Game illegal behavior detection and identification method and system based on AI

Abstract

The invention relates to an AI-based game illegal behavior detection and identification method and system, and relates to the technical field of game security, and the method comprises the following steps: collecting a game type and a game account; determining a demand acquisition type according to the game type; implementing acquisition according to the demand acquisition type to obtain actual acquisition data; historical game data are called in combination with the game account; inputting the historical game data into a preset adaptive AI model cluster for training, and generating an account violation detection model cluster; generating a suspected violation probability in combination with the actually acquired data and the account violation detection model cluster; and generating violation processing information by referring to a falling condition between the suspected violation probability and a preset violation processing probability interval, and outputting the violation processing information. The method and the device have the effect of timely and accurately detecting the game violation behavior.

Description

Technical Field

[0001] This invention relates to the field of game security technology, and in particular to an AI-based method and system for detecting and identifying game violations. Background Technology

[0002] Game security refers to a comprehensive security system that ensures the stable operation of game systems, protects players' rights, and maintains a fair and healthy game environment throughout the entire lifecycle of electronic games (including client games, mobile games, web games, cloud games, and other forms) in terms of development, operation, maintenance, and player interaction. This system utilizes technical means, management mechanisms, and rule systems to prevent various security risks and violations.

[0003] In multiplayer competitive games, player violations such as using cheats, engaging in passive gameplay, or using abusive language severely damage the fair competitive environment and negatively impact the gaming experience of other players. Therefore, it is necessary to detect and punish these violations. Currently, a series of rules are typically pre-established, and whether a player's behavior conforms to these rules is used to determine whether a violation has occurred.

[0004] When a set of pre-defined rules is used to detect game violations, the rules are often manually formulated, which requires a lot of time and effort and is difficult to cover all possible violations. Furthermore, violators can easily bypass the existing rule detection by adjusting their cheating methods, making it difficult to detect game violations in a timely and accurate manner. Summary of the Invention

[0005] In order to detect game violations in a timely and accurate manner, this invention provides an AI-based method and system for detecting and identifying game violations.

[0006] Firstly, the present invention provides an AI-based method for detecting and identifying game violation behaviors, employing the following technical solution: An AI-based method for detecting and identifying game violations includes: S1: Collect game type and game account; S2: Determine the type of data collection required based on the game type; S3: Collect data according to the aforementioned data collection type to obtain actual collected data; S4: Retrieve historical game data based on the aforementioned game account; S5: Input the historical game data into a preset adaptive AI model cluster for training to generate an account violation detection model cluster; S6: Combine the actual collected data with the account violation detection model cluster to generate a suspected violation probability; S7: Based on the occurrence of the suspected violation probability and the preset violation handling probability range, generate violation handling information and output the violation handling information.

[0007] Optionally, the method for generating the suspected violation probability includes: S61: Retrieve the actual collection type and single-type data based on the actual collected data; S62: Select a single-type detection model from the account violation detection model cluster based on the actual collection type; S63: Input the single-type data into the single-type detection model to obtain the single-type probability; S64: Combine the actual data collection type with the required data collection type to generate a type influence weight; S65: The single-type probability and the type influence weight are weighted and calculated to obtain the comprehensive type probability, and the comprehensive type probability is used as the suspected violation probability.

[0008] Optionally, the method for generating the type influence weights includes: S641: Compare the actual collection type with the required collection type to obtain the actual overlapping type and the actual missing type; S642: Generate a type adjustment value by combining the actual overlapping type and the actual missing type; S643: Retrieve the type importance value corresponding to each of the aforementioned requirement collection types based on the game type; S644: Determine the type importance weight with reference to the type importance value; S645: Adjust the type importance weight with the type adjustment value to obtain the actual type weight, and use the actual type weight as the type influence weight.

[0009] Optionally, the method for generating the type adjustment value includes: S6421: Retrieve the numerical values ​​of the overlapping types based on the actual overlapping types; S6422: Retrieve the number of missing values ​​based on the actual missing type; S6423: Calculate the ratio of the number of overlapping types to the number of missing types to obtain the overlap ratio; S6424: Determine the overlap ratio adjustment value with reference to the aforementioned overlap ratio value; S6425: Adjust the preset overlap reference adjustment value using the overlap ratio adjustment value to form an overlap adjustment value; S6426: Adjust the preset missing baseline adjustment value using the overlap ratio adjustment value to form a missing adjustment value; S6427: Combine the overlap adjustment value and the missing adjustment value as the type adjustment value.

[0010] Optionally, after adjusting the type importance weight with the type adjustment value to obtain the actual type weight, the method further includes: S6451: Retrieve basic account information and account activity time based on the aforementioned game account; S6452: Determine the baseline active time corresponding to each of the aforementioned demand collection types based on the account baseline information; S6453: Compare the baseline active time with the account active time to determine the active deviation time value; S6454: Determine the active deviation adjustment value with reference to the active deviation time value; S6455: Adjust and update the actual type weights using the active deviation adjustment value.

[0011] Optionally, the method for determining the baseline active time includes: S64521: Retrieve the account age value, account historical location points, and account friend information based on the aforementioned account baseline information; S64522: Collect the current location of the account; S64523: Generate a position deviation adjustment value by combining the current location of the account with the historical location of the account; S64524: Determine the age baseline time corresponding to each of the aforementioned demand collection types based on the account age value; S64525: Retrieve the active time of friends based on the account's friend information; S64526: Generate an active adjustment time by combining the age baseline time with the friend's active time; S64527: Combine the position deviation adjustment value with the active adjustment time to generate a position adjustment time, and use the position adjustment time as the reference active time.

[0012] Optionally, the method for generating the position deviation adjustment value includes: S645231: Compare the current location of the account with the historical location of the account to determine the location deviation distance value; S645232: Determine the age baseline distance range based on the account age value; S645233: Determine whether the positional deviation distance value is within the age reference distance range; S645234: If yes, then determine the distance movement adjustment value based on the position deviation distance value, and use the distance movement adjustment value as the position deviation adjustment value; S645235: If not, then determine the abnormal deviation distance value by combining the position deviation distance value with the age reference distance range; S645236: Collect the current time point; S645237: Determine the abnormal distance adjustment value by combining the current time point with the abnormal deviation distance value, and use the abnormal distance adjustment value as the position deviation adjustment value.

[0013] Optionally, the method for generating the active adjustment time includes: S645261: When the age reference time overlaps with the friend's active time, compare the age reference time with the friend's active time to determine the active overlap time; S645262: Compare the active overlap time with the age baseline time to determine the overlap ratio; S645263: Determine the overlap adjustment value with reference to the aforementioned overlap ratio value; S645264: Retrieve the number of overlapping values ​​based on the active overlap time; S645265: Determine the number of influence values ​​by referring to the aforementioned overlap value; S645266: Determine the overlap correction value by combining the number influence value and the overlap adjustment value; S645267: Adjust the age reference time using the overlap correction value to obtain the overlap adjustment time, and use the overlap adjustment time as the active adjustment time.

[0014] Optionally, the method for generating the position adjustment time includes: S645271: Retrieve the adjustment start time and adjustment end time based on the active adjustment time; S645272: Compare the adjustment start time point with the adjustment end time point to determine the time span value; S645273: Determine the adjustment benchmark interval with reference to the aforementioned time span value; S645274: Determine whether the position deviation adjustment value falls within the adjustment reference range; S645275: If yes, then the position deviation adjustment value is used to adjust the adjustment start time point and the adjustment end time point in the same direction to form the position adjustment time; S645276: If not, then the position deviation adjustment value is used to perform convergent adjustment on the adjustment start time point and the adjustment end time point to form the position adjustment time.

[0015] Secondly, this invention provides an AI-based system for detecting and recognizing game violations, employing the following technical solution: An AI-based system for detecting and identifying game violations includes: The data collection module is used to collect information such as game type, game account, actual data collected, current account location, and current time. The memory stores a program for implementing an AI-based method for detecting and identifying game violations as described in any one of the first aspects; The processor loads and executes programs stored in memory.

[0016] In summary, the present invention has at least one of the following beneficial technical effects: 1. After collecting data on game types and game accounts and determining the required data collection type, the actual data is collected according to the required data collection type. Then, historical game data is retrieved and input into a preset adaptive AI model cluster for training to generate an account violation detection model cluster. The actual collected data and the account violation detection model cluster are used to generate a suspected violation probability. The violation handling information is generated and output based on the suspected violation probability falling within the preset violation handling probability range. This allows for accurate judgment based on the user situation, thereby timely and accurately detecting game violations. 2. By retrieving actual collected data and single-type data, after selecting a single-type detection model, the single-type data is input to obtain the single-type probability. Then, the type influence weight is generated by the actual collected type and the required collected type. The comprehensive type probability is obtained by weighted calculation and used as the suspected violation probability, thereby improving the accuracy of the obtained suspected violation probability. 3. By comparing the actual collected types with the required collected types, the actual overlapping types and the actual missing types are obtained. Then, type adjustment values ​​are generated based on the actual overlapping types and the actual missing types. The type importance value is retrieved based on the game type to determine the type importance weight. The type adjustment value is then used to adjust the type importance weight to obtain the actual type weight, which is used as the type influence weight, thereby improving the accuracy of the obtained type influence weight. Attached Figure Description

[0017] Figure 1 This is a flowchart of an AI-based method for detecting and identifying game violations. Figure 2 This is a flowchart illustrating the method for generating the probability of suspected violations; Figure 3 This is a flowchart of the method for generating weights based on type influence. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0019] An AI-based method for detecting and identifying game violations involves collecting data on game type, game account, actual data collection, the account's current location, and the current time. Based on the game type, the required data collection type is determined, and actual data is collected. Simultaneously, historical game data of the account is retrieved and input into a pre-set adaptive AI model cluster for training, generating an account violation detection model cluster. Next, combining the actual collected data, a single-type detection model is selected from the model cluster according to the actual collected type to obtain the single-type probability. Then, by comparing the actual collected type with the required collected type and combining the type importance value corresponding to the game type, a type influence weight is generated, and a weighted calculation is performed to obtain the suspected violation probability. During this process, the weights and baseline active time are further optimized by incorporating account baseline information, active time, location, and friend information. Finally, based on whether the suspected violation probability falls within a pre-set violation processing probability range, violation processing information is generated and output. This allows for precise judgment based on user circumstances, thereby enabling timely and accurate detection of game violations.

[0020] Reference Figure 1 This invention discloses an AI-based method for detecting and identifying game violations, comprising: S1: Collect game type and game account.

[0021] The game type refers to the category of the game for which violation detection is currently required. Game types include FPS (First-Person Shooter) games, MOBA (Multiplayer Online Battle Arena) games, and RPG (Role-Playing Game) games. A game account is a unique identifier registered by a player on a game platform or in a specific game. Game accounts are linked to a player's personal information (such as registered phone number and nickname), game behavior data (such as historical operation records and battle results), and account attributes (such as level, rank, and assets). They are the core basis for locating individual players and tracing their historical game behavior.

[0022] The game type is read directly from the game's client or server backend, while the game account is automatically obtained when the player logs in.

[0023] S2: Determine the type of data collection required based on the game type.

[0024] Among them, the data collection type refers to the characteristics of violations and core gameplay logic of a specific game type, and the data categories that need to be collected in advance for violation detection.

[0025] Different game types correspond to different data collection requirements. By inputting the game type into a preset data collection database, the data collection requirements can be matched and obtained for subsequent use.

[0026] The type acquisition database pre-stores a mapping table of different game types and their corresponding acquisition requirements. The type acquisition database is pre-configured by the operator according to actual needs.

[0027] For example, when the game type is FPS, the data collection type focuses on data such as mouse trajectory and field of view angle; when the game type is MOBA, the data collection type focuses on data such as resource acquisition speed and skill release interval.

[0028] S3: Collect data according to the required data collection type to obtain the actual collected data.

[0029] The actual collected data refers to the specific data content that is actually obtained during the game's operation and used for subsequent violation detection.

[0030] By selecting and collecting data from the game client or server based on the required data collection type, it becomes easier to use it later.

[0031] For example, when the data collection type is shooting angle data in an FPS game, the angle value of the character's gun muzzle (such as horizontal angle or vertical angle) is collected to obtain the actual data; when the data collection type is resource acquisition data in a MOBA game, the gold coin increase value and experience value change after the player kills monsters / minions are captured in real time.

[0032] S4: Retrieve historical game data by combining game account.

[0033] Historical game data refers to various data records generated and stored by the game account in the past game process. Historical game data includes the account's historical operation behavior (such as mouse movement trajectory and skill release frequency in the past 30 days), changes in game status (such as historical rank fluctuations and resource acquisition records), and interaction information (such as historical teammate and chat history).

[0034] Historical game data can be retrieved from the game account's historical data for later use.

[0035] S5: Input historical game data into a preset adaptive AI model cluster for training to generate an account violation detection model cluster.

[0036] The adaptive AI model cluster refers to a pre-built collection of AI models targeting various game violation scenarios. This cluster possesses dynamic learning capabilities, automatically adjusting internal parameters (such as model weights and detection thresholds) based on the characteristics of the input data to adapt to the behavioral patterns of different game accounts (e.g., automatically relaxing some behavioral judgment standards for aggressive accounts, and increasing the detection sensitivity of sensitive actions for stable accounts). The adaptive AI model cluster is pre-configured by the operator according to their needs.

[0037] An account violation detection model cluster refers to a set of models trained on historical game data of a specific game account, specifically designed for detecting violations by that account. The model structure of the account violation detection model cluster corresponds to the preset adaptive AI model cluster, but the parameters have been optimized based on the account's behavioral baseline (such as the frequency of regular operations and skill release habits), enabling accurate identification of the account's abnormal behavior.

[0038] By cleaning and extracting features from historical game data, and labeling it with the account's historical violation records, a training sample set is formed. The preprocessed samples are then categorized according to violation scenarios and input into the corresponding sub-models in a pre-defined adaptive AI model cluster for training, thereby obtaining violation detection sub-models for each violation scenario. Finally, these violation detection sub-models are combined to form an account violation detection model cluster.

[0039] For example, the standard deviation of operations and the volatility of skill release intervals from historical game data are input into the sub-model corresponding to the boosting scenario in the preset adaptive AI model cluster for training, thereby obtaining a violation detection sub-model for the boosting scenario. Dialogue data from historical game data are input into the sub-model corresponding to the insult scenario in the preset adaptive AI model cluster for training, thereby obtaining a violation detection sub-model for the insult scenario. Then, the violation detection sub-model for the boosting scenario and the violation detection sub-model for the insult scenario are combined to obtain an account violation detection model cluster for convenient subsequent use.

[0040] S6: Combine actual collected data with the account violation detection model cluster to generate the probability of suspected violations.

[0041] Among them, the probability of suspected violation refers to a quantitative parameter for judging whether the current player has violated any rules.

[0042] By inputting the actual collected data into the sub-models in the corresponding account violation detection model cluster and performing probability calculations, the overall violation probability is obtained and used as the suspected violation probability for subsequent use.

[0043] To further ensure the reasonableness of the suspected violation probability, it is necessary to conduct a further separate analysis and calculation of the suspected violation probability, which will be explained in detail through the steps shown below.

[0044] Reference Figure 2 The method for generating the probability of suspected violations includes the following steps: S61: Retrieve actual collected data and single-type data based on actual collected data.

[0045] In this context, "actual acquisition type" refers to the category of data successfully acquired during the actual data collection process, based on the required acquisition type and considering the actual conditions of the equipment and scenario. "Single-type data" refers to specific data content belonging to a particular actual acquisition type within the actual acquisition data. Actual acquisition data includes both actual acquisition types and single-type data.

[0046] The actual collected data is retrieved to identify the actual data type and single-type data, making it convenient for subsequent use.

[0047] S62: Select a single-type detection model from the account violation detection model cluster based on the actual collection type.

[0048] Among them, the single-type detection model refers to a sub-model that performs violation detection on a single type of data.

[0049] By selecting sub-models from the account violation detection model cluster that require data corresponding to the actual collection type and using them as single-type detection models, it is easier to use them in the future.

[0050] For example, when the actual data collected is dialogue data, the violation detection sub-model for the abusive language scenario is selected as the single-type detection model. When the actual data collected is the skill release interval fluctuation rate, the violation detection sub-model for the substitute training scenario is selected as the single-type detection model.

[0051] S63: Input single-type data into the single-type detection model to obtain the single-type probability.

[0052] Among them, single-type probability refers to the probability value of a player's violation behavior characteristics based on single-type data. The value range of single-type probability is usually from 0 to 1.

[0053] By inputting single-type data into a single-type detection model, the single-type detection model preprocesses and analyzes the single-type data to obtain the violation probability, which is then used as the single-type probability for convenient subsequent use.

[0054] S64: Combine the actual data collection type with the required data collection type to generate the type influence weight.

[0055] Among them, the type influence weight refers to the parameter value used to measure the importance of the violation probability of the actual collection type in violation detection.

[0056] By combining and analyzing factors such as the degree of matching and importance between the actual data collection type and the required data collection type, a weight value is assigned to each actual data collection type and used as the type influence weight for convenient subsequent use.

[0057] To further ensure the rationality of the type influence weight, it is necessary to perform a further separate analysis and calculation of the type influence weight, which will be explained in detail through the steps shown below.

[0058] Reference Figure 3 The method for generating type-affected weights includes the following steps: S641: Compare the actual data collection type with the required data collection type to obtain the actual overlapping types and the actual missing types.

[0059] Among them, "actual overlapping type" refers to the type that was successfully collected for the specific requirement collection type. "Actual missing type" refers to the type that was not collected for the specific requirement collection type.

[0060] By comparing the actual data collection type with the required data collection type, the types that are successfully compared are taken as the actual overlapping types, and the types that are not successfully compared are taken as the actual missing types, which facilitates subsequent use.

[0061] S642: Generate type adjustment values ​​by combining the actual overlapping types and the actual missing types.

[0062] The type adjustment value refers to the adjustment value corresponding to the weight adjustment based on the type.

[0063] By combining and analyzing the actual overlapping types and the actual missing types, type adjustment values ​​are generated for convenient subsequent use.

[0064] To further ensure the rationality of the type adjustment value, it is necessary to perform a further separate analysis and calculation on the type adjustment value, which will be explained in detail through the steps shown below.

[0065] The method for generating type adjustment values ​​includes the following steps: S6421: Retrieve the number of values ​​for the overlapping types based on the actual overlapping types.

[0066] Among them, the number of overlapping types refers to the number of actual overlapping types.

[0067] By counting the actual overlapping types and using the count results as the number of overlapping types, it is convenient to use them later.

[0068] S6422: Retrieve the number of missing values ​​based on the actual missing type.

[0069] The number of missing types refers to the number of actual missing types.

[0070] By counting the actual missing types and using the count results as the number of missing types, it is convenient to use them later.

[0071] S6423: Calculate the ratio of the number of overlapping types to the number of missing types to obtain the overlap ratio.

[0072] The overlap ratio refers to the ratio of the number of overlapping types to the number of missing types.

[0073] Calculating the overlap ratio facilitates subsequent use.

[0074] S6424: Determine the overlap ratio adjustment value by referring to the overlap ratio value.

[0075] Among them, the overlap ratio adjustment value refers to the adjustment value set based on the overlap ratio value to adjust the weight.

[0076] By inputting the overlap ratio value into a preset ratio range database, the system analyzes whether the overlap ratio value falls within the preset ratio range, and selects the corresponding adjustment value based on the falling situation as the overlap ratio adjustment value.

[0077] The ratio range database pre-stores different ratio ranges and their corresponding adjustment values. The ratio range database is preset by the operator according to actual needs.

[0078] For example, the ratio range database can be set to pre-store: a ratio range of 0 to 0.2 with a corresponding adjustment value of 0.3; a ratio range of 0.2 to 0.5 with a corresponding adjustment value of 0.7; a ratio range of 0.5 to 0.8 with a corresponding adjustment value of 1; and a ratio range of 0.8 to 1 with a corresponding adjustment value of 1.2.

[0079] S6425: Adjust the preset overlap reference adjustment value using the overlap ratio adjustment value to form the overlap adjustment value.

[0080] The overlap benchmark adjustment value refers to a fixed reference value pre-set for the actual overlap type, used to adjust and correct the weights. The overlap benchmark adjustment value is pre-set by the operator according to actual needs. The overlap adjustment value refers to the adjusted value after adjusting the overlap benchmark adjustment value.

[0081] The product of the overlap ratio adjustment value and the overlap benchmark adjustment value is calculated, and the result is used as the overlap adjustment value for convenient subsequent use.

[0082] S6426: Adjust the preset missing baseline adjustment value using the overlap ratio adjustment value to form the missing adjustment value.

[0083] The missing baseline adjustment value refers to a fixed reference value pre-set for the actual missing type, used to adjust and correct the weights. The missing baseline adjustment value is pre-set by the operator according to actual needs. The missing adjustment value refers to the adjusted value after adjusting the missing baseline adjustment value.

[0084] The product of the overlap ratio adjustment value and the missing baseline adjustment value is calculated, and the result is used as the missing adjustment value for convenient subsequent use.

[0085] S6427: Combine the overlap adjustment value and the missing adjustment value as the type adjustment value.

[0086] By combining the overlap adjustment value and the missing adjustment value as the type adjustment value, it is convenient to select and use the overlap adjustment value or the missing adjustment value according to whether the types overlap when making subsequent adjustments, thereby improving the accuracy of the type influence weight obtained later.

[0087] S643: Retrieve the type importance value corresponding to each required data collection type based on the game type.

[0088] Among them, the type importance value refers to the degree of importance of each requirement collection type in the game type.

[0089] The system retrieves the corresponding game type database by game type and inputs the required data collection type into the game type database to obtain the type importance value, which is convenient for subsequent use.

[0090] The game type database pre-stores a mapping table of different requirement collection types and their corresponding type importance values. The game type database is pre-configured by the operator according to actual needs.

[0091] For example, when the game type is FPS, the type importance value for player action data can be set to 0.9, and the type importance value for in-game economy data can be set to 0.7. When the game type is RPG, the type importance value for character progression data can be set to 0.8, the type importance value for story-related data can be set to 0.7, and the type importance value for social interaction data can be set to 0.7.

[0092] S644: Determine the type importance weight by referring to the type importance value.

[0093] Among them, the type importance weight refers to the weight value that quantifies the contribution ratio of different types in violation detection based on the type importance value.

[0094] The sum of all type importance values ​​is calculated and used as the overall importance value. Then, the ratio between the type importance value and the overall importance value is calculated, and the result is used as the type importance weight, which facilitates subsequent use.

[0095] S645: Adjust the type importance weights using type adjustment values ​​to obtain the actual type weights, and use the actual type weights as the type influence weights.

[0096] The actual type weight refers to the weight value after adjusting the type importance weight.

[0097] By calculating the sum between the type adjustment value and the type importance weight, and using the calculation result as the actual type weight, and then using the actual type weight as the type influence weight, the accuracy of the obtained type influence weight is improved.

[0098] To further ensure the rationality of the actual type weights, it is necessary to perform further separate analysis and calculation on the actual type weights, which will be explained in detail through the steps shown below.

[0099] After adjusting the type importance weights using type adjustment values ​​to obtain the actual type weights, the following steps are also included: S6451: Combine game account information to retrieve basic account information and account activity time.

[0100] Account baseline information refers to information related to the basic attributes and status of a game account. This includes account age, historical location data, and friend information. Account activity time refers to data related to the time a player uses their game account for in-game activities.

[0101] The system retrieves the account's basic information and active time from the game account for future use.

[0102] S6452: Determine the baseline active time for each demand collection type based on the account baseline information.

[0103] The benchmark active time refers to the typical active time range of the account under different demand collection types.

[0104] By analyzing the account baseline information, the baseline active time corresponding to each type of demand collection is generated, which facilitates subsequent use.

[0105] To further ensure the rationality of the baseline active time, it is necessary to perform a further separate analysis and calculation of the baseline active time, which will be explained in detail through the steps shown below.

[0106] The method for determining the baseline active time includes the following steps: S64521: Retrieve account age, historical location points, and friend information based on account baseline information.

[0107] Here, account age refers to the time elapsed from account registration to the present. Account historical location points refer to the center point of the location range recorded each time the account logs in during its usage. Account friend information refers to information about other users who have established friend relationships with this account. Basic account information includes account age, account historical location points, and account friend information.

[0108] The system retrieves the account's age and friend information based on the account's baseline information, as well as the login location points. It then calculates the center point of each login location point and uses it as the account's historical location points for future use.

[0109] S64522: Collect the current location of the account.

[0110] The current location of the account refers to the location recorded when the account logs in at the current time.

[0111] The account's current location is obtained by querying the location of the game device when the account logs in at the current time.

[0112] S64523: Generate a position deviation adjustment value by combining the current position of the account with the historical position of the account.

[0113] Among them, the position deviation adjustment value refers to the adjustment value corresponding to the time adjustment when the position needs to be adjusted.

[0114] By combining and analyzing the account's current location with its historical location, a location deviation adjustment value is generated for convenient subsequent use.

[0115] To further ensure the rationality of the position deviation adjustment value, it is necessary to perform a further separate analysis and calculation on the position deviation adjustment value, which will be explained in detail through the steps shown below.

[0116] The method for generating position deviation adjustment values ​​includes the following steps: S645231: Compare the current location of the account with the historical location of the account to determine the location deviation distance value.

[0117] The location deviation distance value refers to the distance between the current location of the account and the historical location of the account.

[0118] The distance between the account's current location and its historical locations is calculated and used as the location deviation distance value for future use.

[0119] S645232: Determine the age baseline distance range based on the account age value.

[0120] The age baseline distance range refers to the range within which the age of the player corresponding to the account is allowed to move normally.

[0121] Different account age values ​​correspond to different age baseline distance ranges. Generally speaking, players of different age groups usually have fixed login regions, and as the account age value increases, the age baseline distance range will first increase and then decrease.

[0122] By inputting the account's age value into a preset age database, the system analyzes whether the account's age value falls within a preset age range, and uses the distance interval corresponding to the fall within the range as the age baseline distance interval for convenient subsequent use.

[0123] The age database pre-stores different age ranges and their corresponding distance intervals, and the age database is preset by the operator according to actual needs.

[0124] S645233: Determine whether the positional deviation distance value is within the age baseline distance range. If yes, execute S645234; if no, execute S645235.

[0125] Specifically, by judging whether the location deviation distance value is within the age benchmark distance range, it can be determined whether the player's login location is abnormal.

[0126] S645234: Determine the distance movement adjustment value based on the position deviation distance value, and use the distance movement adjustment value as the position deviation adjustment value.

[0127] Among them, the distance movement adjustment value refers to the adjustment value corresponding to the adjustment when adjusting based on the distance deviation.

[0128] When the position deviation distance value is within the age baseline distance range, it indicates that the player's login position is not abnormal. Therefore, by calculating the product between the position deviation distance value and the preset distance adjustment coefficient, and using the calculation result as the distance movement adjustment value, and then using the distance movement adjustment value as the position deviation adjustment value, the accuracy of the obtained position deviation adjustment value is improved.

[0129] The distance adjustment factor is a factor used to convert the position deviation distance value into a distance movement adjustment value. The distance adjustment factor is preset by the operator according to actual needs.

[0130] S645235: Determine the abnormal deviation distance value by combining the positional deviation distance value with the age baseline distance range.

[0131] Among them, the abnormal deviation distance value refers to the deviation distance value between the positional deviation distance value and the age benchmark distance range.

[0132] When the position deviation distance value is not within the age benchmark distance range, it indicates that the player's login position is abnormal. Therefore, the difference between the position deviation distance value and the nearest value in the age benchmark distance range is calculated, and the calculation result is used as the abnormal deviation distance value for subsequent use.

[0133] S645236: Collect the current time point.

[0134] The current time point refers to the specific point in time corresponding to the current moment, which is obtained by querying the current time database. The time database keeps track of time and stores it in real time.

[0135] S645237: Determine the abnormal distance adjustment value by combining the current time point and the abnormal deviation distance value, and use the abnormal distance adjustment value as the position deviation adjustment value.

[0136] The abnormal distance adjustment value refers to the adjustment value corresponding to the adjustment made based on the abnormal deviation distance value.

[0137] By determining whether the current time point is within a holiday period, if it is within a holiday period, a preset holiday adjustment value is output and used as the abnormal distance adjustment value. If it is not within a holiday period, the product between the abnormal deviation distance value and the preset abnormal distance coefficient is calculated, and the result is used as the abnormal distance adjustment value. The abnormal distance adjustment value is then used as the position deviation adjustment value, thereby improving the accuracy of the obtained position deviation adjustment value.

[0138] Holiday adjustment values ​​refer to the baseline adjustment values ​​that allow for abnormal distance deviations during holidays. These holiday adjustment values ​​are preset by the operator based on actual needs.

[0139] The abnormal distance coefficient is a coefficient used to convert abnormal deviation distance values ​​into abnormal distance adjustment values. The abnormal distance coefficient is preset by the operator according to actual needs.

[0140] S64524: Determine the age base time corresponding to each demand collection type based on the account age value.

[0141] Among them, the age baseline time refers to the active time range under different data collection types based on the account's age.

[0142] By inputting the account's age value into a preset age-time database, an age baseline time can be obtained for convenient subsequent use.

[0143] The age-time database pre-stores different age ranges and their corresponding age reference times. The age-time database is preset by the operator according to the actual situation.

[0144] For example, if the age range is between 14 and 18 years old, the overall active time range is from 8 PM to 9 PM. At this time, data types such as mouse trajectory can be set to 8:05 PM to 8:55 PM, and resource acquisition speed and skill release interval can be set to 8:10 PM to 8:50 PM.

[0145] The age range is between 18 and 22 years old. The overall active time range is from 10:00 to 02:00 the next day. Data types such as mouse trajectory can be set to 10:00 to 02:00 the next day, and resource acquisition speed and skill release interval can be set to 10:05 to 1:55 the next day.

[0146] The age range is between 22 and 60 years old. The overall active time range is from 7 pm to 9 pm. Data types such as mouse trajectory can be set from 7:10 pm to 8:50 pm. Resource acquisition speed and skill release interval can be set from 10:15 am to 1:45 am the next day.

[0147] S64525: Retrieve friend activity time based on account friend information.

[0148] Among them, "friends' active time" refers to the time interval corresponding to the account's friends' activity. Account friend information includes friends' active time.

[0149] The account's friend information allows for the retrieval of friends' active time periods, facilitating subsequent use.

[0150] S64526: Generates an active adjustment time by combining the age base time with the active time of friends.

[0151] The active adjustment time refers to the time interval corresponding to the age base time after adjustment.

[0152] By combining and analyzing the age baseline time with the activity time of friends, an activity adjustment time is generated for convenient subsequent use.

[0153] To further ensure the rationality of the active adjustment time, it is necessary to perform a more detailed separate analysis and calculation of the active adjustment time, which will be explained in detail through the steps shown below.

[0154] The method for generating the active adjustment time includes the following steps: S645261: When there is an overlap between the age baseline time and the friend's active time, compare the age baseline time and the friend's active time to determine the active overlap time.

[0155] Among them, the active overlap time refers to the time interval when there is an overlap between the age baseline time and the active time of friends.

[0156] When the age baseline time overlaps with the friend's active time, it indicates that the player usually plays the game with their friends. Therefore, the age baseline time is compared with the friend's active time, and the overlapping time interval is taken as the active overlap time for future use.

[0157] S645262: Compare the active overlap time with the age baseline time to determine the overlap ratio.

[0158] The overlap ratio refers to the proportion of overlap time to the age baseline time.

[0159] By calculating the time span between the active overlap time and the age baseline time separately, and then calculating the ratio between the two time spans as the overlap ratio, it is convenient for subsequent use.

[0160] S645263: Determine the overlap adjustment value by referring to the overlap ratio value.

[0161] The overlap adjustment value refers to the adjustment value corresponding to the adjustment made based on the overlap ratio value.

[0162] The product of the overlap ratio and the preset overlap adjustment coefficient is calculated, and the result is used as the overlap adjustment value for convenient subsequent use.

[0163] The overlap adjustment factor is a coefficient used to convert the overlap ratio value into an overlap adjustment value. The overlap adjustment factor is preset by the operator according to actual needs.

[0164] S645264: Retrieve the number of overlapping values ​​based on the active overlapping time.

[0165] Here, the number of overlapping values ​​refers to the number of values ​​corresponding to the active overlapping time.

[0166] By counting the active overlap times and retrieving the count results as the number of overlaps, it is convenient to use them later.

[0167] S645265: Determine the number of influence values ​​by referring to the number of overlap values.

[0168] Among them, the number of impact values ​​refers to the adjustment value corresponding to the impact of the number of overlaps on the overlap adjustment value.

[0169] The product of the number of overlaps and the preset number influence coefficient is calculated, and the result is used as the number influence value for convenient subsequent use.

[0170] The number influence coefficient is a coefficient used to convert overlapping numerical values ​​into number influence values. The number influence coefficient is preset by the operator according to actual needs.

[0171] S645266: Determine the overlap correction value by combining the number of influence values ​​and the overlap adjustment value.

[0172] The overlap correction value refers to the adjustment value after adjusting the overlap adjustment value.

[0173] The product of the number of influence values ​​and the overlap adjustment value is calculated, and the result is used as the overlap correction value for convenient subsequent use.

[0174] S645267: Adjust the age base time using the overlap correction value to obtain the overlap adjustment time, and use the overlap adjustment time as the active adjustment time.

[0175] The overlap adjustment time refers to the time interval corresponding to the adjustment of the age base time.

[0176] By converting the overlap correction value into a time value, the two time endpoints of the age base time are respectively added with the time value corresponding to the overlap correction value to form new time interval endpoints, thereby obtaining the overlap adjustment time. The overlap adjustment time is then used as the active adjustment time, improving the accuracy of the obtained active adjustment time.

[0177] For example, when the overlap correction value is 0.2, the corresponding time value is 12 minutes. When the age base time is from 8:10 PM to 8:50 PM, the overlap adjustment time is from 8:22 PM to 9:02 PM.

[0178] S64527: Combine the position deviation adjustment value with the active adjustment time to generate the position adjustment time, and use the position adjustment time as the reference active time.

[0179] Among them, the position adjustment time refers to the time interval corresponding to the adjustment of the active adjustment time based on the position deviation adjustment value.

[0180] By combining the position deviation adjustment value with the active adjustment time, a position adjustment time is generated and used as the benchmark active time, thereby improving the accuracy of the obtained benchmark active time.

[0181] To further ensure the rationality of the location adjustment time, it is necessary to conduct a more detailed analysis and calculation of the location adjustment time separately, which will be explained in detail through the steps shown below.

[0182] The method for generating the position adjustment time includes the following steps: S645271: Retrieve the start and end times of the adjustment based on the active adjustment time.

[0183] The adjustment start time refers to the beginning of the active adjustment period. The adjustment end time refers to the end of the active adjustment period. The active adjustment period includes both the adjustment start time and the adjustment end time.

[0184] The start and end times of the adjustment can be retrieved by actively adjusting the time for convenient subsequent use.

[0185] S645272: Compare the start time point and the end time point of the adjustment to determine the time span value.

[0186] The time span value refers to the time elapsed between the start time of the adjustment and the end time of the adjustment.

[0187] The time elapsed between the start and end times of the adjustment is calculated, and the result is used as the time span value for convenient subsequent use.

[0188] S645273: Determine the adjustment benchmark interval by referring to the time span value.

[0189] The adjustment reference interval refers to the reference interval corresponding to the position deviation adjustment value when the interval needs to be narrowed based on the time span value.

[0190] Different time span values ​​correspond to different adjustment benchmark intervals. By inputting the time span values ​​into a preset time span database, adjustment benchmark intervals are obtained for convenient subsequent use.

[0191] The time span database pre-stores a table of different time span values ​​and their corresponding adjustment benchmark intervals. The time span database is preset by the operator according to the actual situation.

[0192] For example, when the time span is 2 hours, the adjustment base range can be set to 0 to 1; when the time span is 3 hours, the adjustment base range can be set to 0 to 0.7; and when the time span is 4 hours, the adjustment base range can be set to 0 to 0.4.

[0193] S645274: Determine whether the position deviation adjustment value falls within the adjustment reference range. If yes, execute S645275; if no, execute S645276.

[0194] Specifically, by judging whether the position deviation adjustment value falls within the adjustment benchmark range, it is determined whether the time interval needs to be narrowed.

[0195] S645275: Use the position deviation adjustment value to adjust the start and end times of the adjustment in the same direction to form the position adjustment time.

[0196] When the position deviation adjustment value falls within the adjustment reference range, it means that there is no need to narrow the time interval. Therefore, the position deviation adjustment value is used to increase the adjustment start time point and adjustment end time point in the same direction to form the position adjustment time, thereby improving the accuracy of the obtained position adjustment time.

[0197] S645276: Use the position deviation adjustment value to perform convergent adjustments on the adjustment start time point and adjustment end time point to form the position adjustment time.

[0198] When the position deviation adjustment value does not fall within the adjustment reference range, it indicates that the time interval needs to be narrowed. Therefore, the position deviation adjustment value is used to increase the adjustment start time point and decrease the adjustment end time point, thereby narrowing the time interval. The narrowed time interval is then used as the position adjustment time, thus improving the accuracy of the obtained position adjustment time.

[0199] S6453: Compare the baseline active time with the account active time to determine the active deviation time value.

[0200] Among them, the active deviation time value refers to the time value corresponding to when there is a deviation in account activity.

[0201] By comparing the baseline active time with the account active time, and calculating the span of the time interval corresponding to the baseline active time that the account active time does not cover, the calculation result is used as the active deviation time value for convenient subsequent use.

[0202] S6454: Determine the active deviation adjustment value by referring to the active deviation time value.

[0203] The active deviation adjustment value refers to the adjustment value required when adjustments are needed based on the active deviation time value.

[0204] The active deviation time value is calculated by multiplying it with a preset deviation time coefficient, and the result is used as the active deviation adjustment value for easy subsequent use.

[0205] The deviation time coefficient is a coefficient used to convert the active deviation time value into the active deviation adjustment value. The deviation time coefficient is preset by the operator according to actual needs.

[0206] S6455: Adjust and update the actual type weights using the active bias adjustment value.

[0207] Specifically, the product between the active bias adjustment value and the actual type weight is calculated, and the result is used as the new actual type weight to update and replace the actual type weight, thereby improving the accuracy of the obtained actual type weight.

[0208] S65: The single-type probability and the type influence weight are weighted to obtain the comprehensive type probability, and the comprehensive type probability is used as the suspected violation probability.

[0209] Among them, the comprehensive type probability refers to the probability value after comprehensively calculating the probabilities of each type.

[0210] By weighting the single-type probability with the type influence weight, a comprehensive type probability is obtained, and this comprehensive type probability is used as the suspected violation probability, thereby improving the accuracy of the obtained suspected violation probability.

[0211] S7: Based on the probability of suspected violation falling within the preset range of violation handling probability, generate violation handling information and output the violation handling information.

[0212] The violation processing probability range refers to the probability range at which a violation needs to be processed. This range is preset by the operator based on the actual situation. Violation processing information refers to the information used to process violations.

[0213] By analyzing the probability of suspected violations falling within a preset range of violation handling probabilities, and outputting the handling information corresponding to the violation handling probability range where the suspected violation probability falls, a precise judgment can be made based on the user's situation, thereby detecting game violations in a timely and accurate manner.

[0214] For example, the probability range for handling violations can be set as follows: low risk range 0 to 0.3, medium risk range 0.3 to 0.7, and high risk range 0.7 to 1.0. Violation handling information corresponding to the low risk range includes pop-up warnings and recording violation logs; violation handling information corresponding to the medium risk range includes disabling certain functions or ranked matches; and violation handling information corresponding to the high risk range includes temporary account bans.

[0215] Based on the same inventive concept, embodiments of the present invention provide an AI-based game violation detection and identification system, comprising: The data collection module is used to collect information such as game type, game account, actual data collected, current account location, and current time. The memory stores a program for implementing an AI-based method for detecting and recognizing game violations, as described above. The processor loads and executes programs stored in memory.

[0216] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0217] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting and identifying game violations based on AI, characterized in that, include: S1: Collect game type and game account; S2: Determine the type of data collection required based on the game type; S3: Collect data according to the aforementioned data collection type to obtain actual collected data; S4: Retrieve historical game data based on the aforementioned game account; S5: Input the historical game data into a preset adaptive AI model cluster for training to generate an account violation detection model cluster; S6: Combine the actual collected data with the account violation detection model cluster to generate a suspected violation probability; S7: Based on the occurrence of the suspected violation probability and the preset violation handling probability range, generate violation handling information and output the violation handling information.

2. The AI-based method for detecting and identifying game violations according to claim 1, characterized in that, The method for generating the suspected violation probability includes: S61: Retrieve the actual collection type and single-type data based on the actual collected data; S62: Select a single-type detection model from the account violation detection model cluster based on the actual collection type; S63: Input the single-type data into the single-type detection model to obtain the single-type probability; S64: Combine the actual data collection type with the required data collection type to generate a type influence weight; S65: The single-type probability and the type influence weight are weighted and calculated to obtain the comprehensive type probability, and the comprehensive type probability is used as the suspected violation probability.

3. The AI-based method for detecting and identifying game violations according to claim 2, characterized in that, The methods for generating the type-affected weights include: S641: Compare the actual collection type with the required collection type to obtain the actual overlapping type and the actual missing type; S642: Generate a type adjustment value by combining the actual overlapping type and the actual missing type; S643: Retrieve the type importance value corresponding to each of the aforementioned requirement collection types based on the game type; S644: Determine the type importance weight with reference to the type importance value; S645: Adjust the type importance weight with the type adjustment value to obtain the actual type weight, and use the actual type weight as the type influence weight.

4. The AI-based method for detecting and identifying game violations according to claim 3, characterized in that, The method for generating the type adjustment value includes: S6421: Retrieve the numerical values ​​of the overlapping types based on the actual overlapping types; S6422: Retrieve the number of missing values ​​based on the actual missing type; S6423: Calculate the ratio of the number of overlapping types to the number of missing types to obtain the overlap ratio; S6424: Determine the overlap ratio adjustment value with reference to the aforementioned overlap ratio value; S6425: Adjust the preset overlap reference adjustment value using the overlap ratio adjustment value to form an overlap adjustment value; S6426: Adjust the preset missing baseline adjustment value using the overlap ratio adjustment value to form a missing adjustment value; S6427: Combine the overlap adjustment value and the missing adjustment value as the type adjustment value.

5. The AI-based method for detecting and identifying game violations according to claim 3, characterized in that, After adjusting the type importance weights using the type adjustment value to obtain the actual type weights, the following steps are also included: S6451: Retrieve basic account information and account activity time based on the aforementioned game account; S6452: Determine the baseline active time corresponding to each of the aforementioned demand collection types based on the account baseline information; S6453: Compare the baseline active time with the account active time to determine the active deviation time value; S6454: Determine the active deviation adjustment value with reference to the active deviation time value; S6455: Adjust and update the actual type weights using the active deviation adjustment value.

6. The AI-based method for detecting and identifying game violations according to claim 5, characterized in that, The method for determining the baseline active time includes: S64521: Retrieve the account age value, account historical location points, and account friend information based on the aforementioned account baseline information; S64522: Collect the current location of the account; S64523: Generate a position deviation adjustment value by combining the current location of the account with the historical location of the account; S64524: Determine the age baseline time corresponding to each of the aforementioned demand collection types based on the account age value; S64525: Retrieve the active time of friends based on the account's friend information; S64526: Generate an active adjustment time by combining the age baseline time with the friend's active time; S64527: Combine the position deviation adjustment value with the active adjustment time to generate a position adjustment time, and use the position adjustment time as the reference active time.

7. The AI-based method for detecting and identifying game violations according to claim 6, characterized in that, The method for generating the position deviation adjustment value includes: S645231: Compare the current location of the account with the historical location of the account to determine the location deviation distance value; S645232: Determine the age baseline distance range based on the account age value; S645233: Determine whether the positional deviation distance value is within the age reference distance range; S645234: If yes, then determine the distance movement adjustment value based on the position deviation distance value, and use the distance movement adjustment value as the position deviation adjustment value; S645235: If not, then determine the abnormal deviation distance value by combining the position deviation distance value with the age reference distance range; S645236: Collect the current time point; S645237: Determine the abnormal distance adjustment value by combining the current time point with the abnormal deviation distance value, and use the abnormal distance adjustment value as the position deviation adjustment value.

8. The AI-based method for detecting and identifying game violations according to claim 6, characterized in that, The method for generating the active adjustment time includes: S645261: When the age reference time overlaps with the friend's active time, compare the age reference time with the friend's active time to determine the active overlap time; S645262: Compare the active overlap time with the age baseline time to determine the overlap ratio; S645263: Determine the overlap adjustment value with reference to the aforementioned overlap ratio value; S645264: Retrieve the number of overlapping values ​​based on the active overlap time; S645265: Determine the number of influence values ​​by referring to the aforementioned overlap value; S645266: Determine the overlap correction value by combining the number influence value and the overlap adjustment value; S645267: Adjust the age reference time using the overlap correction value to obtain the overlap adjustment time, and use the overlap adjustment time as the active adjustment time.

9. The AI-based method for detecting and identifying game violations according to claim 6, characterized in that, The method for generating the position adjustment time includes: S645271: Retrieve the adjustment start time and adjustment end time based on the active adjustment time; S645272: Compare the adjustment start time point with the adjustment end time point to determine the time span value; S645273: Determine the adjustment benchmark interval with reference to the aforementioned time span value; S645274: Determine whether the position deviation adjustment value falls within the adjustment reference range; S645275: If yes, then the position deviation adjustment value is used to adjust the adjustment start time point and the adjustment end time point in the same direction to form the position adjustment time; S645276: If not, then the position deviation adjustment value is used to perform convergent adjustment on the adjustment start time point and the adjustment end time point to form the position adjustment time.

10. An AI-based system for detecting and identifying game violations, characterized in that, include: The data collection module is used to collect information such as game type, game account, actual data collected, current account location, and current time. The memory stores a program for implementing an AI-based method for detecting and identifying game violations as described in any one of claims 1 to 9; The processor loads and executes programs stored in memory.

Images