematic diagram of the structure of the server provided by the embodiment of the present application. Detailed implementation manners

[0040] To enable those skilled in the art to better understand the solution of this application, the following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of them. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of this application.

[0041] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned accompanying drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of this application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0042] Artificial Intelligence (AI) is a theory, method, 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 knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. Artificial intelligence also studies the design principles and implementation methods of various intelligent machines, enabling the machines to have the functions of perception, reasoning, and decision-making.

[0043] Artificial intelligence technology is an interdisciplinary subject with a wide range of fields involved, including both hardware-level technologies and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, pre-trained model technology, operation / interaction systems, mechatronics, etc. Among them, the pre-trained model, also known as the large model or the foundation model, can be widely applied to downstream tasks in various directions of artificial intelligence after fine-tuning. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0044] Computer Vision (CV) technology is a science that studies how to enable machines to "see". Further, it refers to machine vision that uses cameras and computers to replace human eyes to identify and measure targets, and further performs graphics processing to make the computer-processed images more suitable for human eyes to observe or be transmitted to instruments for detection. As a scientific discipline, computer vision studies related theories and technologies and attempts to build artificial intelligence systems that can obtain information from images or multi-dimensional data. Large model technology has brought important changes to the development of computer vision technology. Pretrained models in the field of vision such as swin-transformer, ViT, V-MOE, and MAE can be quickly and widely applied to specific downstream tasks after fine-tuning. Computer vision technology usually includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, three-dimensional object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous localization and mapping, etc. It also includes common biometric recognition technologies such as face recognition and fingerprint recognition.

[0045] The solution provided in the embodiments of this application relates to technologies such as computer vision in artificial intelligence, and is specifically described through the following embodiments:

[0046] The data processing method provided in the embodiments of this application can be executed by a computer device, which can be a terminal device or a server. Among them, terminal devices include, but are not limited to, mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle-mounted terminals, aircraft, etc. The server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server.

[0047] The information (such as virtual character information in games, etc.), data (such as replay data of game rounds, etc.), and signals involved in the embodiments of this application are all authorized by relevant objects or fully authorized by all parties, and the collection, use, and processing of relevant data all comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0048] To facilitate understanding of the data processing method provided in the embodiments of this application, the application scenario of this data processing method will be exemplarily introduced below with the execution subject of this data processing method being a server as an example.

[0049] See Figure 1 , Figure 1 is a schematic diagram of the application scenario of the data processing method provided in the embodiments of this application. As Figure 1As shown, the application scenario includes a server 110 and a database 120. The server 110 can access the data 120 through a network, or the database 120 can also be integrated in the server 110. Among them, the server 110 is used to execute the data processing method provided in the embodiments of the present application, and detect whether a virtual character in the game session uses a cheating program based on the replay data of the game session; the database 120 is used to store the replay data of the game session.

[0050] In practical applications, the server 110 can retrieve the replay data of a certain game session from the database 120, and the replay data can restore the actions generated by the virtual character in the game session and the game environment where the virtual character is located.

[0051] Then, for the target virtual character in the game session, the server 110 can generate a pheromone image corresponding to the reference moment of the target virtual character in the game session according to the above replay data. It should be understood that the target virtual character here can be a virtual character controlled by any player in the game session; the reference moment here can be any moment in the game session. The pheromone image here is a new concept proposed in the embodiments of the present application, which is used to represent the game environment information perceived by the target virtual character at the reference moment; it should be noted that the game environment information represented by the pheromone image is different from the real game environment information at the reference moment. The game environment information represented by the pheromone image is the information obtained from the perspective of the target virtual character by relying on the feelings of the target virtual character, corresponding to the perception of the target virtual character from its own perspective, and the information not covered by the perspective of the target virtual character cannot be accurately represented. For example, the information of the enemy character not covered by the perspective of the target virtual character cannot be accurately represented.

[0052] Furthermore, the server 110 can predict the reference decision intention of the target virtual character at the target moment in the game session based on the pheromone image corresponding to the target virtual character at the reference moment. It should be understood that the target moment is a moment after the reference moment in the game session, for example, it can be a moment adjacent to the reference moment after the reference moment. The reference decision intention here is determined from the perspective of the target virtual character and is the decision intention that should be generated in the currently perceived game environment.

[0053] Finally, the server 110 compares the reference decision intention of the target virtual character at the target moment with the actual decision intention of the target virtual character at the target moment, and determines whether the target virtual character uses a cheating program according to the comparison result. It should be understood that the actual decision intention here is the decision intention actually generated by the target virtual character at the target moment in the game session determined according to the replay data. The comparison operation of the decision intention here is aimed at comparing the decision intention actually generated by the target virtual character in the game session with the decision intention that the target virtual character should generate in the currently perceived game environment. If the two are inconsistent, it means that the game environment information relied on by the target virtual character when generating the above actual decision intention in the game session is not the game environment information it perceives. It may have obtained richer and more accurate game environment information through a cheating program and generated a decision intention that does not conform to the actually perceived game environment information, so as to achieve accurate and effective detection of game cheating programs.

[0054] It should be understood that Figure 1 The application scenarios shown are only examples. In actual applications, the data processing method provided by the embodiments of the present application can also be applied to other scenarios, and no limitations are imposed on the application scenarios of the data processing method provided by the embodiments of the present application here.

[0055] The data processing method provided by the present application will be introduced in detail below through method embodiments.

[0056] See Figure 2 , Figure 2 which is a schematic flowchart of the data processing method provided by the embodiments of the present application. For ease of description, the following takes the server as the execution subject of the data processing method as an example for introduction. As Figure 2 shown, the data processing method includes the following steps:

[0057] Step 201: Obtain the replay data of the game session; the replay data is used to restore the behaviors generated by the virtual characters in the game session and the game environment where they are located.

[0058] In an embodiment of the present application, when it is necessary to detect whether there are players using cheats in a certain game session, the server needs to first obtain the replay data of the game session. Exemplarily, the server can obtain the replay data of the game session from a relevant database; for example, the server can, according to the session identifier of the game session (information used to uniquely identify the game session), retrieve the replay data corresponding to the session identifier from the database storing the replay data of the ended game sessions. Of course, in practical applications, the server can also obtain the replay data of the game session through other means. For example, the server can obtain the replay data of the game session from the game backend server that provides support for the progress of the game session, etc. The embodiment of the present application does not make any limitation on the acquisition method of the replay data of the game session here.

[0059] It should be noted that the game session in the embodiment of the present application refers to a session generated by the confrontation of virtual characters respectively controlled by multiple players. For example, the game session can be a game session of a MOBA game. Exemplarily, when the game session is a game session of a MOBA game, there can be two mutually hostile teams in the game session, and the number of virtual characters included in each of the two teams is the same. For example, each team includes 5 virtual characters controlled by players; in the game session, each virtual character, under the control of the player, adopts specific strategies and tactics, cooperates with other virtual characters in its own team, obtains rewards in the game, improves its combat ability, and attacks the virtual characters in the other team, with the goal of destroying the target building that the other team needs to defend as the winning goal.

[0060] It should be noted that the replay data (replay data) of the game session in the embodiments of the present application refers to the data used to restore the game session. Based on this replay data, the overall situation of the game session at each moment can be restored; that is, from the perspective of the overall game, the position, actions, and states (such as combat power, remaining health, level, etc.) of each virtual character controlled by the player in the game map at each moment of the game session can be restored. It can also restore the position, actions, and states of non-player characters (such as wild monsters, minions, etc.) in the game map at each moment of the game session, and can also restore the state of the target building (such as a tower) at each moment of the game session (such as the damage degree of the tower, etc.). In addition, based on this replay data, the game session situation of each virtual character at each moment can also be restored; that is, from the perspective of a single virtual character, the information known from the perspective of this virtual character at each moment of the game session can be restored, that is, the game environment from the perspective of this virtual character can be restored, such as the position where this virtual character is currently located in the game map and the objects it faces (such as enemy characters, wild monsters, etc.), and also the positions of other friendly characters in the game map and the positions of enemy characters in the game map known by this virtual character through the overall game map, etc. It can also restore the actions (such as moving, attacking, etc.) generated by this virtual character at each moment of the game session.

[0061] That is to say, based on the above replay data of the game session, it is possible to obtain the overall game session information at each moment of the game session from the "God" perspective (that is, the relevant information of all objects in the game session), and it is also possible to obtain the game session information that each player can perceive at each moment of the game session from the perspective of a single player (that is, the relevant information perceived by the virtual character controlled by this player in the game session). It can also be understood that based on the above replay data, the game session from the "God" perspective and the game session from the perspective of each player can be fully replayed.

[0062] Step 202: For the target virtual character in the game session, generate a pheromone image corresponding to the reference moment of the target virtual character in the game session according to the replay data; the pheromone image is used to represent the game environment information perceived by the target virtual character at the reference moment.

[0063] After the server obtains the replay data of the game session, it can generate a pheromone image corresponding to the reference moment of the target virtual character in the game session for the target virtual character in the game session by using the obtained replay data, so as to reflect the game environment information perceived by the target virtual character at the reference moment.

[0064] It should be noted that the target virtual character in the embodiments of the present application can be any virtual character controlled by a player in a game session. The reference moment in the embodiments of the present application can be any moment in the game session; for example, each moment in the game session can be regarded as a reference moment; or a specific moment in the game session can be regarded as a reference moment according to actual needs. For example, when it is necessary to detect whether there is a behavior of using external software at a target moment in the game session, several moments before and adjacent to the target moment in the game session can be regarded as reference moments. For example, assuming that the target moment is the 30th second in the game session, then the 20th second to the 29th second in the game session can be regarded as reference moments.

[0065] It should be noted that the pheromone image in the embodiments of the present application is a new concept inspired by the "ant colony mechanism", which can represent the game environment information perceived by the virtual character at a specific moment; the pheromone image corresponding to the target virtual character at the reference moment generated above is used to represent the game environment information perceived by the target virtual character at the reference moment. It should be understood that the game environment information represented by the pheromone image here can include, for example, the information of the enemy perceived by the virtual character (such as the position of the enemy character, the moving speed of the enemy character, the moving path of the enemy character, etc.), the information of the player's own side perceived by the virtual character (such as the position of the player's own character, the moving speed of the player's own character, the moving path of the player's own character, etc.), and the information of the target building perceived by the virtual character (such as the positions of the enemy tower and the player's own tower).

[0066] It should be noted that the game environment information represented by the above pheromone image may be different from the actual game environment information. Specifically, in an actual game round, a virtual character often cannot accurately obtain information about the enemy, and can only rely on its own experience and feeling to estimate the enemy's information. The enemy information represented by the pheromone image is the enemy information estimated by the virtual character itself. For example, when the field of view of the virtual character covers an enemy character, the virtual character can obtain the position of the enemy character. When the field of view of the virtual character cannot cover the enemy character, the virtual character can estimate the moving speed and moving path of the enemy character according to its own experience, and thus determine the possible position of the enemy character. The enemy information represented by the pheromone image is the enemy information estimated by the virtual character by combining the enemy character it actually sees and its own experience in the above process. Since the enemy information represented by the pheromone image is estimated based on experience, in many cases, the enemy information represented by the pheromone image is not consistent with the real enemy information. For the information of one's own side represented by the pheromone image, since the virtual character can obtain the information of other virtual characters belonging to the same team as itself in real time and accurately, the information of one's own side represented by the pheromone image is usually accurate. For the information of the target building represented by the pheromone image, since the virtual character can also obtain the information of the friendly target building and the enemy target building in real time and accurately, the information of the target building represented by the pheromone image is usually also accurate.

[0067] In summary, the pheromone image proposed in the embodiment of the present application is used to represent the game environment information perceived by the virtual character itself, that is, the information obtained from the perspective of the virtual character based on the feeling of the virtual character, and it can accurately restore the game environment information that the virtual character should rely on when making corresponding decision-making intentions.

[0068] Step 203: Predict the reference decision-making intention of the target virtual character at the target moment in the game round based on the pheromone image corresponding to the target virtual character at the reference moment; the target moment is after the reference moment.

[0069] After the server generates the pheromone image corresponding to the target virtual character at the reference moment, it can predict the reference decision-making intention that the target virtual character should generate at the target moment after the reference moment based on this pheromone image. Exemplarily, assuming that there are multiple reference moments, the server can first arrange the pheromone images corresponding to each reference moment in the order of their respective sequences in the game round to obtain a pheromone image sequence. Then, input this pheromone image sequence into a pre-trained decision-making intention prediction model, and analyze and process the input pheromone image sequence through this decision-making intention prediction model, and then output the reference decision-making intention that the target virtual character should generate at the target moment.

[0070] It should be noted that the target moment in the embodiments of the present application can be any moment after the reference moment in the game session, and the time interval between the target moment and the reference moment should be less than a preset interval threshold; exemplarily, the target moment can be a moment within a suspicious segment in the game session. For example, a suspicious segment in the game session where there may be cheating behavior can be mined, and then any moment within the suspicious segment can be used as the target moment.

[0071] It should be noted that the reference decision intention in the embodiments of the present application is the decision intention predicted based on the pheromone image corresponding to the previous moment, which is used to reflect the decision intention that the virtual character should generate under the game environment perceived by a certain virtual character. Essentially, the decision intention is the task that the virtual character needs to execute within a period of time in the game session (such as jungle clearing, attacking a tower, waiting for other virtual characters, etc.).

[0072] In addition, when determining the reference decision intention using a decision intention prediction model, the decision intention prediction model used can be a combination of Transformer, Long Short-Term Memory (LSTM) and Vision Transformer (VIT), a time-delay network, a gated convolutional neural network, etc. The embodiments of the present application do not make any limitations on the decision intention prediction model here.

[0073] Step 204: Compare the reference decision intention of the target virtual character at the target moment with the actual decision intention of the target virtual character at the target moment, and determine whether the target virtual character uses cheating according to the comparison result.

[0074] Furthermore, the server can compare the predicted reference decision intention of the target virtual character at the target moment with the actual decision intention of the target virtual character at the target moment, and determine whether the target virtual character uses cheating according to the obtained comparison result, that is, determine whether the player controlling the target virtual character uses cheating.

[0075] It should be noted that the actual decision intention in the embodiments of the present application is the decision intention truly generated by the target virtual character at the target moment in the game session, and it is determined according to the acquired replay data. Exemplarily, the server can determine each task performed by the target virtual character in the game session (such as killing monsters, attacking towers, waiting for other virtual characters, attacking enemy characters, etc.) based on the replay data; then, divide the game session into each game segment according to each task performed. For example, assuming that the target virtual character completed the task of killing monsters at the 10th second of the game session and completed the task of attacking the enemy character at the 18th second, then the 0th second to the 10th second of the game session can be divided into one game segment, and the 11th second to the 18th second of the game session can be divided into another game segment; for each game segment, the server can use the task completed in this game segment as the actual decision intention at each moment in this game segment. For example, for the game segment from the 0th second to the 10th second mentioned above, the actual decision intention at each moment is killing monsters, and for the game segment from the 11th second to the 18th second mentioned above, the actual decision intention at each moment is attacking the enemy character.

[0076] It should be noted that the purpose of comparing the reference decision intention and the actual decision intention is to compare the decision intention actually generated by the target virtual character in the game session with the decision intention that the target virtual character should generate in the currently perceived game environment. If the two are inconsistent, it means that the game environment information based on which the target virtual character generated the above actual decision intention in the game session is not the game environment information it perceived. It may have obtained richer and more accurate game environment information through cheats and generated a decision intention that does not conform to the actually perceived game environment information.

[0077] Exemplarily, when using the decision intention prediction model to determine the reference decision intention, the reference decision intention output by the decision intention prediction model may be in the form of a feature vector. To facilitate the comparison between the reference decision intention and the actual decision intention, the actual decision intention can also be converted into the form of a feature vector. For example, the actual decision intention can be processed through a pre-trained Generative Pre-Trained Transformer (GPT) model to obtain the actual decision intention in the corresponding feature vector form. Furthermore, the server can calculate the similarity between the feature vector corresponding to the reference decision intention and the feature vector corresponding to the actual decision intention to obtain the comparison result.

[0078] Exemplarily, the target moment can be each moment in the suspicious segment. Thus, the server needs to determine the reference decision intention corresponding to the target virtual character at each target moment, and obtain the reference decision intentions corresponding to each of the multiple target moments. When comparing the reference decision intention and the actual decision intention, the server can, for each target moment, compare the reference decision intention corresponding to the target moment with the actual decision intention in the above manner to obtain the comparison result corresponding to the target moment. Furthermore, the server can count the comparison results corresponding to each of the target moments. If it is statistically determined that the proportion of the comparison results indicating that the reference decision intention is different from the actual decision intention among all the comparison results exceeds a preset proportion, it can be determined that the target virtual character uses a cheating program. On the contrary, if it is statistically determined that the proportion of the comparison results indicating that the reference decision intention is different from the actual decision intention among all the comparison results does not exceed the preset proportion, it can be determined that the target virtual character does not use a cheating program. It should be understood that when the comparison result is the similarity between the feature vector corresponding to the reference decision intention and the feature vector corresponding to the actual decision intention, if the similarity does not exceed a preset similarity threshold, it is considered that the comparison result indicates that the reference decision intention is different from the actual decision intention. If the similarity exceeds the preset similarity threshold, it is considered that the comparison result indicates that the reference decision intention is the same as the actual decision intention.

[0079] Of course, in practical applications, the server can also use other methods to compare the reference decision intention and the actual decision intention, and use other methods to determine whether the target virtual character uses a cheating program based on the comparison result. The embodiments of the present application do not make any limitations in this regard.

[0080] The data processing method provided by the embodiments of this application innovatively proposes a mechanism for detecting game cheats based on pheromone images. Specifically, in this method, first, replay data of a game session is obtained, and this replay data is used to restore the behaviors of virtual characters and the game environment in the game session. Then, for a target virtual character in the game session, a pheromone image corresponding to the reference moment in the game session of the target virtual character is generated according to the replay data; the pheromone image is a new concept proposed by the embodiments of this application and is used to represent the game environment information perceived by the target virtual character at the reference moment; it should be noted that the game environment information represented by the pheromone image is different from the real game environment information at the reference moment. The game environment information represented by the pheromone image is the information obtained from the perspective of the target virtual character relying on the senses of the target virtual character, corresponding to the perception of the target virtual character from its own perspective, and the information not covered by the perspective of the target virtual character cannot be accurately represented. Furthermore, based on the pheromone image corresponding to the target virtual character at the reference moment, the reference decision intention of the target virtual character at a subsequent target moment is predicted; that is, starting from the perspective of the target virtual character, the decision intention that should be generated in the currently perceived game environment is predicted. Finally, the reference decision intention of the target virtual character at the target moment is compared with the actual decision intention of the target virtual character at the target moment (determined according to the replay data), and it is determined whether the target virtual character uses a cheat based on the comparison result; it should be understood that the above comparison operation of decision intentions aims to compare the actual decision intention generated by the target virtual character in the game session with the decision intention that should be generated by the target virtual character in the currently perceived game environment. If the two are inconsistent, it means that the game environment information relied on by the target virtual character when generating the above actual decision intention in the game session is not the game environment information it perceives. It may have obtained richer and more accurate game environment information through a cheat and generated a decision intention that does not conform to the actually perceived game environment information, thus achieving accurate and effective detection of game cheats.

[0081] In a possible implementation manner, the server can generate the pheromone image corresponding to the target virtual character at the reference moment according to the replay data in the following way:

[0082] Generate a dangerous pheromone image and a safe pheromone image corresponding to the target virtual character at the reference moment according to the replay data; wherein, the dangerous pheromone image is used to represent the enemy information perceived by the target virtual character at the reference moment, and the safe pheromone image is used to represent the friendly information perceived by the target virtual character at the reference moment.

[0083] Specifically, the server can generate corresponding dangerous pheromone images and safe pheromone images for the enemy information and friendly information perceived by the target virtual character at the reference moment. It should be understood that the dangerous pheromone image represents the enemy information perceived by the target virtual character at the reference moment, such as the position of a certain enemy character perceived at the reference moment and the positions of the enemy character at several moments before the reference moment; the safe pheromone image represents the friendly information perceived by the target virtual character at the reference moment, such as the position of the friendly character perceived at the reference moment and the positions of the friendly character at several moments before the reference moment.

[0084] In this way, by separately constructing a dangerous pheromone image for representing the perceived enemy information and a safe pheromone image for representing the perceived friendly information, the enemy information and friendly information perceived by the target virtual character at the reference moment can be more clearly reflected, that is, the dangerous pheromone image and the safe pheromone image are respectively used to reflect the dangerous information and safe information currently known by the target virtual character; correspondingly, it helps to more accurately predict the reference decision-making intention that the target virtual character should generate at the target moment based on the dangerous pheromone image and the safe pheromone image.

[0085] It should be understood that in actual applications, when the server predicts the reference decision-making intention, it can input the above-mentioned dangerous pheromone image and safe pheromone image into the decision-making intention prediction model together for predicting the reference decision-making intention, or synthesize the above-mentioned dangerous pheromone image and safe pheromone image into a comprehensive pheromone image that represents both enemy information and friendly information, and input the comprehensive pheromone image into the decision-making intention prediction model for predicting the reference decision-making intention.

[0086] In a possible implementation manner, the above-mentioned dangerous pheromone image can be specifically generated in the following way:

[0087] According to the replay data, determine the enemy characters perceived by the target virtual character and construct a hypothetical enemy model corresponding to the enemy characters; the hypothetical enemy model is used to simulate the enemy characters perceived by the target virtual character. When the field of view of the target virtual character does not cover the enemy characters, the hypothetical enemy model moves towards the target virtual character at a hypothetical speed and along a hypothetical path;

[0088] Determine the dangerous pheromone image according to the position of the hypothetical enemy model at the reference moment and the positions of the hypothetical enemy model at n moments before the reference moment; n is an integer greater than or equal to 1.

[0089] Specifically, the server can first determine the enemy characters that the target virtual character can perceive based on the playback data. For example, at several moments before the reference moment, enemy character A appeared within the field of view of the target virtual character. After that, enemy character A disappeared from the field of view of the target virtual character. Then, the player controlling the target virtual character will think that enemy character A is near the target virtual character and may attack the target virtual character at any time. At this time, enemy character A is the enemy character perceived by the target virtual character.

[0090] Since the field of view of the target virtual character may not continuously cover the enemy characters it perceives, it is necessary to construct a dummy enemy model corresponding to the enemy character to simulate the enemy character perceived by the target virtual character. Specifically, when the field of view of the target virtual character can cover the enemy character, the dummy enemy model will move according to the behavior of the enemy character actually seen by the target virtual character; when the field of view of the target virtual character cannot cover the enemy character, the dummy enemy model will move towards the target virtual character at a dummy speed and along a dummy path. Here, the dummy speed can be, for example, the moving speed of the enemy character when it moves normally, and the dummy path can be, for example, the shortest straight-line path from the position where the target virtual character last saw the enemy character to the position of the target virtual character. For example, assume that enemy character A appeared within the field of view of the target virtual character from 5 seconds to 3 seconds before the reference moment. Then, the dummy enemy model corresponding to enemy character A will be at the position of enemy character A seen by the target virtual character from 5 seconds to 3 seconds before the reference moment. Assume that enemy character A was not within the field of view of the target virtual character at the reference moment and within the previous 2 seconds. Then, within the reference moment and the previous 2 seconds, the dummy enemy model corresponding to enemy character A will move towards the target virtual character at the normal moving speed of enemy character A along the straight-line path closest to the target virtual character.

[0091] Figure 3 It is a schematic diagram of the moving path of the dummy enemy model provided by the embodiment of the present application. As Figure 3 shown, among them, path 1 is the moving path of the target virtual character, path 2 is the moving path of the dummy enemy model corresponding to the enemy character A perceived by the target virtual character, and path 3 is the actual moving path of enemy character A. It can be seen that in the embodiment of the present application, the dummy enemy model is only a model used to represent the enemy character perceived by the target virtual character, and the actions performed by the dummy enemy model may be the same as or different from the actions performed by the real enemy character.

[0092] Furthermore, the server can generate a dangerous pheromone image corresponding to the target virtual character at the reference time based on the above imaginary enemy model, according to the position of the imaginary enemy model at the reference time and the positions of the imaginary enemy model at n moments before the reference time. It should be noted that n here is usually an integer greater than or equal to 1, and the value of n is determined by the decay rate of the pheromone in the dangerous pheromone image. For example, if it takes 5 seconds for a pheromone to completely dissipate from its initial deployment, then n is equal to 5.

[0093] In this way, through the above method, the imaginary enemy model is used to simulate the enemy characters perceived by the target virtual character, and a dangerous pheromone image is generated according to the movement path of the imaginary enemy model. Since the imaginary enemy model reflects the behavior of the enemy characters imagined by the player, based on this imaginary enemy model to construct a dangerous pheromone image can ensure that the dangerous pheromone image can accurately represent the enemy characters perceived by the target virtual character, that is, accurately represent the dangerous information in the target virtual character's own cognition.

[0094] In a possible implementation manner, the server can specifically determine the dangerous pheromone image in the following way according to the position of the imaginary enemy model at the reference time and the positions of the imaginary enemy model at n moments before the reference time:

[0095] In the map image corresponding to the game session, initial dangerous pheromones are deployed at the first reference position; the first reference position is the position of the imaginary enemy model at the reference time; the initial dangerous pheromones correspond to the highest information intensity;

[0096] In the first range centered on the first reference position, distance decay pheromones associated with the initial dangerous pheromones are deployed; the information intensity corresponding to the distance decay pheromones is negatively correlated with the distance between the deployment position of the distance decay pheromones and the first reference position;

[0097] Time decay pheromones are deployed at the historical positions, and in the second range centered on the historical positions, distance decay pheromones associated with the time decay pheromones are deployed; the historical positions are the positions of the imaginary enemy model at any of the n moments before the reference time; the information intensity corresponding to the time decay pheromones is negatively correlated with the time interval between the moment corresponding to the time decay pheromones and the reference time.

[0098] It should be noted that in the embodiments of the present application, the dangerous pheromones included in the dangerous pheromone image have distance decay characteristics, time decay characteristics, and superposition characteristics.

[0099] The so-called distance decay characteristic can also be called distance outward diffusion, which means that distance decay pheromones with information intensity weakening with the increase of distance are deployed around a certain pheromone. Figure 4A schematic diagram of the distance attenuation characteristic provided by the embodiment of the present application. Figure 4 The darker the color of the grid, the stronger the information intensity of the pheromone deployed here. As Figure 4 shown, for the newly deployed pheromone 401, distance attenuation pheromones 402 are deployed at the positions adjacent to its periphery. The information intensity of the distance attenuation pheromones 402 is weaker than that of the pheromone 401. In addition, distance attenuation pheromones 403 are deployed at the positions adjacent to the periphery of the distance attenuation pheromones 402. The information intensity of the distance attenuation pheromones 403 is weaker than that of the distance attenuation pheromones 402, and so on until the extended information intensity is lower than the preset intensity threshold. It should be understood that the above distance attenuation pheromones can be attenuated according to a preset rule. For example, the diffusion range of the information intensity can be limited between ±2 distance units, and the information intensity can be attenuated according to exp(-d), where d represents the distance from the deployment position of the pheromone with the strongest associated information intensity.

[0100] The so-called time attenuation characteristic means that as time goes by, the information intensity of the historically deployed pheromones gradually weakens. The historically deployed pheromones here can be called time attenuation pheromones. Figure 5 A schematic diagram of the time attenuation characteristic provided by the embodiment of the present application. Figure 5 The darker the color of the grid, the stronger the information intensity of the pheromone deployed here. As Figure 5 shown, assuming that the currently constructed pheromone image corresponds to the 5th second of the game round, then the pheromone 501 deployed for the 5th second will correspond to the strongest information intensity. The pheromone 502 deployed for the 4th second will be used as a time attenuation pheromone, and the information intensity of this pheromone 502 is weaker than that of the pheromone 501. The pheromone 503 deployed for the 3rd second will also be used as a time attenuation pheromone, and the information intensity of this pheromone 503 is weaker than that of the pheromone 502. The pheromone 504 deployed for the 2nd second will also be used as a time attenuation pheromone, and the information intensity of this pheromone 504 is weaker than that of the pheromone 503, and so on until the information intensity of the time attenuation pheromone is lower than the preset intensity threshold. It should be understood that the above time attenuation pheromones can be attenuated according to a preset rule. For example, every time a time unit passes, the intensity of the pheromone will be weakened to 80% of the original. This time attenuation characteristic can be understood as the pheromones gradually dissipating in the environment.

[0101] The so-called superposition characteristic means that in the case where the information intensity of the pheromone has not completely decayed and dissipated, if a new pheromone is deployed at the position where the pheromone is located, then the information intensity of the original pheromone at that position will be superimposed with the information intensity of the new pheromone.

[0102] In the embodiments of the present application, when constructing the dangerous pheromone image corresponding to the target virtual character at the reference moment, an initial dangerous pheromone is deployed at the first reference position in the map image corresponding to the game round. The first reference position is the position where the imaginary enemy model is located in the game map at the reference moment, and the initial dangerous pheromone corresponds to the highest information intensity. Figure 6 It is a schematic diagram of a dangerous pheromone image provided by the embodiments of the present application, as Figure 6 shown, an initial dangerous pheromone corresponding to the highest information intensity is deployed at 601 in the map image.

[0103] In addition, distance decay pheromones associated with the initial dangerous pheromone are correspondingly deployed at each position within the first range centered on the first reference position. Specifically, within the first range, for positions closer to the first reference position, distance decay pheromones with higher information intensity will be correspondingly deployed here, and for positions farther from the first reference position, distance decay pheromones with lower information intensity will be correspondingly deployed here. That is, the closer a position is to the first reference position, the higher the information intensity of the distance decay pheromone deployed at that position. Of course, the information intensity of all distance decay pheromones is lower than that of the initial dangerous pheromone. The first range here can be determined according to the diffusion range of the information intensity of the initial dangerous pheromone. As Figure 6 shown, several distance decay pheromones are deployed around the initial dangerous pheromone 601, and the information intensity of the distance decay pheromone is negatively correlated with the distance between its deployment position and the first reference position.

[0104] In addition, time decay pheromones are also correspondingly deployed at the historical positions where the imaginary enemy model was located at n moments before the reference moment; the information intensity of the deployed time decay pheromone is related to the corresponding moment. In essence, the moment corresponding to the time decay pheromone is the moment corresponding to the historical position where the time decay pheromone is located; the longer the time interval between the moment corresponding to the time decay pheromone and the reference moment, the weaker the information intensity of the time decay pheromone, and the shorter the time interval between the moment corresponding to the time decay pheromone and the reference moment, the stronger the information intensity of the time decay pheromone. As Figure 6 shown, time decay pheromone 602 and time decay pheromone 603 are correspondingly deployed at each historical position where the imaginary enemy model was located at n moments before the reference moment.

[0105] It should be understood that each time-decaying pheromone is initially also an initial danger pheromone, which transforms into a time-decaying pheromone as time passes. Correspondingly, around each time-decaying pheromone, a distance-decaying pheromone associated with the time-decaying pheromone is also deployed, that is, the distance-decaying pheromone associated with the time-decaying pheromone is deployed within a second range centered on the time-decaying pheromone. On the one hand, this distance-decaying pheromone decays as the distance between it and the associated time-decaying pheromone increases, and on the other hand, it also decays as time passes. It should be understood that the second range corresponding to each time-decaying pheromone may be different. The shorter the time interval between the moment corresponding to the time-decaying pheromone and the reference moment, the larger the second range corresponding to the time-decaying pheromone. The longer the time interval between the moment corresponding to the time-decaying pheromone and the reference moment, the smaller the second range corresponding to the time-decaying pheromone, and even the second range can be zero. As Figure 6 shown, around the time-decaying pheromone 602, distance-decaying pheromones with different information intensities are deployed, and the deployment range of the distance-decaying pheromone associated with the time-decaying pheromone 602 is relatively large. Since the time interval between the moment corresponding to the time-decaying pheromone 603 and the reference moment is relatively large, the deployment range of the distance-decaying pheromone associated with the time-decaying pheromone 603 has become zero.

[0106] In this way, by generating the danger pheromone image in the above manner, it can be ensured that the generated danger pheromone image accurately reflects the enemy information perceived by the target virtual character, that is, the danger pheromones in the danger pheromone image can accurately and reliably simulate the change situation of the enemy information perceived in the game round, and simulate the process of the enemy information gradually weakening and dissipating from two dimensions of distance and time.

[0107] In a possible implementation manner, the above safety pheromone image can be specifically generated through the following method:

[0108] According to the replay data, determine the position of the friendly character of the target virtual character at the reference moment, and the positions of the friendly character at n moments before the reference moment; n is an integer greater than or equal to 1;

[0109] Determine the safety pheromone image according to the position of the friendly character at the reference moment and the positions of the friendly character at n moments before the reference moment.

[0110] Specifically, for each friendly character belonging to the same team as the target virtual character, the server can determine the position of the friendly character at the reference moment and the positions of the friendly character at n moments before the reference moment according to the replay data of the game session. It should be understood that since the target virtual character can accurately know the position of each friendly character in real time during the game session, the position of the friendly character at the reference moment and the positions of the friendly character at n moments before the reference moment obtained here are both accurate. It should be noted that n here is usually an integer greater than or equal to 1, and the value of n is determined by the decay rate of the pheromones in the safety pheromone image. For example, assuming that it takes 5 seconds for a pheromone to completely dissipate from the initial deployment, then n is equal to 5.

[0111] Furthermore, the server can generate a safety pheromone image corresponding to the target virtual character at the reference moment according to the positions of each friendly character at the reference moment and the positions of each friendly character at n moments before the reference moment.

[0112] In this way, through the above method, using the information of the friendly characters truly perceived by the target virtual character to generate a safety pheromone image can ensure that the safety pheromone image can accurately represent the friendly characters perceived by the target virtual character, that is, accurately represent the safety information in the target virtual character's own cognition.

[0113] In a possible implementation manner, the server can specifically determine the safety pheromone image in the following way according to the position of the friendly character at the reference moment and the positions of the friendly character at n moments before the reference moment:

[0114] Deploy the initial safety pheromone at the second reference position in the map image corresponding to the game session; the second reference position is the position of the friendly character at the reference moment; the initial safety pheromone corresponds to the highest information intensity;

[0115] Deploy the distance decay pheromone associated with the initial safety pheromone within the third range centered on the second reference position; the information intensity corresponding to the distance decay pheromone is negatively correlated with the distance between the deployment position of the distance decay pheromone and the second reference position;

[0116] Deploy the time decay pheromone at the historical position and deploy the distance decay pheromone associated with the time decay pheromone within the fourth range centered on the historical position; the historical position is the position of the friendly character at any one of the n moments before the reference moment; the information intensity corresponding to the time decay pheromone is negatively correlated with the time interval between the moment corresponding to the time decay pheromone and the reference moment.

[0117] It should be noted that in the embodiments of the present application, the safety pheromone images include safety pheromones with distance attenuation characteristics, time attenuation characteristics, and superposition characteristics. These three characteristics are the same as those of the dangerous pheromones in the dangerous pheromone images. For details, please refer to the relevant introduction content above and will not be elaborated here.

[0118] In the embodiments of the present application, when constructing the safety pheromone image corresponding to the target virtual character at the reference moment, an initial safety pheromone will be deployed at the second reference position in the map image corresponding to the game session. The second reference position is the position where each friendly character is located in the game map at the reference moment, and the initial safety pheromone corresponds to the highest information intensity. Figure 7 A schematic diagram of a safety pheromone image provided by an embodiment of the present application, as Figure 7 shown. Assume that there are two friendly characters belonging to the same team as the target virtual character. These two friendly characters are at 710 and 720 in the map image at the reference moment. Then, correspondingly, initial safety pheromones with the highest information intensity are deployed at 710 and 720 in the map image.

[0119] In addition, distance attenuation pheromones associated with the initial safety pheromone will be correspondingly deployed at each position within a third range centered on the second reference position. Specifically, within the third range, for positions closer to the second reference position, distance attenuation pheromones with higher information intensity will be correspondingly deployed here. For positions farther from the second reference position, distance attenuation pheromones with lower information intensity will be correspondingly deployed here. That is, the closer a position is to the second reference position, the higher the information intensity of the distance attenuation pheromone deployed at that position. Of course, the information intensity of all distance attenuation pheromones is lower than that of the initial safety pheromone. The third range here can be determined according to the diffusion range of the information intensity of the initial safety pheromone. As Figure 7 shown, several distance attenuation pheromones are deployed around the initial safety pheromones 710 and 720, and the information intensity of the distance attenuation pheromone is negatively correlated with the distance between its deployment position and the corresponding second reference position.

[0120] In addition, time attenuation pheromones will also be correspondingly deployed at the historical positions where the friendly character was located at n moments before the reference moment; the information intensity of the deployed time attenuation pheromone is related to the corresponding moment. The moment corresponding to the time attenuation pheromone is essentially the moment corresponding to the historical position where the time attenuation pheromone is located; the longer the time interval between the moment corresponding to the time attenuation pheromone and the reference moment, the weaker the information intensity of the time attenuation pheromone. The shorter the time interval between the moment corresponding to the time attenuation pheromone and the reference moment, the stronger the information intensity of the time attenuation pheromone. As Figure 7As shown, for each friendly character, time-decaying pheromones are deployed accordingly at each historical position of the friendly character at n moments before the reference moment. For example, for the initial safety pheromone 710, the associated time-decaying pheromone 711 and time-decaying pheromone 712 are deployed, and the initial safety pheromone 710, time-decaying pheromone 711, and time-decaying pheromone 712 correspond to the same friendly character; for the initial safety pheromone 720, the associated time-decaying pheromone 721 and time-decaying pheromone 722 are also deployed accordingly.

[0121] It should be understood that each time-decaying pheromone is also initially an initial safety pheromone, which is transformed into a time-decaying pheromone as time passes. Accordingly, around each time-decaying pheromone, a distance-decaying pheromone associated with the time-decaying pheromone will also be deployed, that is, the distance-decaying pheromone associated with the time-decaying pheromone will be deployed within a fourth range centered on the time-decaying pheromone. On the one hand, the distance-decaying pheromone will decay as the distance between it and the associated time-decaying pheromone increases, and on the other hand, it will also decay as time passes. It should be understood that the fourth range corresponding to each time-decaying pheromone may be different. The shorter the time interval between the moment corresponding to the time-decaying pheromone and the reference moment, the larger the fourth range corresponding to the time-decaying pheromone. The longer the time interval between the moment corresponding to the time-decaying pheromone and the reference moment, the smaller the second range corresponding to the time-decaying pheromone. Figure 7 As shown, around the time decay pheromone 711, distance decay pheromones corresponding to different information strengths are deployed, and the deployment range of the distance decay pheromone associated with the time decay pheromone 711 is larger, and the deployment range of the distance decay pheromone associated with the time decay pheromone 712 is zero.

[0122] In this way, by generating a safety pheromone image in the above manner, it can be ensured that the generated safety pheromone image accurately reflects the self-information perceived by the target virtual character, that is, the safety pheromones in the safety pheromone image can accurately and reliably simulate the changes in the self-information perceived during the game, and simulate the process of the safety information gradually weakening and dissipating from the two dimensions of distance and time.

[0123] In a possible implementation, the server may also add information of the target building to the pheromone image, namely:

[0124] Add information about target buildings belonging to the enemy to the danger pheromone image;

[0125] Adds information about the target building belonging to the friendly side to the safety pheromone image.

[0126] For example, Figure 8Another schematic diagram of a dangerous pheromone image provided by an embodiment of the present application, as Figure 8 shown. This dangerous pheromone image is based on the dangerous pheromone image shown in Figure 6 and further adds pheromones 801, 802, and 803 corresponding to the towers that the enemy needs to defend. Moreover, distance attenuation pheromones are also deployed around this pheromone. Figure 9 Another schematic diagram of a safe pheromone image provided by an embodiment of the present application, as Figure 9 shown. This safe pheromone image is based on the safe pheromone image shown in Figure 7 and further adds pheromones 901, 902, and 903 corresponding to the towers that one's own side needs to defend. Moreover, distance attenuation pheromones are also deployed around this pheromone.

[0127] In this way, it is ensured that the generated dangerous pheromone image can include richer dangerous information that the target virtual character can perceive, and it is ensured that the generated safe pheromone image can include richer safe information that the target virtual character can perceive, that is, to ensure the information diversity and information richness of the generated pheromone image. Thus, it helps to accurately predict the decision-making intention based on this pheromone image subsequently.

[0128] In a possible implementation manner, the target moment mentioned in step 203 above can be any moment within a suspicious segment in the game round. Correspondingly, this suspicious segment can be mined through the following method:

[0129] Obtain the measurement condition for the out-of-sight approaching segment; the out-of-sight approaching segment is a game segment in which a virtual character approaches an enemy character when its own field of vision does not cover the enemy character;

[0130] According to the replay data, detect the game segments in the game round where the target virtual character meets the above measurement condition as the suspicious segments corresponding to the target virtual character.

[0131] Specifically, in the embodiments of the present application, the out-of-sight approach segment is regarded as a suspicious segment, which refers to a game segment where there may be cheating behavior. The reason for regarding the out-of-sight approach segment as a suspicious segment is that in the out-of-sight approach segment, the virtual character approaches an enemy character even when its own field of vision cannot continuously cover the enemy character. The occurrence of this phenomenon is very likely because the player controlling the virtual character uses a cheat to see a larger field of vision. In the out-of-sight approach segment, the virtual character may perform suspicious behaviors such as active attack behavior (attacking actively when the enemy character cannot be seen), active retreat behavior (retreating actively when the enemy character cannot be seen), stealing or counter-jungling behavior (stealing or counter-jungling when the enemy character cannot be seen), etc. If the above suspicious behaviors occur in the out-of-sight approach segment, it indicates a higher possibility that the virtual character uses a cheat. Figure 10 FIG. is a schematic diagram of an out-of-sight approach scenario provided by an embodiment of the present application, as Figure 10 shown, the virtual character B approaches the enemy character C when its own field of vision does not cover the enemy character C.

[0132] Based on this, the server can obtain the pre-set measurement conditions for the out-of-sight approach segment. Then, according to the replay data of the game round, it can detect whether there is an out-of-sight approach segment that meets the measurement conditions for the target virtual character in the game round. If it detects that the target virtual character has such an out-of-sight approach segment, it can regard the out-of-sight approach segment of the target virtual character as the corresponding suspicious segment of the target virtual character, and further detect whether the target virtual character uses a cheat based on this suspicious segment through the method provided by the embodiments of the present application.

[0133] In this way, by using the above method to mine the out-of-sight approach segments in the game round as suspicious segments, the recall rate of the suspicious segments can be effectively improved, a large number of suspicious segments can be recalled, and cheating detection can be performed based on the recalled suspicious segments, which helps to perform more comprehensive and reliable cheating detection and avoid missed detections. In addition, only performing cheating detection based on the suspicious segments can also reduce the waste of relevant processing resources.

[0134] In a possible implementation manner, the measurement conditions for the out-of-sight approach segment may include start-phase conditions, middle-phase conditions, and end-phase conditions. Among them, the start-phase conditions are used to specify the distance threshold that needs to be met between the virtual character and the enemy character, and the virtual character has no field of vision for the enemy character; the middle-phase conditions are used to specify that the distance between the virtual character and the enemy character shows a decreasing trend; the end-phase conditions are used to specify that the distance between the virtual character and the enemy character shows an increasing trend.

[0135] Figure 11This is a schematic diagram of the measurement conditions for the non-vision approaching segment provided by the embodiments of this application. As Figure 11 shown, for the start stage of the non-vision approaching segment, it is required that the distance between the virtual character and the enemy character is less than or equal to the first distance threshold (such as 200 distance units), and the virtual character has no vision of the enemy character; in addition, the following optional requirements can also be included: neither the virtual character nor the enemy character is dead, at least one of the virtual character and the enemy character is dynamically changing, and the vision range of the virtual character covers the enemy character a preset number of times (such as at least twice). For the middle stage of the non-vision approaching segment, it is required that the distance between the virtual character and the enemy character shows a decreasing trend; in addition, the following optional requirements can also be included: the enemy character is far from the target building (such as the enemy's tower), and a preset number (such as at least one) of friendly characters have approached the virtual character, that is, the distance between the preset number of friendly characters and the virtual character is greater than the second distance threshold (such as 150 distance units). For the end stage of the non-vision approaching segment, it is required that the distance between the virtual character and the enemy character shows an increasing trend. For example, after the virtual character and the enemy character reach the closest distance, the neutral zone between the two continues to expand, and the expansion time reaches the preset time threshold (such as 3s).

[0136] Furthermore, the server can detect, according to the replay data, the game segments in which the target virtual character correspondingly meets the above start stage conditions, the above middle stage conditions, and the above end stage conditions in the game session, as the suspicious segments corresponding to the target virtual character. That is, the server can detect in the game session whether there is the following game segment: the front sub-segment in the game segment meets the above start stage conditions, the middle sub-segment in the game segment meets the above middle stage conditions, and the rear sub-segment in the game segment meets the above end stage conditions; if there is the above game segment, then the game segment is used as the suspicious segment corresponding to the target virtual character.

[0137] In this way, by the above method, the start stage conditions, the middle stage conditions, and the end stage conditions are divided, and the suspicious segments in the game session are detected and recalled based on the conditions of these three stages, which can ensure that the recalled suspicious segments are accurate non-vision approaching segments and can ensure a high recall rate.

[0138] In addition, in the embodiments of this application, the suspicious segments can also be detected according to the generated pheromone image; for example, when the dangerous pheromone image indicates that there is no enemy character near the target virtual character, and the target virtual character initiates an active attack behavior, an active retreat behavior, a wild monster robbing behavior, or a counter-wild monster behavior, etc., then the game segment corresponding to the dangerous pheromone image can be considered as a suspicious segment. Specifically, other measurement conditions for suspicious segments can also be set for the pheromone image, and the embodiments of this application do not make any limitations on this.

[0139] In a possible implementation, the method provided by the embodiments of the present application further includes:

[0140] Determine the macro game characteristics and micro game characteristics corresponding to the target virtual character at the reference moment according to the playback data; the macro game characteristics are used to represent the global game information obtained by the target virtual character through the game global map at the reference moment; the micro game characteristics are used to represent the local game information obtained by the target virtual character from its own perspective at the reference moment.

[0141] Specifically, in the embodiments of the present application, the server can also determine the macro game characteristics and micro game characteristics corresponding to the target virtual character at the reference moment.

[0142] Among them, the macro game characteristics are the characteristics determined from the global perspective of the game session, and are used to reflect the global game information obtained by the target virtual character through the game global map at the reference moment. Figure 12 It is a schematic diagram of the game interface provided by the embodiments of the present application for the target virtual character, as Figure 12 shown, the game global map corresponding to the game session is displayed in the upper left corner of the game interface, and the information obtained by the target virtual character through the game global map can be represented by the macro game characteristics. For example, the positions of each friendly character, the positions of visible enemy characters, the positions of wild monsters, the positions of minions, etc. known by the target virtual character at the reference moment can be represented by the macro game characteristics.

[0143] Among them, the micro game characteristics are the characteristics determined from the local perspective of the game session, and are used to reflect the local game information obtained by the target virtual character from its own perspective at the reference moment. As Figure 12 shown, the game environment currently faced by the target virtual character D has been framed in the game interface, and the relevant information in this game environment can be represented by the micro game characteristics. For example, the position of the opponent character currently faced by the target virtual character relative to the target virtual character, the state of the opponent character faced, etc. can be represented by the micro game characteristics.

[0144] Accordingly, when the server predicts the reference decision intention of the target virtual character at the target moment based on the pheromone image corresponding to the target virtual character at the reference moment, the above-mentioned macroscopic game features and microscopic game features corresponding to the target virtual character at the reference moment can be introduced; that is, based on the pheromone image, macroscopic game features, and microscopic game features corresponding to the target virtual character at the reference moment, the reference decision intention of the target virtual character at the target moment is predicted. For example, the pheromone image, macroscopic game features, and microscopic game features corresponding to the target virtual character at the reference moment can be input into a pre-trained decision intention prediction model, and the decision intention prediction model analyzes and processes the input pheromone image, macroscopic game features, and microscopic game features, and outputs the corresponding reference decision intention of the target virtual character at the target moment.

[0145] In this way, through the above method, battlefield information from different perspectives (i.e., macroscopic game features and microscopic game features) is introduced, combined with the pheromone image, to jointly predict the reference decision intention of the target virtual character at the target moment, so that the prediction task of the reference decision intention can be based on richer and valuable information, which helps to improve the accuracy of the predicted reference decision intention.

[0146] In a possible implementation, the reference moment can be m moments before the target moment, where m is an integer greater than 1. For example, the reference moment can be the previous 5 moments adjacent to the target moment. If the target moment is the 10th second in the game session, then the reference moment can be from the 5th second to the 9th second in the game session.

[0147] Accordingly, when predicting the reference decision intention of the target virtual character at the target moment based on the pheromone image, macroscopic game features, and microscopic game features corresponding to the target virtual character at the reference moment, the server can arrange the pheromone images corresponding to these m moments in the order of their occurrence in the game session to obtain a pheromone image sequence, arrange the macroscopic game features corresponding to these m moments to obtain a macroscopic feature sequence, and arrange the microscopic game features corresponding to these m moments to obtain a microscopic feature sequence; furthermore, through the decision intention prediction model, based on the above pheromone image sequence, macroscopic feature sequence, and microscopic feature sequence, the reference decision intention of the target virtual character at the target moment is determined.

[0148] Specifically, the server can first arrange the pheromone images of danger, pheromone images of safety, macroscopic features of the game, and microscopic features of the game corresponding to each of the m reference times, respectively, so as to obtain a sequence of pheromone images of danger, a sequence of pheromone images of safety, a sequence of macroscopic features, and a sequence of microscopic features. Furthermore, the above-mentioned sequence of pheromone images of danger, sequence of pheromone images of safety, sequence of macroscopic features, and sequence of microscopic features are then correspondingly input into each input data channel of the decision intention prediction model, so as to implement inputting the above-mentioned sequence of pheromone images of danger, sequence of pheromone images of safety, sequence of macroscopic features, and sequence of microscopic features into the decision intention model.

[0149] Figure 13 It is a schematic diagram of the operation of the decision intention prediction model provided by the embodiment of the present application. As Figure 13 shown, after inputting the sequence of pheromone images of danger, sequence of pheromone images of safety, sequence of macroscopic features, and sequence of microscopic features into the decision intention prediction model, the decision intention prediction model can first perform encoding processing on the sequence of pheromone images of danger, sequence of pheromone images of safety, sequence of macroscopic features, and sequence of microscopic features respectively, to obtain the encoded features corresponding to the sequence of pheromone images of danger, sequence of pheromone images of safety, sequence of macroscopic features, and sequence of microscopic features respectively, and then splice the encoded features corresponding to the sequence of pheromone images of danger, sequence of pheromone images of safety, sequence of macroscopic features, and sequence of microscopic features, and predict the reference decision intention based on the spliced features.

[0150] Exemplarily, the above-mentioned decision intention prediction model can specifically be a Transformer structure, which processes features based on the multi-head attention mechanism (Multiheadattention) to predict the reference decision intention. The specific formula of the multi-head attention mechanism is as follows:

[0151]

[0152] where Q, K, and V are all spliced features obtained by splicing the encoded features corresponding to the sequence of pheromone images of danger, sequence of pheromone images of safety, sequence of macroscopic features, and sequence of microscopic features respectively, and d k is a preset attention parameter. In the embodiment of the present application, the decision intention prediction model can perform iterative processing on the spliced features through multiple layers of multi-head attention mechanisms, so as to obtain the predicted reference decision intention in the form of a feature vector.

[0153] In order to facilitate the comparison between the reference decision intention and the actual decision intention, a decision intention conversion model with a GPT structure can also be set up to convert the actual decision intention through this decision intention conversion model to obtain the actual decision intention in the form of a feature vector.

[0154] In this way, through the above method, using a neural network model, based on the pheromone images of danger, pheromone images of safety, macroscopic features of the game, and microscopic features of the game corresponding to multiple reference times before the target time, the reference decision-making intention is predicted, so as to realize the prediction of the reference decision-making intention by combining time series information, and ensure the accuracy and reliability of the predicted reference decision-making intention.

[0155] In a possible implementation manner, the above decision-making intention prediction model can be specifically trained in the following way:

[0156] Obtain training samples; the training samples include a training pheromone image sequence, a training macroscopic feature sequence, a training microscopic feature sequence, and corresponding cheating annotation results;

[0157] Through the decision-making intention prediction model as a generator, based on the training pheromone image sequence, training macroscopic feature sequence, and training microscopic feature sequence in the training samples, determine the predicted decision-making intention;

[0158] Through the cheating recognition model as a discriminator, determine the cheating prediction result according to the predicted decision-making intention;

[0159] Based on the cheating prediction result and the cheating annotation result, construct a loss function;

[0160] Based on the loss function, train at least one of the decision-making intention prediction model and the cheating recognition model.

[0161] Specifically, the above decision-making intention prediction model can be trained based on the training mechanism of a generative adversarial network. Before training the decision-making intention prediction model, a large number of training samples need to be obtained. Each training sample includes a training pheromone image sequence, a training macroscopic feature sequence, a training microscopic feature sequence, and corresponding cheating annotation results. Exemplarily, based on game segments where it has been accurately determined whether there is cheating behavior, the above training pheromone image sequence, training macroscopic feature sequence, and training microscopic feature sequence can be extracted. For example, according to the generation method of pheromone images introduced above, generate the pheromone images (including pheromone images of danger and pheromone images of safety) corresponding to each moment in the game segment, and arrange the pheromone images to obtain a training pheromone image sequence. At the same time, determine the macroscopic features of the game and microscopic features of the game at each moment in the game segment, and arrange the macroscopic features of each game to obtain a training macroscopic feature sequence, and arrange the microscopic features of each game to obtain a training microscopic feature sequence.

[0162] Then, through the decision intention prediction model as a generator, according to the training pheromone image sequence, training macro feature sequence, and training micro feature sequence in the training samples, the corresponding predicted decision intention is determined; the specific working mode of this decision intention prediction model has been introduced in detail above and will not be elaborated here. Next, through the external cheat recognition model as a discriminator, according to this predicted decision intention, the external cheat prediction result is determined. This external cheat recognition model is a model used to identify whether it is an external cheat behavior based on the decision intention, and it can be a classification network with any structure. The embodiment of the present application does not make any limitation on the structure of this external cheat recognition model.

[0163] Furthermore, based on the obtained predicted decision intention and the external cheat annotation result in the training samples, a loss function is constructed; aiming to minimize this loss function, the model parameters of the above decision intention prediction model and external cheat recognition model are alternately adjusted until the training end condition is met. The training end condition here can be, for example, that the number of training rounds for the decision intention prediction model and the external cheat recognition model reaches a preset round threshold, or for another example, the recognition accuracy of this external cheat recognition model reaches a preset accuracy threshold, etc.

[0164] Exemplarily, as shown in Table 1 below, are the definitions and values of some important hyperparameters in model training:

[0165] Table 1

[0166]

[0167] In addition, in model training, the optimizer can be selected as RMSprop, and the learning rate can be set to 1e - 4.

[0168] In this way, through the above method, training the decision intention prediction model based on the architecture of the generative adversarial network can ensure that the trained decision intention prediction model has better performance and can accurately predict the corresponding decision intention.

[0169] To facilitate further understanding of the data processing method provided by the embodiment of the present application, the following combines Figure 14 to give an overall exemplary introduction to this data processing method, Figure 14 which is a schematic diagram of the implementation architecture of the data processing method provided by the embodiment of the present application.

[0170] As Figure 14 shown, in the data processing method provided by the embodiment of the present application, it is necessary to first obtain the replay data of the game session (i.e., replay data). Based on this replay data, the game session can be replayed from each perspective. For example, the game session can be replayed from the global perspective and the perspective of each player participating in the game session.

[0171] Then, through the feature extraction module, feature extraction is performed based on the above replay data of the game session. Specifically, it is necessary to extract the macro features, micro features, and actual decision-making intentions of each virtual character in the game session at each moment. When extracting the macro features and micro features of the game session corresponding to a certain moment for each virtual character, the game interface faced by the player controlling the virtual character at that moment can be restored based on the replay data, and the macro features of the game session can be determined according to the information reflected in the global game map in the game interface (i.e., the mini-map in the upper left corner of the game interface), and the micro features of the game session can be determined according to the game environment faced by the virtual character in the game interface. When extracting the actual decision-making intention of each virtual character, the game session from the perspective of the virtual character can be divided into game segments corresponding to each task (such as jungle clearing, attacking a tower, waiting for other characters, etc.) performed by the virtual character in the game session. Then, for each moment in a game segment, the actual decision-making intention of the virtual character at that moment is the task corresponding to the game segment.

[0172] In addition, based on the obtained replay data of the game session, suspicious segments in the game session can also be recalled. Specifically, for each virtual character in the game session, its corresponding suspicious segment can be recalled. In the embodiments of the present application, the segment of approaching without vision of the virtual character is used as the suspicious segment. Exemplarily, the measurement conditions for the segment of approaching without vision can be preset. The measurement conditions can include start-phase conditions, middle-phase conditions, and end-phase conditions. Among them, the start-phase conditions are used to stipulate that the distance between the virtual character and the enemy character should be less than a preset distance threshold, and the virtual character has no vision of the enemy character. The middle-phase conditions are used to stipulate that the distance between the virtual character and the enemy character shows a decreasing trend. The end-phase conditions are used to stipulate that the distance between the virtual character and the enemy character shows an increasing trend. Furthermore, when mining the segment of approaching without vision corresponding to each virtual character, the game session from the perspective of the virtual character can be restored based on the replay data, and the game segments that meet the above start-phase conditions, middle-phase conditions, and end-phase conditions can be mined therein as the suspicious segments corresponding to the virtual character.

[0173] Furthermore, for each virtual character, the decision-making intention at each moment in its corresponding suspicious segment can be predicted, and it can be determined whether the predicted decision-making intention is consistent with the actual decision-making intention, so as to determine whether the virtual character uses a cheating program.

[0174] Before predicting the decision-making intention, it is necessary to construct a pheromone image based on which the prediction is made. In practical applications, there are many team behaviors in the game. Specifically, team behaviors can be classified into explicit team behaviors (such as team gank, team invading the jungle area, team squatting, etc.), support behaviors (such as team support), and implicit team behaviors (such as team actively retreating, key initiating a team fight, releasing skills to protect, etc.). The above-mentioned team behaviors usually need to be realized based on the interaction information between individuals and the team. To represent this information, the concept of pheromone is proposed in the embodiments of this application. Pheromone is inspired by the "ant colony mechanism". In this application, the behavior trajectory of a virtual character is analogized to the pheromone left by an ant on the path it has walked. Specifically, when a virtual character walks through a certain path, corresponding pheromone will be left on this path. For the path walked by an enemy character, the pheromone left is the dangerous pheromone, which is used to represent relevant dangerous information. For the path walked by one's own character, the pheromone left is the safe pheromone, which is used to represent relevant safe information.

[0175] In an actual game, the above-mentioned pheromone can specifically reflect the information exchange between teammates. That is, based on the path walked by one's own character that the player has learned, the player can understand the actions and intentions of one's own character, so as to cooperate more coordinately with it. The above-mentioned pheromone can also help in the behavior determination of virtual characters. For example, this pheromone can more clearly reflect events such as the virtual character's active attack and active retreat. The above-mentioned pheromone is also helpful for realizing the full-map path evaluation, that is, comprehensively analyzing the information of the entire game map and evaluating the path behavior of the player.

[0176] In the embodiments of this application, the above-mentioned pheromone has the characteristics of distance attenuation, time attenuation, and superposition. The so-called distance attenuation means that the pheromone will spread outward within a certain range. The time attenuation characteristic means that the information intensity of the pheromone will weaken over time. The superposition characteristic means that if a certain virtual character walks through the same path, the pheromone on that path will be superimposed.

[0177] In this way, based on the characteristics of the above-mentioned pheromone and the information represented by the replay data about the perception of the virtual character, a dangerous pheromone image for representing the enemy information perceived by the virtual character and a safe pheromone image for representing the friendly information perceived by the virtual character are generated, and the pheromone of the towers of each side is correspondingly added to the dangerous pheromone image and the safe pheromone image. The specific way of generating the pheromone image has been introduced in detail above and will not be elaborated here.

[0178] Furthermore, for each virtual character, based on the generated pheromone image and the extracted macroscopic and microscopic features of the game, the reference decision-making intention at each moment in the corresponding suspicious segment can be predicted. Specifically, when predicting the reference decision-making intention at a certain moment, the dangerous pheromone images, safe pheromone images, macroscopic features, and microscopic features corresponding to several moments before that moment can be arranged respectively to obtain a dangerous pheromone image sequence, a safe pheromone image sequence, a macroscopic feature sequence, and a microscopic feature sequence. Then, the above-mentioned dangerous pheromone image sequence, safe pheromone image sequence, macroscopic feature sequence, and microscopic feature sequence are input into the decision-making intention prediction model of the Transformer structure, so as to predict the reference decision-making intention at that moment in the form of a feature vector. To facilitate the comparison between the reference decision-making intention and the actual decision-making intention, the actual decision-making intention at that moment can also be converted into the corresponding feature vector form through the feature conversion model of the GPT2 structure. Furthermore, calculate the similarity between the feature vector of the reference decision-making intention and the feature vector of the actual decision-making intention, and compare whether the two decision-making intentions are consistent according to this similarity to obtain the comparison result corresponding to that moment.

[0179] If, in the suspicious segment corresponding to the virtual character, the proportion of the comparison results indicating inconsistent compared decision-making intentions among the comparison results corresponding to each moment in the suspicious segment exceeds the preset proportion threshold, it indicates that the player controlling the virtual character may have used a cheating program.

[0180] In the method provided in the embodiment of the present application above, the recall effect of the suspicious segment has been significantly improved. Figure 15 This is a data comparison chart of the recall effect of the suspicious segment provided by the embodiment of the present application. As Figure 15 shown, the recall rate of the skill segment has increased from 59% to 79%, the recall rate of the active withdrawal segment has increased from 40% to 85%, and the recall rate of the overall segment has increased from 59% to 81%.

[0181] Moreover, the cheating program detection effect of the method provided by the embodiment of the present application has also been significantly improved. The following Table 2 shows the model performance comparison results provided by the embodiment of the present application. Among them, the first six rows are the existing models commonly used for cheating program detection, and the seventh row is the decision-making intention model of the Transformer structure provided by the embodiment of the present application. As shown in Table 2, the F1 value of the model in the embodiment of the present application can be increased to 77.43%, which is 23% higher than the Baseline model. It can be seen that the model provided by the present application has a better effect in cheating program detection.

[0182] Table 2

[0183]

[0184] As shown in Table 3 below, in the game evaluation, the F1 value of the model provided by the embodiment of the present application is 89.9%.

[0185] Table 3

[0186]

[0187] For the data processing method described above, the present application further provides a corresponding data processing device to enable the above data processing method to be applied and implemented in practice.

[0188] See Figure 16 , Figure 16 is a schematic structural diagram of a data processing device 1600 corresponding to the data processing method shown above. As Figure 2 shown, the data processing device 1600 includes: Figure 16 shown, the data processing device 1600 includes:

[0189] A data acquisition module 1601, configured to acquire replay data of a game session; the replay data is used to restore the behaviors and the game environment of virtual characters in the game session;

[0190] A pheromone image generation module 1602, configured to generate, for a target virtual character in the game session, a pheromone image corresponding to a reference moment in the game session according to the replay data; the pheromone image is used to represent game environment information perceived by the target virtual character at the reference moment;

[0191] A decision intention prediction module 1603, configured to predict a reference decision intention of the target virtual character at a target moment in the game session based on the pheromone image corresponding to the reference moment of the target virtual character; the target moment is after the reference moment;

[0192] An external plug-in detection module 1604, configured to compare the reference decision intention of the target virtual character at the target moment with the actual decision intention of the target virtual character at the target moment, and determine whether the target virtual character uses an external plug-in according to the comparison result.

[0193] Optionally, the pheromone image generation module 1602 is specifically configured to:

[0194] Generate a dangerous pheromone image and a safe pheromone image corresponding to the reference moment of the target virtual character according to the replay data; the dangerous pheromone image is used to represent enemy information perceived by the target virtual character at the reference moment, and the safe pheromone image is used to represent friendly information perceived by the target virtual character at the reference moment.

[0195] Optionally, the pheromone image generation module 1602 is specifically configured to generate the dangerous pheromone image in the following manner:

[0196] Based on the playback data, determine the enemy characters perceived by the target virtual character, and construct an imaginary enemy model corresponding to the enemy characters; the imaginary enemy model is used to simulate the enemy characters perceived by the target virtual character. When the field of view of the target virtual character does not cover the enemy characters, the imaginary enemy model moves towards the target virtual character at an imaginary speed and along an imaginary path;

[0197] Based on the position of the imaginary enemy model at the reference moment and the positions of the imaginary enemy model at n moments before the reference moment, determine the dangerous pheromone image; n is an integer greater than or equal to 1.

[0198] Optionally, the pheromone image generation module 1602 is specifically configured to:

[0199] Deploy an initial dangerous pheromone at a first reference position in the map image corresponding to the game round; the first reference position is the position of the imaginary enemy model at the reference moment; the initial dangerous pheromone corresponds to the highest information intensity;

[0200] In a first range centered on the first reference position, deploy distance attenuation pheromones associated with the initial dangerous pheromone; the information intensity corresponding to the distance attenuation pheromones is negatively correlated with the distance between the deployment position of the distance attenuation pheromones and the first reference position;

[0201] Deploy time attenuation pheromones at historical positions, and in a second range centered on the historical positions, deploy distance attenuation pheromones associated with the time attenuation pheromones; the historical positions are any of the positions of the imaginary enemy model at n moments before the reference moment; the information intensity corresponding to the time attenuation pheromones is negatively correlated with the time interval between the moment corresponding to the time attenuation pheromones and the reference moment.

[0202] Optionally, the pheromone image generation module 1602 is specifically configured to generate the safe pheromone image in the following manner:

[0203] Based on the playback data, determine the position of the friendly character of the target virtual character at the reference moment and the positions of the friendly character at n moments before the reference moment; n is an integer greater than or equal to 1;

[0204] Based on the position of the friendly character at the reference moment and the positions of the friendly character at n moments before the reference moment, determine the safe pheromone image.

[0205] Optionally, the pheromone image generation module 1602 is specifically configured to:

[0206] In the map image corresponding to the game session, deploy initial safety pheromones at a second reference position; the second reference position is the position of the own character at the reference moment; the initial safety pheromones correspond to the highest pheromone intensity.

[0207] In a third range centered on the second reference position, deploy distance attenuation pheromones associated with the initial safety pheromones; the pheromone intensity corresponding to the distance attenuation pheromones is negatively correlated with the distance between the deployment position of the distance attenuation pheromones and the second reference position.

[0208] Deploy time attenuation pheromones at historical positions, and in a fourth range centered on the historical positions, deploy distance attenuation pheromones associated with the time attenuation pheromones; the historical positions are the positions of the own character at any one of the n moments before the reference moment; the pheromone intensity corresponding to the time attenuation pheromones is negatively correlated with the time interval between the moment corresponding to the time attenuation pheromones and the reference moment.

[0209] Optionally, the pheromone image generation module 1602 is further configured to:

[0210] Add information about target buildings belonging to the enemy to the dangerous pheromone image.

[0211] Add information about target buildings belonging to the own side to the safety pheromone image.

[0212] Optionally, the target moment is any moment within a suspicious segment in the game session; the device further includes a segment mining module, and the segment mining module is configured to:

[0213] Obtain the measurement conditions for the out-of-sight approach segment; the out-of-sight approach segment is a game segment in which a virtual character approaches the enemy character without its own field of vision covering the enemy character.

[0214] According to the playback data, detect the game segments in the game session where the target virtual character meets the measurement conditions as the suspicious segments corresponding to the target virtual character.

[0215] Optionally, the measurement conditions include start-phase conditions, intermediate-phase conditions, and end-phase conditions; the start-phase conditions are used to specify a distance threshold that needs to be satisfied between the virtual character and the enemy character, and the virtual character has no line of sight to the enemy character; the intermediate-phase conditions are used to specify that the distance between the virtual character and the enemy character shows a decreasing trend; the end-phase conditions are used to specify that the distance between the virtual character and the enemy character shows an increasing trend; then the segment mining module is specifically used for:

[0216] According to the replay data, detect game segments in the game session where the target virtual character correspondingly satisfies the start-phase conditions, the intermediate-phase conditions, and the end-phase conditions, and use them as the suspicious segments corresponding to the target virtual character.

[0217] Optionally, the device further includes:

[0218] A feature extraction module, configured to determine the macro-game features and micro-game features corresponding to the target virtual character at the reference moment according to the replay data; the macro-game features are used to characterize the global game information obtained by the target virtual character through the game global map at the reference moment; the micro-game features are used to characterize the local game information obtained by the target virtual character from its own perspective at the reference moment;

[0219] Then the decision intention prediction module 1603 is specifically used for:

[0220] Based on the pheromone image, the macro-game features, and the micro-game features corresponding to the target virtual character at the reference moment, predict the reference decision intention of the target virtual character at the target moment.

[0221] Optionally, the reference moment includes m moments before the target moment, where m is an integer greater than 1; the decision intention prediction module 1603 is specifically used for:

[0222] Arrange the pheromone images corresponding to the m moments in the order of their occurrence in the game session to obtain a pheromone image sequence, arrange the macro-game features corresponding to the m moments to obtain a macro-feature sequence, and arrange the micro-game features corresponding to the m moments to obtain a micro-feature sequence;

[0223] Through a decision intention prediction model, determine the reference decision intention of the target virtual character at the target moment according to the pheromone image sequence, the macro-feature sequence, and the micro-feature sequence.

[0224] Optionally, the device further includes a model training module, and the model training module is used for:

[0225] Obtain training samples; the training samples include a training pheromone image sequence, a training macro feature sequence, a training micro feature sequence, and corresponding external annotation results;

[0226] Through the decision intention prediction model as a generator, determine a predicted decision intention according to the training pheromone image sequence, the training macro feature sequence, and the training micro feature sequence in the training samples;

[0227] Through an external recognition model as a discriminator, determine an external prediction result according to the predicted decision intention;

[0228] Construct a loss function based on the external prediction result and the external annotation result;

[0229] Based on the loss function, train at least one of the decision intention prediction model and the external recognition model.

[0230] The data processing device provided by the embodiment of the present application innovatively proposes a mechanism for detecting game cheats based on pheromone images. Specifically, in this device, the data acquisition module first acquires the replay data of the game session, and this replay data is used to restore the behaviors of the virtual characters and the game environment in the game session. Then, the pheromone image generation module generates, for the target virtual character in the game session, a pheromone image corresponding to the reference moment in the game session according to the replay data; the pheromone image is a new concept proposed by the embodiment of the present application and is used to represent the game environment information perceived by the target virtual character at the reference moment; it should be noted that the game environment information represented by the pheromone image is different from the actual game environment information at the reference moment. The game environment information represented by the pheromone image is the information obtained from the perspective of the target virtual character based on the perception of the target virtual character, corresponding to the perception of the target virtual character from its own perspective, and the information not covered by the perspective of the target virtual character cannot be accurately represented. Furthermore, the decision intention prediction module predicts the reference decision intention of the target virtual character at the target moment after the reference moment based on the pheromone image corresponding to the target virtual character at the reference moment; that is, starting from the perspective of the target virtual character, it predicts the decision intention that should be generated in the currently perceived game environment. Finally, the cheat detection module compares the reference decision intention of the target virtual character at the target moment with the actual decision intention of the target virtual character at the target moment (determined according to the replay data), and determines whether the target virtual character uses a cheat based on the comparison result; it should be understood that the above comparison operation of the decision intention aims to compare the actual decision intention generated by the target virtual character in the game session with the decision intention that should be generated by the target virtual character in the currently perceived game environment. If the two are inconsistent, it means that the game environment information relied on by the target virtual character when generating the above actual decision intention in the game session is not the game environment information it perceives. It may have obtained richer and more accurate game environment information through a cheat and generated a decision intention inconsistent with the actually perceived game environment information, thus realizing accurate and effective detection of game cheats.

[0231] The embodiment of the present application also provides a computer device for detecting game cheats. This computer device can specifically be a terminal device or a server. The terminal device and the server provided by the embodiment of the present application will be introduced from the perspective of hardware implementation below.

[0232] See Figure 17 , Figure 17 is a schematic structural diagram of the terminal device provided by the embodiment of the present application. As Figure 17As shown, for ease of explanation, only parts related to the embodiments of the present application are shown. For specific technical details not disclosed, please refer to the method part of the embodiments of the present application. The terminal can be any terminal device including a mobile phone, a tablet computer, a personal digital assistant (PDA), a point of sales (POS), an in-vehicle computer, etc. Taking the terminal as a computer as an example:

[0233] Figure 17 The figure shows a block diagram of a part of the structure of a computer related to the terminal provided by the embodiments of the present application. Refer to Figure 17 , the computer includes: a radio frequency (RF) circuit 1710, a memory 1720, an input unit 1730 (including a touch panel 1731 and other input devices 1732), a display unit 1740 (including a display panel 1741), a sensor 1750, an audio circuit 1760 (which can be connected to a speaker 1761 and a microphone 1762), a wireless fidelity (WiFi) module 1770, a processor 1780, and a power supply 1790 and other components. Those skilled in the art can understand that Figure 17 the computer structure shown in does not constitute a limitation on the computer, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0234] The memory 1720 can be used to store software programs and modules. The processor 1780 executes various functional applications and data processing of the computer by running the software programs and modules stored in the memory 1720. The memory 1720 may mainly include a program storage area and a data storage area. Among them, the program storage area can store an operating system, applications required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the computer (such as audio data, a phone book, etc.). In addition, the memory 1720 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0235] The processor 1780 is the control center of the computer, connecting various parts of the entire computer using various interfaces and circuits. By running or executing software programs and / or modules stored in the memory 1720, and by invoking the data stored in the memory 1720, it performs various functions of the computer and processes data. Optionally, the processor 1780 may include one or more processing units; preferably, the processor 1780 may integrate an application processor and a modem processor. Among them, the application processor mainly processes the operating system, user interface, application programs, etc., and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor may not be integrated into the processor 1780 either.

[0236] In the embodiment of the present application, the processor 1780 included in the terminal is further configured to execute the steps of any implementation manner of the data processing method provided in the embodiment of the present application.

[0237] See Figure 18 , Figure 18 which is a schematic structural diagram of a server 1800 provided in the embodiment of the present application. The server 1800 may vary greatly due to configuration or performance differences, and may include one or more central processing units (CPUs) 1822 (for example, one or more processors) and a memory 1832, and one or more storage media 1830 (for example, one or more mass storage devices) for storing application programs 1842 or data 1844. Among them, the memory 1832 and the storage media 1830 may be transient storage or persistent storage. The programs stored in the storage media 1830 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Further, the central processor 1822 may be configured to communicate with the storage media 1830 and execute a series of instruction operations in the storage media 1830 on the server 1800.

[0238] The server 1800 may further include one or more power supplies 1826, one or more wired or wireless network interfaces 1850, one or more input / output interfaces 1858, and / or, one or more operating systems, such as Windows Server TM , Mac OS X TM , Unix TM , Linux TM , FreeBSD TM and so on.

[0239] The steps performed by the server in the above embodiments may be based on the Figure 18 server structure shown.

[0240] Among them, the CPU 1822 can also be used to execute the steps of any implementation manner of the data processing method provided in the embodiments of the present application.

[0241] The embodiments of the present application also provide a computer-readable storage medium for storing a computer program, and the computer program is used to execute any implementation manner of the data processing method described in the foregoing embodiments.

[0242] The embodiments of the present application also provide a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes any implementation manner of the data processing method described in the foregoing embodiments.

[0243] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0244] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some interfaces, and the indirect coupling or communication connection of the device or unit can be in an electrical, mechanical or other form.

[0245] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0246] In addition, the functional units in the various embodiments of the present application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0247] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media that can store computer programs, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0248] It should be understood that in this application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects and indicates that three relationships can exist. For example, "A and / or B" can represent: only A exists, only B exists, and both A and B exist simultaneously. Among them, A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one (item) of the following" or its similar expression refers to any combination of these items, including any combination of single item (s) or plural item (s). For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0249] As described above, the above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit them; although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of various embodiments of this application.

Claims

1. A data processing method, characterized in that, The method includes: Obtaining replay data of a game session; the replay data is used to restore the behaviors of virtual characters and the game environment in the game session. For a target virtual character in the game session, generating a pheromone image corresponding to a reference moment in the game session according to the replay data; the pheromone image is used to represent the game environment information perceived by the target virtual character at the reference moment. Based on the pheromone image corresponding to the reference moment of the target virtual character, predicting a reference decision intention of the target virtual character at a target moment in the game session; the target moment is after the reference moment. Comparing the reference decision intention of the target virtual character at the target moment with the actual decision intention of the target virtual character at the target moment, and determining whether the target virtual character uses a cheating program according to the comparison result.

2. The method according to claim 1, wherein The generating the pheromone image corresponding to the reference moment of the target virtual character in the game session according to the replay data includes: Generating a dangerous pheromone image and a safe pheromone image corresponding to the reference moment of the target virtual character according to the replay data; the dangerous pheromone image is used to represent the enemy information perceived by the target virtual character at the reference moment, and the safe pheromone image is used to represent the friendly information perceived by the target virtual character at the reference moment.

3. The method according to claim 2, wherein The dangerous pheromone image is generated by the following method: According to the replay data, determining the enemy characters perceived by the target virtual character, and constructing a hypothetical enemy model corresponding to the enemy characters; the hypothetical enemy model is used to simulate the enemy characters perceived by the target virtual character. When the field of view of the target virtual character does not cover the enemy characters, the hypothetical enemy model moves towards the target virtual character at a hypothetical speed and along a hypothetical path. Determining the dangerous pheromone image according to the position of the hypothetical enemy model at the reference moment and the positions of the hypothetical enemy model at n moments before the reference moment; n is an integer greater than or equal to 1.

4. The method according to claim 3, characterized in that The determining the dangerous pheromone image according to the position of the hypothetical enemy model at the reference moment and the positions of the hypothetical enemy model at n moments before the reference moment includes: Deploying initial dangerous pheromones at a first reference position in the map image corresponding to the game session; the first reference position is the position of the hypothetical enemy model at the reference moment; the initial dangerous pheromones correspond to the highest information intensity. Deploying distance attenuation pheromones associated with the initial dangerous pheromones within a first range centered on the first reference position; the information intensity corresponding to the distance attenuation pheromones is negatively correlated with the distance between the deployment position of the distance attenuation pheromones and the first reference position. Deploy time-decaying pheromones at a historical position, and within a second range centered on the historical position, deploy distance-decaying pheromones associated with the time-decaying pheromones; the historical position is the position of the enemy model at any one of the n moments before the reference moment; the information intensity corresponding to the time-decaying pheromones is negatively correlated with the time interval between the moment corresponding to the time-decaying pheromones and the reference moment.

5. The method according to claim 2, characterized in that The safety pheromone image is generated in the following manner: Based on the replay data, determine the position of the friendly character of the target virtual character at the reference moment, and the positions of the friendly character at n moments before the reference moment; n is an integer greater than or equal to 1; Based on the position of the friendly character at the reference moment and the positions of the friendly character at n moments before the reference moment, determine the safety pheromone image.

6. The method according to claim 5, wherein The determining the safety pheromone image based on the position of the friendly character at the reference moment and the positions of the friendly character at n moments before the reference moment includes: In the map image corresponding to the game session, deploy initial safety pheromones at a second reference position; the second reference position is the position of the friendly character at the reference moment; the initial safety pheromones correspond to the highest information intensity; Within a third range centered on the second reference position, deploy distance-decaying pheromones associated with the initial safety pheromones; the information intensity corresponding to the distance-decaying pheromones is negatively correlated with the distance between the deployment position of the distance-decaying pheromones and the second reference position; Deploy time-decaying pheromones at a historical position, and within a fourth range centered on the historical position, deploy distance-decaying pheromones associated with the time-decaying pheromones; the historical position is the position of the friendly character at any one of the n moments before the reference moment; the information intensity corresponding to the time-decaying pheromones is negatively correlated with the time interval between the moment corresponding to the time-decaying pheromones and the reference moment.

7. The method according to any one of claims 2 to 6, characterized in that, The method further includes: Add information about the target buildings belonging to the enemy on the dangerous pheromone image; Add information about the target buildings belonging to the friendly side on the safety pheromone image.

8. The method according to claim 1, characterized in that, The target moment is any moment within the suspicious segment in the game session; the suspicious segment is mined in the following manner: Obtain the measurement conditions for the out-of-sight approaching segment; the out-of-sight approaching segment is a game segment in which a virtual character approaches an enemy character without its own field of vision covering the enemy character; Based on the replay data, detect the game segments in the game session where the target virtual character meets the measurement conditions as the suspicious segments corresponding to the target virtual character.

9. The method according to claim 8, wherein The measurement conditions include start - stage conditions, middle - stage conditions, and end - stage conditions; the start - stage conditions are used to specify the distance threshold that needs to be met between the virtual character and the enemy character, and the virtual character has no line of sight to the enemy character; the middle - stage conditions are used to specify that the distance between the virtual character and the enemy character shows a decreasing trend; the end - stage conditions are used to specify that the distance between the virtual character and the enemy character shows an increasing trend; Detecting, according to the replay data, the game segments in which the target virtual character meets the measurement conditions in the game session as the suspicious segments corresponding to the target virtual character, includes: Detecting, according to the replay data, the game segments in which the target virtual character correspondingly meets the start - stage conditions, the middle - stage conditions, and the end - stage conditions in the game session as the suspicious segments corresponding to the target virtual character.

10. The method according to claim 1 or 2, characterized in that, The method further includes: Determining, according to the replay data, the macro - game characteristics and micro - game characteristics corresponding to the target virtual character at the reference time; the macro - game characteristics are used to characterize the global game information obtained by the target virtual character through the game global map at the reference time; the micro - game characteristics are used to characterize the local game information obtained by the target virtual character from its own perspective at the reference time; Predicting the reference decision - making intention of the target virtual character at the target time in the game session based on the pheromone image corresponding to the target virtual character at the reference time, includes: Predicting the reference decision - making intention of the target virtual character at the target time based on the pheromone image, the macro - game characteristics, and the micro - game characteristics corresponding to the target virtual character at the reference time.

11. The method according to claim 10, wherein The reference time includes m moments before the target time, where m is an integer greater than 1; predicting the reference decision - making intention of the target virtual character at the target time based on the pheromone image, the macro - game characteristics, and the micro - game characteristics corresponding to the target virtual character at the reference time, includes: Arranging the pheromone images corresponding to the m moments in the order of their occurrence in the game session to obtain a pheromone image sequence, arranging the macro - game characteristics corresponding to the m moments to obtain a macro - characteristic sequence, and arranging the micro - game characteristics corresponding to the m moments to obtain a micro - characteristic sequence; Determining the reference decision - making intention of the target virtual character at the target time through a decision - making intention prediction model according to the pheromone image sequence, the macro - characteristic sequence, and the micro - characteristic sequence.

12. The method according to claim 11, wherein The decision - making intention prediction model is trained in the following way: Obtaining training samples; the training samples include a training pheromone image sequence, a training macro - characteristic sequence, a training micro - characteristic sequence, and the corresponding cheating annotation results; Through the decision intention prediction model as a generator, determine a predicted decision intention according to the training pheromone image sequence, the training macro feature sequence, and the training micro feature sequence in the training sample; Through the external cheat recognition model as a discriminator, determine an external cheat prediction result according to the predicted decision intention; Based on the external cheat prediction result and the external cheat annotation result, construct a loss function; Based on the loss function, train at least one of the decision intention prediction model and the external cheat recognition model.

13. A data processing device, characterized in that, The device includes: A data acquisition module, configured to acquire replay data of a game session; the replay data is used to restore the behaviors of virtual characters and the game environment in the game session; A pheromone image generation module, configured to generate, for a target virtual character in the game session, a pheromone image corresponding to a reference moment in the game session according to the replay data; the pheromone image is used to represent the game environment information perceived by the target virtual character at the reference moment; A decision intention prediction module, configured to predict a reference decision intention of the target virtual character at a target moment in the game session based on the pheromone image corresponding to the reference moment of the target virtual character; the target moment is after the reference moment; An external cheat detection module, configured to compare the reference decision intention of the target virtual character at the target moment with the actual decision intention of the target virtual character at the target moment, and determine whether the target virtual character uses an external cheat according to the comparison result.

14. A computer device, characterized in that, The device includes a processor and a memory; The memory is used to store a computer program; The processor is configured to execute the data processing method according to any one of claims 1 to 12 according to the computer program.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program, and the computer program is used to execute the data processing method according to any one of claims 1 to 12.

16. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or the instruction is executed by the processor, the data processing method according to any one of claims 1 to 12 is implemented.

Images