Identity verification method, device and equipment

Abstract

The embodiment of the invention discloses an identity verification method, device and equipment, and the method comprises the steps: obtaining the historical dialogue data of a user in human-computer interaction and supplementary verification information for the user when the identity verification of the user is detected; inputting the historical dialogue data and the supplementary verification information into a large language model to generate interaction problem data, wherein the interaction problem data can guide a user to provide dialogue content matched with a dialogue style of the user and is embedded with information related to the supplementary verification information; obtaining interaction answer data of the user, and if it is judged that supplementary verification processing needs to be carried out on the user again, carrying out man-machine interaction with the user again through the large language model until the user identity supplementary verification result obtained based on the man-machine interaction data is used for judging that the accuracy reached by the user identity exceeds a preset threshold value; and determining whether the identity verification of the user is passed based on the identity verification result and the identity supplementary verification result.

Description

Technical Field

[0001] This document relates to the field of computer technology, and particularly to an identity verification method, device, and equipment. Background Art

[0002] With the support of large model capabilities, the effect of multi-round conversations has been continuously improved and is currently applied to various fields, such as intelligent customer service, information recommendation, image processing, identity verification, etc. At the same time, the booming development of AIGC (Artificial Intelligence Generated Content) has also enabled the AIGC capabilities to be exploited by illegal elements. How to defend against AIGC attacks and protect users' private data has become an important issue in various business scenarios. For identity verification, it is usually to classify the video, audio, text data, etc. collected in the identity verification scenario by AIGC for identification. The above method is easily bypassed, resulting in identity verification being attacked and users' private data being leaked. Therefore, a better identity verification method is needed. Summary of the Invention

[0003] The purpose of the embodiments of this specification is to provide a better identity verification method.

[0004] To achieve the above technical solution, the embodiments of this specification are implemented as follows: An identity verification method provided by the embodiments of this specification, the method includes: when detecting that a user performs identity verification, obtaining the historical conversation data of the user in the human-computer interaction and supplementary verification information for the user, where the supplementary verification information includes information related to the attributes and / or user behaviors of the user; inputting the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information, where the interactive question data can guide the user to provide conversation content matching the user's conversation style and is embedded with information related to the supplementary verification information; obtaining the interactive answer data provided by the user for the interactive question data, if it is determined based on the interactive question data and the interactive answer data that the user needs to be subjected to supplementary verification processing again, then based on the historical conversation data of the user in the human-computer interaction and the supplementary verification information for the user, performing human-computer interaction with the user again through the large language model until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity reaches more than a preset threshold; obtaining the identity verification result of the user's identity verification, and determining whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result.

[0005] An identity verification device provided by an embodiment of this specification, the device includes: a data acquisition module, when detecting that a user performs identity verification, acquires the historical conversation data of the user in human-computer interaction and supplementary verification information for the user, where the supplementary verification information includes information related to the attributes and / or user behavior of the user; an interaction question generation module, inputs the historical conversation data and the supplementary verification information into a large language model, to guide the large language model to generate interaction question data through the historical conversation data and the supplementary verification information, the interaction question data can guide the user to provide conversation content matching the user's conversation style and is embedded with information related to the supplementary verification information; a supplementary verification module, acquires the interaction answer data provided by the user for the interaction question data, if it is determined based on the interaction question data and the interaction answer data that the user needs to be subjected to supplementary verification processing again, then based on the historical conversation data of the user in human-computer interaction and the supplementary verification information for the user, performs human-computer interaction with the user again through the large language model until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity reaches more than a preset threshold; an identity verification module, acquires the identity verification result of the user's identity verification, and determines whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result.

[0006] An identity verification device provided by an embodiment of this specification, the identity verification device includes: a processor; and a memory arranged to store computer-executable instructions, the executable instructions, when executed, cause the processor: when detecting that a user performs identity verification, obtain historical conversation data of the user in human-computer interaction and supplementary verification information for the user, the supplementary verification information includes information related to the attributes and / or user behavior of the user; input the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information, the interactive question data can guide the user to provide conversation content matching the user's conversation style and is embedded with information related to the supplementary verification information; obtain interactive answer data provided by the user for the interactive question data, if it is determined based on the interactive question data and the interactive answer data that it is necessary to perform supplementary verification processing on the user again, then based on the historical conversation data of the user in human-computer interaction and the supplementary verification information for the user, perform human-computer interaction with the user again through the large language model until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity exceeds a preset threshold; obtain the identity verification result of the user's identity verification, and determine whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result.

[0007] The embodiments of this specification also provide a storage medium for storing computer-executable instructions. When the executable instructions are executed by a processor, the following processes are implemented: When it is detected that a user performs identity verification, obtain the user's historical conversation data in human-computer interaction and supplementary verification information for the user, where the supplementary verification information includes information related to the user's attributes and / or user behavior; input the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interaction problem data through the historical conversation data and the supplementary verification information. The interaction problem data can guide the user to provide conversation content matching the user's conversation style and is embedded with information related to the supplementary verification information; obtain the interaction answer data provided by the user for the interaction problem data. If it is determined based on the interaction problem data and the interaction answer data that supplementary verification processing needs to be performed on the user again, then based on the user's historical conversation data in human-computer interaction and the supplementary verification information for the user, conduct human-computer interaction with the user again through the large language model until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity exceeds a preset threshold; obtain the identity verification result of the user's identity verification, and determine whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result.

[0008] The embodiments of this specification also provide a computer program product, including a computer program, which when executed by a processor implements the following processes: when it is detected that a user performs identity verification, obtain the historical conversation data of the user in human-computer interaction and supplementary verification information for the user, where the supplementary verification information includes information related to the attributes and / or user behavior of the user; input the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information, and the interactive question data can guide the user to provide conversation content matching the user's conversation style and embed information related to the supplementary verification information; obtain the interactive answer data provided by the user for the interactive question data, if it is determined based on the interactive question data and the interactive answer data that the user needs to be subjected to supplementary verification processing again, then based on the historical conversation data of the user in human-computer interaction and the supplementary verification information for the user, perform human-computer interaction with the user again through the large language model until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity exceeds a preset threshold; obtain the identity verification result of the user's identity verification, and determine whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result. Brief Description of the Drawings

[0009] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings; Figure 1 It is a schematic flowchart of an identity verification method in this specification; Figure 2 It is a schematic diagram of an identity verification process in this specification; Figure 3 It is a schematic diagram of a model training process in this specification; Figure 4 It is a schematic diagram of an identity verification process based on video in this specification; Figure 5 It is a schematic diagram of an identity verification device in this specification; Figure 6 It is a schematic diagram of an identity verification device in this specification. Detailed Embodiments

[0010] The embodiments of this specification provide an identity verification method, device and equipment.

[0011] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0012] The embodiments of this specification provide an identity verification mechanism. With the support of large model capabilities, the effect of multiple rounds of dialogues is constantly improving, and it has been applied to various fields, such as intelligent customer service, information recommendation, image processing, identity verification, etc. At the same time, the vigorous development of AIGC has also caused AIGC capabilities to be exploited by illegal persons. How to defend against AIGC attacks has become an important issue in many business scenarios. For identity verification, AIGC classification is usually performed on the video, audio, text data, etc. collected in the identity verification scenario for identification. The above method can be easily bypassed, resulting in identity verification attacks and user data leakage. To this end, it is necessary to provide a better identity verification method. The embodiment of this specification provides an identity verification method. This identity verification method takes into account that identity verification is a long process that requires in-depth communication with the user. Therefore, the user's conversation style recognition and KBA (Knowledge-Based Authentication) problems are considered to be applied to the identity verification process. However, the blunt insertion of questions does not bring a good user experience. Based on this, the embodiment of this specification provides an identity verification method based on large model dialogue guidance, embedded user conversation style judgment and implicit KBA questions and answers, so as to enhance the defense capability of the identity verification scenario against AIGC, that is, embedding user conversation style judgment and implicit KBA questions and answers in the identity verification scenario to supplement the user's identity. In this way, in identity verification, by guiding the user to reply with more stylized content, while including knowledge information related to the user, implicit verification is naturally completed, thereby improving the security of identity verification. For specific processing, please refer to the specific content in the following embodiments.

[0013] like Figure 1As shown in the figure, an embodiment of this specification provides an identity verification method. The execution subject of this method can be a terminal device or a server, etc. The terminal device can be a mobile terminal device such as a mobile phone or a tablet computer, or a computer device such as a laptop computer or a desktop computer, or it can also be an IoT device (specifically, such as a smart watch, a vehicle-mounted device, etc.). The server can be an independent server or a server cluster composed of multiple servers. The server can be a background server for financial services or online shopping services, etc., or it can be a background server for a certain application program, etc. In this embodiment, the server is used as an example of the execution subject for detailed description. For the case where the execution subject is a terminal device, refer to the following handling of the server case, which will not be elaborated here. The method can specifically include the following steps: In step S102, when it is detected that the user is performing identity verification, obtain the historical conversation data of the user in the human-computer interaction and the supplementary verification information for the user. The supplementary verification information includes information related to the user's attributes and / or user behavior.

[0014] Among them, the user can be any user who needs to perform identity verification. Human-computer interaction is a process in which a person interacts and converses with an intelligent agent (such as a device equipped with a specified network model, specifically, a mobile phone, a device, etc. equipped with a large language model). The historical conversation data can be the data of the conversation generated by the interaction between the person and the intelligent agent in the human-computer interaction within a certain time period. The supplementary verification information can be information related to supplementing the identity verification of the above user. The supplementary verification information can include information related to the user's attributes and / or user behavior. The user's attributes can include various types. For example, the user's commonly used transaction address, transaction time, user preferences, place of birth, place of work, user identification, etc. can be specifically set according to the actual situation. User behavior can include various types. For example, the user's transaction behavior, query behavior, information recording behavior, etc. can be specifically set according to the actual situation. The supplementary verification information can include various types. The supplementary verification information can include the Q&A information in the KBA Q&A knowledge base. Based on this, the supplementary verification information can be presented in the form of questions and answers. For example, question: Which cities has the user visited recently (such as within the last 3 days or within the last week, etc.)? Answer: Beijing and Tianjin. Another example, question: Which financial management services has the user handled recently (such as within the last 3 days or within the last week, etc.)? Answer: Insurance services. Among them, the above questions and answers can be determined in various different ways. For example, the corresponding transaction location can be obtained through the user's transaction behavior, and the relevant information about the cities the user has visited can be determined through the transaction location. The relevant information about the financial management services handled can be determined through the user's transaction type, etc. It can be specifically set according to the actual situation.

[0015] In implementation, when a user directly triggers identity verification or needs to undergo identity verification during the execution of a certain business, etc., the user can, before the identity verification, or during the identity verification process, or after the identity verification is completed, continue to conduct supplementary verification on the user to improve the security of the user's identity verification. Specifically, it is possible to obtain the conversation data formed during the interaction between the user and the intelligent agent within a specified time period (such as within 3 days or within a month before the current moment, etc.). The above-obtained conversation data can be used as the historical conversation data of the user in human-computer interaction. For the supplementary verification information of the user, it can be constructed in a variety of different ways. For example, by summarizing the user's behavior within a specified time period, data of questions and corresponding answers can be formed. For another example, the user's attribute information can be obtained, and the supplementary verification information can be constructed based on the processing results obtained through combination, integration, etc. of this attribute information. Specifically, it can be set according to the actual situation. The supplementary verification information of the user can be obtained through the above methods.

[0016] In step S104, the historical conversation data and the supplementary verification information are input into the large language model to guide the large language model to generate interaction question data through the historical conversation data and the supplementary verification information. The interaction question data can guide the user to provide conversation content that matches the user's conversation style and embeds information related to the supplementary verification information.

[0017] Among them, the conversation style can include various types. For example, humorous, concise, accustomed to using dialects, euphemistic, etc. Specifically, it can be set according to the actual situation. The large language model can be a natural language processing model composed of hundreds of millions of parameters, based on deep learning algorithms, and pre-trained using a large-scale corpus for text understanding, generation, and transformation, etc. The large language model can include GPT-3, ERNIE Bot, etc. Specifically, it can be set according to the actual situation.

[0018] In implementation, a corresponding algorithm can be obtained, and a large language model can be constructed based on this algorithm. For example, a corresponding large language model can be constructed based on the Transfomer architecture, etc. The input data of this large language model can be prompt information, historical conversation data, and supplementary verification information, etc. The output data can be the interaction question data used to interact with the user. Then, training samples for training the large language model (that is, the historical conversation data in the human-computer interaction of different users and the supplementary verification information of different users or the prompt information constructed from the above information, etc.) can be obtained, and this training sample can be used to train the corresponding large language model. During the training process, a target function can be preset, and the model parameters in the large language model can be optimized based on this target function to finally obtain the trained large language model.

[0019] The corresponding prompt information Prompt can be constructed using historical conversation data and supplementary verification information. Then, the prompt information Prompt can be input into the trained large language model. In addition, the historical conversation data and supplementary verification information can also be input into the trained large language model. The large language model can use the prompt information Prompt or the historical conversation data and supplementary verification information as guiding information to guide the large language model to generate interactive question data. The generated interactive question data can guide the user to provide conversation content that matches the user's conversation style and implicitly embeds information related to the supplementary verification information. In this way, in the identity verification scenario, the large language model embeds the user's conversation style judgment and implicit KBA question and answer information (i.e., supplementary verification information) to supplement the verification of the user's identity, thereby enhancing the security of identity verification.

[0020] In step S106, the interactive answer data provided by the user for the interactive question data is obtained. If it is determined based on the interactive question data and the interactive answer data that the user needs to be supplemented with verification again, then based on the historical conversation data of the user in the human-computer interaction and the supplementary verification information for the user, the large language model is used to conduct a human-computer interaction with the user again until the accuracy of the identity supplementary verification result for the user obtained based on the human-computer interaction data reaches a preset threshold.

[0021] Among them, the preset threshold can be set according to the actual situation. Specifically, for example, the preset threshold can be 80% or 90%, etc. The human-computer interaction data can be composed of interactive question data and interactive answer data.

[0022] In implementation, after generating interaction problem data in the above manner, the interaction problem data can be provided to the user (it can be presented to the user in text form, or read aloud to the user by voice, etc.). The user can give corresponding responses to the interaction problem data, thereby obtaining the user's interaction answer data. The user's interaction answer data can be analyzed to determine the matching degree between the user's conversation style presented in the user's interaction answer data and the user's true conversation style. At the same time, the matching degree between the user's interaction answer data and the answers in the supplementary verification information can also be determined. If the user's interaction answer data does not fully match the answers in the supplementary verification information (for example, there are 2 pieces of information in the answers of the supplementary verification information, but only 1 piece of information in the user's interaction answer data. Specifically, for example, in the supplementary verification information, the question is: Which cities has the user visited recently (such as within the last 3 days or within the last week, etc.)? The answer is: Beijing and Tianjin, and the user's interaction answer data only contains Tianjin), and the matching degree between the user's conversation style presented in the user's interaction answer data and the user's true conversation style does not exceed the set value, then it can be determined that the user needs to be processed for supplementary verification again. In practical applications, in addition to judging whether to process the user for supplementary verification again in the above manner, it can also be achieved through other methods. For example, the number of supplementary verifications can be preset, specifically such as 3 times or 2 times, etc. In addition, the number of supplementary verifications can also be set according to the user's accumulated credit situation, or the number of supplementary verifications can be set according to the user's attributes, etc. Specifically, it can be set according to the actual situation. If it is determined based on the interaction problem data and the interaction answer data that the user needs to be processed for supplementary verification again, then based on the user's historical conversation data in the human-computer interaction and the supplementary verification information for the user, the large language model is used to interact with the user again, that is, the processing of steps S102 to S106 is repeated. Each time, the interaction answer data provided by the user for the interaction problem data can be obtained. Based on the obtained interaction problem data and interaction answer data (i.e., the human-computer interaction data), it can be judged whether the accuracy of determining the user's identity exceeds the preset threshold. If the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of determining the user's identity exceeds the preset threshold, then the finally obtained identity supplementary verification result can be obtained.

[0023] In step S108, obtain the identity verification result of the user for identity verification, and determine whether the user's identity verification is passed based on the identity verification result and the identity supplementary verification result.

[0024] In implementation, it is also possible to obtain the identity verification result of the user for identity verification. For example, it is possible to obtain the identity verification result of identity verification through facial recognition (or video verification), or it is possible to obtain the identity verification result of identity verification through fingerprint recognition, etc. Then, the identity verification result and the identity supplementary verification result can be combined to comprehensively verify the identity of the user to determine whether the user's identity verification passes. Specifically, if the identity verification result is passed and the identity supplementary verification result is passed, it can be determined that the user's final identity verification result is passed; otherwise, it can be determined that the user's final identity verification result is not passed. If it is determined that the user's final identity verification result is passed, subsequent business processing can be continued. If it is determined that the user's final identity verification result is not passed, the user can be notified that the current identity verification has not passed, and the user can be reminded to re-perform identity verification processing.

[0025] An embodiment of this specification provides an identity verification method. When it is detected that a user performs identity verification, historical conversation data of the user in human-computer interaction and supplementary verification information for the user are obtained. The supplementary verification information includes information related to the attributes and / or behaviors of the user. Then, the historical conversation data and the supplementary verification information are input into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information. The interactive question data can guide the user to provide conversation content matching the user's conversation style and embed information related to the supplementary verification information. After that, the interactive answer data provided by the user for the interactive question data can be obtained. If it is determined based on the interactive question data and the interactive answer data that the user needs to be re-verified supplementary, then based on the historical conversation data of the user in human-computer interaction and the supplementary verification information for the user, the large language model is used to perform human-computer interaction with the user again until the accuracy of the identity supplementary verification result for the user obtained based on the human-computer interaction data reaches a preset threshold. Finally, the identity verification result of the user for identity verification can be obtained, and based on the identity verification result and the identity supplementary verification result, it can be determined whether the user's identity verification passes. In this way, an identity verification method that embeds user conversation style judgment and implicit KBA question and answer (i.e., supplementary verification information for the user) is used to enhance the defense ability of the identity verification scenario against AIGC. That is, in the identity verification scenario, user conversation style judgment and implicit KBA question and answer are embedded to supplement the verification of the user's identity. In this way, in identity verification, by guiding the user to reply with more stylized content and including knowledge information related to the user, implicit verification is naturally completed, thereby improving the security of identity verification. Moreover, implicit questions can be asked based on knowledge information such as the user's supplementary verification information to ensure the user experience during the identity verification process.

[0026] In practical applications, the supplementary authentication result for a user's identity can be determined in various ways. The following provides an optional processing method, which specifically includes the processing steps from Step A2 to Step A6.

[0027] In Step A2, obtain the interaction answer data in the human-computer interaction data obtained each time.

[0028] In implementation, after generating the interaction question data each time in the above manner, the interaction question data can be provided to the user, and the user can give corresponding answers to each interaction question data, so as to determine the interaction answer data in the human-computer interaction data obtained each time.

[0029] In Step A4, based on the interaction answer data obtained each time in the human-computer interaction, perform supplementary authentication processing on the user respectively through preset verification rules, and determine the supplementary authentication result corresponding to the interaction answer data obtained each time. The verification rules include the rule for judging whether the interaction answer data is consistent with the user's supplementary authentication information and the matching verification rule for the dialogue style.

[0030] Among them, the matching verification rule for the dialogue style can be a rule for matching the dialogue style presented by the user in the human-computer interaction this time with the user's true (or actual) dialogue style. The user's true (or actual) dialogue style can be determined in various different ways. For example, the user's true (or actual) dialogue style can be extracted from the user's historical dialogue data (which can be the historical dialogue data in the human-computer interaction or the historical dialogue data of the user chatting with other users, etc.), or it can also be relevant information about the user's own true (or actual) dialogue style provided by the user. It can be specifically set according to the actual situation.

[0031] In implementation, the interaction answer data obtained each time in the human-computer interaction can be analyzed, the information corresponding to the questions and answers in the user's supplementary authentication information can be extracted, and it can be determined whether the extracted information is consistent with the user's supplementary authentication information. In addition, the user's dialogue style can be extracted from the interaction answer data, and the extracted dialogue style can be matched with the user's true (or actual) dialogue style. Through the above consistency verification and matching processing, the supplementary authentication result corresponding to the interaction answer data obtained each time in the human-computer interaction can be obtained.

[0032] In Step A6, based on the supplementary authentication result corresponding to the interaction answer data obtained each time in the human-computer interaction, determine the supplementary authentication result for the user's identity.

[0033] In implementation, a comprehensive judgment can be made by combining the supplementary authentication results corresponding to the interaction answer data obtained each time in the human-computer interaction, and finally the supplementary authentication result for the user's identity can be obtained.

[0034] In practical applications, the matching and verification rules for the dialogue style include rules for determining whether the first dialogue style corresponding to the interactive answer data and the historical dialogue data matches and / or rules for determining whether the second dialogue style corresponding to the interactive answer data and the preset generated dialogue data matches. The generated dialogue data is the dialogue data generated by a specified network model.

[0035] Among them, the network model can include a neural network model, a large language model, etc., and can be specifically set according to the actual situation.

[0036] In practical applications, the verification rules include rules for determining whether the interactive answer data is consistent with the user's supplementary verification information. Based on this, the specific processing method of the above step A4 can be various. Hereinafter, an optional processing method is provided, which can specifically include the processing of the following step A402 and step A404.

[0037] In step A402, the interactive answer data obtained from each human-machine interaction and the user's supplementary verification information are input into a pre-trained verification large model to determine whether the interactive answer data is consistent with the user's supplementary verification information through the verification large model, and a first judgment result corresponding to the interactive answer data obtained from each human-machine interaction is obtained.

[0038] Among them, the verification large model can be composed of hundreds of millions of parameters, based on deep learning algorithms, and pre-trained using a large-scale corpus to verify whether the interactive answer data is consistent with the user's supplementary verification information. The verification large model can be constructed through GPT-3, ERNIE Bot, etc., and can be specifically set according to the actual situation.

[0039] In implementation, the corresponding algorithm can be obtained, and the verification large model can be constructed based on this algorithm. For example, a corresponding verification large model can be constructed based on the Transfomer architecture. The input data of the verification large model can be the interactive answer data obtained from human-machine interaction, the user's supplementary verification information, etc., and the output data can be whether the interactive answer data is consistent with the user's supplementary verification information or can also be the score of the consistency between the interactive answer data and the user's supplementary verification information, etc. Then, the training samples for training the verification large model (i.e., the interactive answer data obtained from human-machine interaction, the user's supplementary verification information, etc.) can be obtained, and the corresponding verification large model can be trained using the training samples. During the training process, the objective function can be preset, and the model parameters in the verification large model can be optimized based on the objective function, and finally the trained verification large model can be obtained.

[0040] Such as Figure 2As shown, the corresponding prompt information Prompt can be constructed using the interactive answer data obtained from human-computer interaction and the user's supplementary verification information. Then, the prompt information Prompt can be input into the trained verification large model. In addition, the interactive answer data obtained from human-computer interaction and the user's supplementary verification information can also be input into the trained verification large model. The verification large model can use the prompt information Prompt or the interactive answer data obtained from human-computer interaction and the user's supplementary verification information as guiding information to guide the verification large model to determine whether the interactive answer data is consistent with the user's supplementary verification information, and a consistent or inconsistent result or a score of their consistency (such as 0.5 or 0.8, etc.) can be obtained, thereby obtaining a first judgment result corresponding to the interactive answer data obtained from each human-computer interaction.

[0041] In step A404, based on the first judgment result, determine the supplementary verification result corresponding to the interactive answer data obtained from each human-computer interaction.

[0042] In implementation, the first judgment result can be used as the supplementary verification result corresponding to the interactive answer data obtained from this human-computer interaction. Or, the first judgment result can also be combined with other relevant information to finally determine the supplementary verification result corresponding to the interactive answer data obtained from this human-computer interaction, etc., which can be specifically set according to the actual situation.

[0043] In practical applications, the verification rules include rules for judging whether the first dialogue style corresponding to the interactive answer data matches the historical dialogue data. Based on this, the specific processing method of the above step A4 can be various. Hereinafter, another optional processing method is provided, such as Figure 2 As shown, it can specifically include the processing of the following steps A406 and A410.

[0044] In step A406, extract the dialogue style features from the interactive answer data obtained from each human-computer interaction and the historical dialogue data respectively, to obtain the first dialogue style feature corresponding to the interactive answer data obtained from each human-computer interaction and the second dialogue style feature corresponding to the historical dialogue data.

[0045] Among them, the extraction of the dialogue style features from the interactive answer data obtained from each human-computer interaction and the historical dialogue data can be implemented through a specified feature extraction algorithm, and can also be implemented through a network model (such as a convolutional neural network model or a recurrent neural network model, etc.), etc., which can be specifically set according to the actual situation.

[0046] In step A408, based on the similarity between the first dialogue style feature and the second dialogue style feature, judge whether the first dialogue style corresponding to the interactive answer data matches the historical dialogue data, and obtain a second judgment result corresponding to the interactive answer data obtained from each human-computer interaction.

[0047] Among them, the similarity between the first dialogue style feature and the second dialogue style feature can be determined by a variety of different similarity algorithms, and the similarity algorithms can include cosine distance similarity algorithm, Euclidean distance similarity algorithm, etc., which can be specifically set according to the actual situation.

[0048] In implementation, the similarity between the first dialogue style feature and the second dialogue style feature can be calculated through a specified similarity algorithm. If the calculated similarity is greater than a preset similarity threshold, it can be determined that the interactive answer data matches the first dialogue style corresponding to the historical dialogue data. If the calculated similarity is less than the preset similarity threshold, it can be determined that the interactive answer data does not match the first dialogue style corresponding to the historical dialogue data. The similarity threshold can be set according to the actual situation, such as 80% or 90%, etc.

[0049] In step A410, based on the second judgment result, determine the supplementary verification result corresponding to the interactive answer data obtained from each human-machine interaction.

[0050] In implementation, the second judgment result can be used as the supplementary verification result corresponding to the interactive answer data obtained from this human-machine interaction. Or, the second judgment result can also be combined with other relevant information to finally determine the supplementary verification result corresponding to the interactive answer data obtained from this human-machine interaction, etc., which can be specifically set according to the actual situation.

[0051] In practical applications, the verification rule includes a rule for judging whether the interactive answer data matches the second dialogue style corresponding to the preset generated dialogue data. Based on this, the specific processing method of the above step A4 can be various. Hereinafter, an optional processing method is provided, such as Figure 2 As shown, it can specifically include the processing from step A412 to step A416 as follows.

[0052] In step A412, extract the dialogue style feature from the interactive answer data obtained from each human-machine interaction to obtain the third dialogue style feature corresponding to the interactive answer data obtained from each human-machine interaction.

[0053] In step A414, based on the similarity between the third dialogue style feature and the fourth dialogue style feature corresponding to the generated dialogue data, judge whether the interactive answer data matches the second dialogue style corresponding to the preset generated dialogue data, and obtain the third judgment result corresponding to the interactive answer data obtained from each human-machine interaction.

[0054] In step A416, based on the third judgment result, determine the supplementary verification result corresponding to the interactive answer data obtained from each human-machine interaction.

[0055] In practical applications, large language models and verification models can be obtained through model training in the following manner of step B02 to step B10.

[0056] In step B02, obtain the first historical conversation data of the first user in the human-computer interaction and the supplementary verification information for the first user.

[0057] In step B04, input the first historical conversation data and the supplementary verification information for the first user into the large language model to guide the large language model to generate multiple first interactive question data through the first historical conversation data and the supplementary verification information for the first user. The first interactive question data can guide the first user to provide conversation content matching the conversation style of the first user and is embedded with information related to the first user.

[0058] In step B06, obtain the first interactive answer data provided by the first user for each first interactive question data, and input each first interactive answer data and the supplementary verification information for the first user into the verification model to judge the conversation style situation corresponding to each first interactive question data and the embedding situation of the supplementary verification information for the first user through the verification model, and obtain the fourth judgment result corresponding to each first interactive answer data.

[0059] In implementation, as Figure 3 shown, obtain the first interactive answer data provided by the first user for each first interactive question data. The specific processing can refer to the foregoing relevant content and will not be elaborated here. Then, each first interactive answer data and the supplementary verification information for the first user can be input into the verification model. The verification model can judge each first interactive answer data, thereby judging the conversation style situation corresponding to each first interactive question data and the embedding situation of the supplementary verification information for the first user, and obtaining the fourth judgment result corresponding to each first interactive answer data. Among them, the conversation style situation corresponding to each first interactive question data and the embedding situation of the supplementary verification information for the first user can correspond to multiple different evaluation metrics. For example, the conversation style situation corresponding to each first interactive question data can correspond to a conversation style guidance metric, and this evaluation metric can be used to judge the guidance situation of this first interactive question data on the user's conversation style (the larger this evaluation metric, the more the first interactive question data can guide the user's true conversation style). In addition, other metrics can also be corresponding, and can be specifically set according to the actual situation; in addition, the embedding situation of the supplementary verification information for the first user can correspond to an implicit embedding metric for the supplementary verification information, and this evaluation metric can be used to judge the implicit embedding situation of the supplementary verification information for the first user (the larger this evaluation metric, the more the supplementary verification information for the first user can be implicitly embedded into the interactive question data). In addition, other metrics can also be corresponding, and can be specifically set according to the actual situation. For example, asFigure 3 As shown, the large language model generated 2 pieces of first interaction question data, specifically A and B, to verify that the large model is set with an implicit embedding index for dialogue style guidance and supplementary verification information. Through verification, the following results can be obtained from the large model: for the dialogue style guidance index, A > B; for the implicit embedding index of supplementary verification information, B > A.

[0060] In step B08, based on the fourth judgment result corresponding to each first interaction answer data, the corresponding reward model is trained to obtain the trained reward model corresponding to each fourth judgment result.

[0061] In implementation, a corresponding reward model can be set for each of the above evaluation metrics. For example, as Figure 3 shown, if there are 2 evaluation metrics set, then a corresponding reward model is set for each evaluation metric, that is, if there are 2 evaluation metrics set, then 2 reward models are correspondingly set. Based on the above example, as Figure 3 shown, based on the dialogue style guidance index A > B, the reward model corresponding to the dialogue style guidance index can be trained to obtain the trained reward model corresponding to the dialogue style guidance index. Based on the implicit embedding index B > A of the supplementary verification information, the reward model corresponding to the implicit embedding index of the supplementary verification information can be trained to obtain the trained reward model corresponding to the implicit embedding index of the supplementary verification information. During the training process of the reward model, it can be processed in the following way: based on the fourth judgment result corresponding to each first interaction answer data and the optimal reward function, determine the optimal probability distribution corresponding to the corresponding reward model. Based on the optimal probability distribution corresponding to the reward model, determine the loss function corresponding to the reward model. Furthermore, based on the fourth judgment result corresponding to each first interaction answer data, train the reward model through the loss function corresponding to the reward model to obtain the trained reward model.

[0062] In step B10, based on the trained reward model corresponding to each fourth judgment result, the large language model and the verification large model are jointly trained through reinforcement learning to obtain the trained large language model and the trained verification large model.

[0063] In implementation, based on the trained reward model, each fourth judgment result, the processing strategy corresponding to the large language model, and the processing strategy corresponding to the fine-tuned large language model, the objective function for fine-tuning the large language model can be determined. Furthermore, based on the trained reward model and the large language model, through each fourth judgment result, and using the preset objective function as the fine-tuning target, the large language model is fine-tuned. In this way, the large language model and the verification large model are jointly trained through reinforcement learning to obtain the trained large language model and the trained verification large model.

[0064] The following provides a detailed description of an identity verification method provided in the embodiments of this specification in combination with specific application scenarios. The identity verification herein can be to verify the identity of a user through video, and the large language model is set with an implicit embedding index for guiding the dialogue style and supplementary verification information. In step C02, obtain the first historical dialogue data of the first user in the human-computer interaction and the supplementary verification information for the first user.

[0065] In step C04, input the first historical dialogue data and the supplementary verification information for the first user into the large language model, so as to generate two first interactive question data through the first historical dialogue data and the supplementary verification information of the first user. The first interactive question data can guide the first user to provide dialogue content matching the dialogue style of the first user and is embedded with information related to the first user.

[0066] In step C06, obtain the first interactive answer data provided by the first user for each first interactive question data, and input each first interactive answer data and the supplementary verification information of the first user into the verification large language model, so as to respectively judge the dialogue style situation corresponding to each first interactive question data and the embedding situation of the supplementary verification information of the first user through the dialogue style guiding index and the implicit embedding index of the supplementary verification information in the verification large language model, and obtain a fourth judgment result corresponding to each first interactive answer data.

[0067] In step C08, based on the two fourth judgment results, train the two reward models respectively to obtain two trained reward models.

[0068] In step C10, based on the two trained reward models, jointly train the large language model and the verification large language model through reinforcement learning to obtain a trained large language model and a trained verification large language model.

[0069] In step C12, verify the identity of the user through video, and obtain the historical dialogue data of the user in the human-computer interaction and the supplementary verification information for the user. The supplementary verification information includes information related to the attributes and / or behaviors of the user.

[0070] In step C14, input the historical dialogue data and the supplementary verification information into the large language model, so as to generate interactive question data through the historical dialogue data and the supplementary verification information. The interactive question data can guide the user to provide dialogue content matching the dialogue style of the user and is embedded with information related to the supplementary verification information.

[0071] In step C16, obtain the interactive answer data provided by the user for the interactive question data. If it is determined based on the interactive question data and the interactive answer data that the user needs to be further verified, then based on the historical conversation data of the user in the human-computer interaction and the supplementary verification information for the user, conduct a human-computer interaction with the user again through a large language model.

[0072] As Figure 4 shown, the interactive answer data obtained from each human-computer interaction can be supplemented and verified from three perspectives, namely whether the conversation style is consistent with the historical conversation data of the user, whether it is similar to the conversation style corresponding to the AIGC-generated conversation data, and whether it is consistent with the Q&A information (i.e., supplementary verification information) in the user's KBA Q&A knowledge base. For the first two perspectives, the similarity can be extracted and calculated through conversation style features. For the third perspective, it can be judged by the verification large model. Specifically, refer to the processing in steps C18 to C32 below.

[0073] In step C18, obtain the interactive answer data in the human-computer interaction data obtained from each human-computer interaction.

[0074] In step C20, input the interactive answer data obtained from each human-computer interaction and the supplementary verification information of the user into the pre-trained verification large model to judge whether the interactive answer data is consistent with the supplementary verification information of the user through the verification large model, and obtain the first judgment result corresponding to the interactive answer data obtained from each human-computer interaction.

[0075] In step C22, based on the first judgment result, determine the supplementary verification result corresponding to the interactive answer data obtained from each human-computer interaction.

[0076] In step C24, extract the conversation style features of the interactive answer data and the historical conversation data obtained from each human-computer interaction respectively, and obtain the first conversation style feature corresponding to the interactive answer data obtained from each human-computer interaction and the second conversation style feature corresponding to the historical conversation data.

[0077] In step C26, judge whether the interactive answer data matches the first conversation style corresponding to the historical conversation data based on the similarity between the first conversation style feature and the second conversation style feature, and obtain the second judgment result corresponding to the interactive answer data obtained from each human-computer interaction.

[0078] In step C28, based on the second judgment result, determine the supplementary verification result corresponding to the interactive answer data obtained from each human-computer interaction.

[0079] In step C30, based on the similarity between the first dialogue style feature and the fourth dialogue style feature corresponding to the generated dialogue data, it is determined whether the interactive answer data matches the second dialogue style corresponding to the preset generated dialogue data, and a third judgment result corresponding to the interactive answer data obtained from each human-machine interaction is obtained.

[0080] In step C32, based on the third judgment result, the supplementary verification result corresponding to the interactive answer data obtained from each human-machine interaction is determined.

[0081] In step C34, based on the supplementary verification result corresponding to the interactive answer data obtained from each human-machine interaction, the supplementary verification result for the user's identity is determined, and the supplementary verification result for the user's identity can determine that the accuracy of the user's identity reaches more than a preset threshold.

[0082] In step C36, the identity verification result of verifying the user's identity through video is obtained, and based on the identity verification result and the supplementary verification result of the identity, it is determined whether the user's identity verification is passed.

[0083] It should be noted that considering the user-related information such as the user attributes and user behaviors of the users involved in this specification, as well as the relevant data such as historical dialogue data and supplementary verification information may be privacy data of the users to a certain extent. Therefore, if you want to collect the above-mentioned relevant information, historical dialogue data and supplementary verification information of the users, you can obtain the user's authorization before collecting the data, so that the operation of collecting data complies with relevant data management regulations. For example, data authorization can be carried out during the process of the user applying for identity verification on the identity verification platform, or data authorization can also be carried out during the process of the user's first service access on the identity verification platform; the specific method of data authorization can be to send a user data authorization reminder to the user, and the user can obtain data collection authorization after confirming the reminder through an instruction. Or, the method of data authorization can also be to obtain data collection authorization by signing a data authorization agreement.

[0084] The embodiments of this specification provide an identity verification method. When it is detected that a user is performing identity verification, historical conversation data of the user in human-computer interaction and supplementary verification information for the user are obtained. The supplementary verification information includes information related to the attributes and / or behaviors of the user. Then, the historical conversation data and the supplementary verification information are input into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information. The interactive question data can guide the user to provide conversation content that matches the user's conversation style and embeds information related to the supplementary verification information. After that, the interactive answer data provided by the user for the interactive question data can be obtained. If it is determined based on the interactive question data and the interactive answer data that the user needs to be subjected to supplementary verification processing again, then based on the historical conversation data of the user in human-computer interaction and the supplementary verification information for the user, the large language model is used to perform human-computer interaction with the user again until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity exceeds a preset threshold. Finally, the identity verification result of the user's identity verification can be obtained, and based on the identity verification result and the identity supplementary verification result, it is determined whether the user's identity verification is passed. In this way, an identity verification method that embeds user conversation style judgment and implicit KBA question and answer (i.e., supplementary verification information for the user) is used to enhance the defense ability of the identity verification scenario against AIGC. That is, user conversation style judgment and implicit KBA question and answer are embedded in the identity verification scenario to perform supplementary verification on the user's identity. In this way, in identity verification, by guiding the user to reply with more stylized content and including knowledge information related to the user, implicit verification is naturally completed, thereby improving the security of identity verification. Moreover, implicit questions can be asked based on knowledge information such as the user's supplementary verification information to ensure the user experience during the identity verification process.

[0085] The above is the identity verification method provided by the embodiments of this specification. Based on the same idea, the embodiments of this specification also provide an identity verification device, as Figure 5 shown.

[0086] The identity verification device includes: a data acquisition module 501, an interactive question generation module 502, a supplementary verification module 503, and an identity verification module 504, where: The data acquisition module 501, when detecting that a user is performing identity verification, acquires the historical conversation data of the user in human-computer interaction and the supplementary verification information for the user, and the supplementary verification information includes information related to the attributes and / or behaviors of the user; The interaction problem generation module 502 inputs the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interaction problem data through the historical conversation data and the supplementary verification information. The interaction problem data can guide the user to provide conversation content matching the user's conversation style and embed information related to the supplementary verification information. The supplementary verification module 503 obtains the interaction answer data provided by the user for the interaction problem data. If it is determined based on the interaction problem data and the interaction answer data that the user needs to be subjected to supplementary verification processing again, then based on the historical conversation data of the user in the human-computer interaction and the supplementary verification information for the user, the large language model is used to conduct human-computer interaction with the user again until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity reaches beyond a preset threshold. The identity verification module 504 obtains the identity verification result of the user for identity verification and determines whether the identity verification of the user passes based on the identity verification result and the identity supplementary verification result.

[0087] In the embodiments of this specification, the device further includes: The answer acquisition module acquires the interaction answer data in the human-computer interaction data obtained from each human-computer interaction. The verification module conducts supplementary verification processing on the user respectively through preset verification rules based on the interaction answer data obtained from each human-computer interaction, and determines the supplementary verification result corresponding to the interaction answer data obtained from each human-computer interaction. The verification rules include a rule for judging whether the interaction answer data is consistent with the user's supplementary verification information and a matching verification rule for the conversation style. The verification result determination module determines the identity supplementary verification result for the user based on the supplementary verification result corresponding to the interaction answer data obtained from each human-computer interaction.

[0088] In the embodiments of this specification, the matching verification rule for the conversation style includes a rule for judging whether the interaction answer data matches the first conversation style corresponding to the historical conversation data and / or a rule for judging whether the interaction answer data matches the second conversation style corresponding to the preset generated conversation data. The generated conversation data is conversation data generated by a specified network model.

[0089] In the embodiments of this specification, the verification rule includes a rule for judging whether the interaction answer data is consistent with the user's supplementary verification information. The verification module includes: The first verification unit inputs the interaction answer data obtained from each human-machine interaction and the supplementary verification information of the user into a pre-trained verification large model, so as to judge whether the interaction answer data is consistent with the supplementary verification information of the user through the verification large model, and obtain a first judgment result corresponding to the interaction answer data obtained from each human-machine interaction; The first verification result determination unit determines the supplementary verification result corresponding to the interaction answer data obtained from each human-machine interaction based on the first judgment result.

[0090] In the embodiments of the present specification, the verification rule includes a rule for judging whether the interaction answer data matches the first dialogue style corresponding to the historical dialogue data. The verification module includes: The first feature extraction unit respectively extracts the dialogue style features of the interaction answer data obtained from each human-machine interaction and the historical dialogue data, and obtains the first dialogue style feature corresponding to the interaction answer data obtained from each human-machine interaction and the second dialogue style feature corresponding to the historical dialogue data; The second verification unit judges whether the interaction answer data matches the first dialogue style corresponding to the historical dialogue data based on the similarity between the first dialogue style feature and the second dialogue style feature, and obtains a second judgment result corresponding to the interaction answer data obtained from each human-machine interaction; The second verification result determination unit determines the supplementary verification result corresponding to the interaction answer data obtained from each human-machine interaction based on the second judgment result.

[0091] In the embodiments of the present specification, the verification rule includes a rule for judging whether the interaction answer data matches the second dialogue style corresponding to the preset generated dialogue data. The verification module includes: The second feature extraction unit extracts the dialogue style features of the interaction answer data obtained from each human-machine interaction, and obtains the third dialogue style feature corresponding to the interaction answer data obtained from each human-machine interaction; The third verification unit judges whether the interaction answer data matches the second dialogue style corresponding to the preset generated dialogue data based on the similarity between the third dialogue style feature and the fourth dialogue style feature corresponding to the generated dialogue data, and obtains a third judgment result corresponding to the interaction answer data obtained from each human-machine interaction; The third verification result determination unit determines the supplementary verification result corresponding to the interaction answer data obtained from each human-machine interaction based on the third judgment result.

[0092] In the embodiments of the present specification, the device further includes: The information acquisition module acquires the first historical dialogue data of the first user in the human-machine interaction and the supplementary verification information for the first user; The first question generation module inputs the first historical dialogue data and supplementary verification information for the first user into a large language model, so as to guide the large language model to generate multiple first interactive question data through the first historical dialogue data and the supplementary verification information of the first user. The first interactive question data can guide the first user to provide dialogue content matching the dialogue style of the first user and is embedded with information related to the first user; The judgment module obtains the first interactive answer data provided by the first user for each of the first interactive question data, and inputs each first interactive answer data and the supplementary verification information of the first user into a verification large model, so as to judge the dialogue style situation corresponding to each first interactive question data and the embedding situation of the supplementary verification information of the first user through the verification large model, and obtain a fourth judgment result corresponding to each first interactive answer data; The first training module trains the corresponding reward model based on the fourth judgment result corresponding to each first interactive answer data to obtain a trained reward model corresponding to each fourth judgment result; The joint training module jointly trains the large language model and the verification large model in a reinforcement learning manner based on the trained reward model corresponding to each fourth judgment result to obtain a trained large language model and a trained verification large model.

[0093] In the embodiments of the present specification, the identity verification result of the user's identity verification is the identity verification result of verifying the user's identity by video.

[0094] An embodiment of this specification provides an identity verification device. When it detects that a user is performing identity verification, it obtains the user's historical conversation data in human-computer interaction and supplementary verification information for the user. The supplementary verification information includes information related to the user's attributes and / or user behavior. Then, it inputs the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information. The interactive question data can guide the user to provide conversation content that matches the user's conversation style and embeds information related to the supplementary verification information. After that, it can obtain the interactive answer data provided by the user for the interactive question data. If it is determined based on the interactive question data and the interactive answer data that the user needs to be subjected to supplementary verification processing again, then based on the user's historical conversation data in human-computer interaction and the supplementary verification information for the user, it conducts human-computer interaction with the user again through the large language model until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity exceeds a preset threshold. Finally, it can obtain the identity verification result of the user's identity verification and determine whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result. In this way, an identity verification method that embeds user conversation style judgment and implicit KBA question and answer (i.e., supplementary verification information for the user) is used to enhance the defense ability of the identity verification scenario against AIGC. That is, in the identity verification scenario, user conversation style judgment and implicit KBA question and answer are embedded to supplement the verification of the user's identity. In this way, in identity verification, by guiding the user to reply with more stylized content, which also contains knowledge information related to the user, implicit verification is naturally completed, thereby improving the security of identity verification. Moreover, it can implicitly ask questions based on knowledge information such as the user's supplementary verification information to ensure the user experience during the identity verification process.

[0095] The above is the identity verification device provided by the embodiment of this specification. Based on the same idea, the embodiment of this specification also provides an identity verification device, as Figure 6 shown.

[0096] The identity verification device may be the terminal device or server provided in the above embodiment, etc.

[0097] Identity verification devices can vary significantly due to differences in configuration or performance. They may include one or more processors 601 and a memory 602. The memory 602 can store one or more stored application programs or data. Among them, the memory 602 can be short-term storage or persistent storage. The application programs stored in the memory 602 can include one or more modules (not shown in the figure), and each module can include a series of computer-executable instructions in the identity verification device. Further, the processor 601 can be set to communicate with the memory 602 and execute a series of computer-executable instructions in the memory 602 on the identity verification device. The identity verification device can also include one or more power supplies 603, one or more wired or wireless network interfaces 604, one or more input / output interfaces 605, and one or more keyboards 606.

[0098] Specifically, in this embodiment, the identity verification device includes a memory and one or more programs. One or more of the programs are stored in the memory, and one or more of the programs can include one or more modules. Each module can include a series of computer-executable instructions in the identity verification device and is configured to be executed by one or more processors. The one or more programs include the following computer-executable instructions: When it is detected that the user is performing identity verification, obtain the user's historical conversation data in the human-computer interaction and supplementary verification information for the user. The supplementary verification information includes information related to the user's attributes and / or user behavior; Input the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information. The interactive question data can guide the user to provide conversation content matching the user's conversation style and embed information related to the supplementary verification information; Obtain the interactive answer data provided by the user for the interactive question data. If it is determined based on the interactive question data and the interactive answer data that the user needs to be subjected to supplementary verification processing again, then based on the user's historical conversation data in the human-computer interaction and the supplementary verification information for the user, conduct human-computer interaction with the user again through the large language model until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity exceeds a preset threshold; Obtain the identity verification result of the user's identity verification and determine whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result.

[0099] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiment of the identity verification device, since it is basically similar to the method embodiment, the description is relatively simple. For the relevant parts, reference can be made to the corresponding description in the method embodiment.

[0100] An embodiment of this specification provides an identity verification device. When detecting that a user performs identity verification, the device obtains the historical conversation data of the user in the human-computer interaction and the supplementary verification information for the user. The supplementary verification information includes information related to the attributes and / or behaviors of the user. Then, the historical conversation data and the supplementary verification information are input into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information. The interactive question data can guide the user to provide conversation content matching the user's conversation style and embed information related to the supplementary verification information. After that, the interactive answer data provided by the user for the interactive question data can be obtained. If it is determined based on the interactive question data and the interactive answer data that the user needs to be subjected to supplementary verification processing again, then based on the historical conversation data of the user in the human-computer interaction and the supplementary verification information for the user, the large language model is used to perform human-computer interaction with the user again until the accuracy of the identity supplementary verification result for the user obtained based on the human-computer interaction data exceeds a preset threshold. Finally, the identity verification result of the user's identity verification can be obtained, and based on the identity verification result and the identity supplementary verification result, it is determined whether the user's identity verification is passed. In this way, an identity verification method that embeds the judgment of the user's conversation style and implicit KBA Q&A (i.e., the supplementary verification information for the user) is used to enhance the defense ability of the identity verification scenario against AIGC. That is, in the identity verification scenario, the judgment of the user's conversation style and implicit KBA Q&A are embedded to perform supplementary verification on the user's identity. In this way, in identity verification, by guiding the user to reply with more stylized content and including knowledge information related to the user, implicit verification is naturally completed, thereby improving the security of identity verification. Moreover, implicit questions can be asked based on knowledge information such as the user's supplementary verification information to ensure the user experience during the identity verification process.

[0101] Furthermore, based on the above Figures 1 to 4 , one or more embodiments of this specification also provide a storage medium for storing computer-executable instruction information. In a specific embodiment, the storage medium can be a USB flash drive, an optical disc, a hard disk, etc. When the computer-executable instruction information stored in the storage medium is executed by a processor, the following process can be implemented: When it is detected that the user conducts identity verification, obtain the user's historical conversation data in human-computer interaction and supplementary verification information for the user, where the supplementary verification information includes information related to the user's attributes and / or user behavior; Input the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information. The interactive question data can guide the user to provide conversation content matching the user's conversation style and embed information related to the supplementary verification information; Obtain the interactive answer data provided by the user for the interactive question data. If it is determined based on the interactive question data and the interactive answer data that the user needs to be subjected to supplementary verification processing again, then based on the user's historical conversation data in human-computer interaction and the supplementary verification information for the user, conduct human-computer interaction with the user again through the large language model until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity exceeds a preset threshold; Obtain the identity verification result of the user's identity verification, and determine whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result.

[0102] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments. In particular, for the above-mentioned storage medium embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment.

[0103] An embodiment of this specification provides a storage medium. When it is detected that a user is performing identity verification, historical conversation data of the user in human-computer interaction and supplementary verification information for the user are obtained. The supplementary verification information includes information related to the attributes and / or behaviors of the user. Then, the historical conversation data and the supplementary verification information are input into a large language model to guide the large language model to generate interaction question data through the historical conversation data and the supplementary verification information. The interaction question data can guide the user to provide conversation content matching the user's conversation style and embed information related to the supplementary verification information. After that, interaction answer data provided by the user for the interaction question data can be obtained. If it is determined based on the interaction question data and the interaction answer data that the user needs to be subjected to supplementary verification processing again, then based on the historical conversation data of the user in human-computer interaction and the supplementary verification information for the user, the large language model is used to perform human-computer interaction with the user again until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity reaches a level exceeding a preset threshold. Finally, an identity verification result of the user's identity verification can be obtained, and based on the identity verification result and the identity supplementary verification result, it can be determined whether the user's identity verification is passed. In this way, an identity verification method that embeds user conversation style judgment and implicit KBA question and answer (i.e., supplementary verification information for the user) is used to enhance the defense ability of the identity verification scenario against AIGC. That is, user conversation style judgment and implicit KBA question and answer are embedded in the identity verification scenario to perform supplementary verification of the user's identity. Thus, in identity verification, by guiding the user to reply with more stylized content and simultaneously including knowledge information related to the user, implicit verification is naturally completed, thereby improving the security of identity verification. Moreover, implicit questions can be asked based on knowledge information such as the user's supplementary verification information to ensure the user experience during the identity verification process.

[0104] Furthermore, based on the above Figures 1 to 4 , one or more embodiments of this specification also provide a computer program product, including a computer program. When the computer program in this computer program product is executed by a processor, the following processes can be implemented: When it is detected that a user is performing identity verification, obtain the historical conversation data of the user in human-computer interaction and the supplementary verification information for the user, where the supplementary verification information includes information related to the attributes and / or behaviors of the user; Input the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interaction question data through the historical conversation data and the supplementary verification information, where the interaction question data can guide the user to provide conversation content matching the user's conversation style and embed information related to the supplementary verification information; Obtain the interaction answer data provided by the user for the interaction problem data. If it is determined based on the interaction problem data and the interaction answer data that the user needs to be supplemented and verified again, then based on the historical conversation data of the user in the human-computer interaction and the supplementary verification information for the user, interact with the user again through the large language model until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity exceeds a preset threshold; Obtain the identity verification result of the user's identity verification, and determine whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result.

[0105] Each embodiment in this specification is described in a progressive manner. For the same or similar parts between the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the above-mentioned embodiment of a computer program product, since it is basically similar to the method embodiment, the description is relatively simple, and for the relevant parts, reference can be made to the partial description of the method embodiment.

[0106] An embodiment of this specification provides a computer program product. When it detects that a user is performing identity verification, it obtains the user's historical conversation data in human-computer interaction and supplementary verification information for the user. The supplementary verification information includes information related to the user's attributes and / or user behavior. Then, it inputs the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interaction question data through the historical conversation data and the supplementary verification information. The interaction question data can guide the user to provide conversation content that matches the user's conversation style and embeds information related to the supplementary verification information. After that, it can obtain the interaction answer data provided by the user for the interaction question data. If it is determined based on the interaction question data and the interaction answer data that the user needs to be subjected to supplementary verification processing again, it will conduct human-computer interaction with the user again through the large language model based on the user's historical conversation data in human-computer interaction and the supplementary verification information for the user, until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity exceeds a preset threshold. Finally, it can obtain the identity verification result of the user's identity verification and determine whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result. In this way, an identity verification method that embeds user conversation style judgment and implicit KBA question and answer (i.e., supplementary verification information for the user) is used to enhance the defense ability of the identity verification scenario against AIGC. That is, user conversation style judgment and implicit KBA question and answer are embedded in the identity verification scenario to conduct supplementary verification of the user's identity. Thus, in identity verification, by guiding the user to reply with more stylized content, which also contains knowledge information related to the user, implicit verification is naturally completed, thereby improving the security of identity verification. Moreover, it can implicitly ask questions based on knowledge information such as the user's supplementary verification information to ensure the user experience during the identity verification process.

[0107] The above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require the specific order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0108] In the 1990s, it was quite obvious to distinguish whether an improvement in a technology was an improvement in hardware (e.g., improvement in circuit structures such as diodes, transistors, switches, etc.) or an improvement in software (improvement in method flows). However, with the development of technology, many improvements in method flows today can be regarded as direct improvements in hardware circuit structures. Almost all designers obtain the corresponding hardware circuit structures by programming the improved method flows into the hardware circuits. Therefore, it cannot be said that an improvement in a method flow cannot be implemented with a hardware entity module. For example, a programmable logic device (PLD) (e.g., a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by a user's programming of the device. Designers can program by themselves to "integrate" a digital system on a PLD, without having to ask a chip manufacturer to design and fabricate a dedicated integrated circuit chip. Moreover, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly implemented using "logic compiler" software, which is similar to the software compiler used in program development and writing. The original code before compilation also has to be written in a specific programming language, which is called a hardware description language (HDL), and there is not only one kind of HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. Currently, the most commonly used are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also be aware that by simply performing a little logical programming on the method flow with the above-mentioned several hardware description languages and programming it into an integrated circuit, it is easy to obtain a hardware circuit that implements the logical method flow.

[0109] The controller can be implemented in any suitable manner. For example, the controller can take the form of, for example, a microprocessor or a processor and a computer-readable medium storing computer-readable program code (such as software or firmware) executable by the (micro)processor, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of the controller include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in the form of pure computer-readable program code, it is entirely possible to make the controller implement the same function in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. by logically programming the method steps. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be regarded as the structures within the hardware component. Or even, the devices for implementing various functions can be regarded as either software modules for implementing the method or structures within the hardware component.

[0110] The systems, devices, modules, or units illustrated in the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0111] For the convenience of description, when describing the above devices, they are described separately as various units according to their functions. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0112] Those skilled in the art should understand that the embodiments of this specification can be provided as a method, a system, or a computer program product. Therefore, one or more embodiments of this specification can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code.

[0113] Embodiments of this specification are described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It should be understood that each flow and / or block in the flowchart and / or block diagram, and combinations of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable serial-parallel devices for fraud cases to generate a machine, such that the instructions executed by the processors of the computer or other programmable serial-parallel devices for fraud cases generate means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0114] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable serial-parallel device for fraud cases to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0115] These computer program instructions can also be loaded onto a computer or other programmable serial-parallel device for fraud cases, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0116] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0117] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0118] A computer-readable medium includes both permanent and non-permanent, removable and non-removable media and can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to store information accessible by a computing device. As defined herein, a computer-readable medium does not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0119] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but also other elements not expressly listed, or elements that are inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0120] Those skilled in the art should understand that the embodiments of this specification can be provided as a method, system, or computer program product. Therefore, one or more embodiments of this specification can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0121] One or more embodiments of this specification can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. One or more embodiments of this specification can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0122] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and for the relevant parts, reference can be made to the partial description of the method embodiment.

[0123] The above is only the embodiment of this specification and is not used to limit this document. For those skilled in the art, various modifications and changes can be made to this specification. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this specification shall be included within the scope of the claims of this specification.

Claims

1. An identity verification method, the method comprising: When it is detected that a user performs identity verification, obtaining historical conversation data of the user in human-computer interaction and supplementary verification information for the user, where the supplementary verification information includes information related to the attributes and / or user behavior of the user; Inputting the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information, where the interactive question data can guide the user to provide conversation content matching the user's conversation style and is embedded with information related to the supplementary verification information; Obtaining interactive answer data provided by the user for the interactive question data. If it is determined based on the interactive question data and the interactive answer data that the user needs to be subjected to supplementary verification processing again, then based on the historical conversation data of the user in human-computer interaction and the supplementary verification information for the user, the large language model is used to perform human-computer interaction with the user again until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity reaches beyond a preset threshold; Obtaining the identity verification result of the user's identity verification, and determining whether the user's identity verification passes based on the identity verification result and the identity supplementary verification result.

2. The method according to claim 1, the method further comprising: Obtaining interactive answer data in the human-computer interaction data obtained each time; Based on the interactive answer data obtained each time, respectively performing supplementary verification processing on the user through preset verification rules to determine the supplementary verification result corresponding to the interactive answer data obtained each time, where the verification rules include rules for determining whether the interactive answer data is consistent with the supplementary verification information of the user and matching verification rules for the conversation style; Based on the supplementary verification result corresponding to the interactive answer data obtained each time, determining the identity supplementary verification result for the user.

3. The method according to claim 2, the matching verification rules for the conversation style include rules for determining whether the interactive answer data matches the first conversation style corresponding to the historical conversation data and / or rules for determining whether the interactive answer data matches the second conversation style corresponding to preset generated conversation data, where the generated conversation data is conversation data generated by a specified network model.

4. The method according to claim 2, the verification rules include rules for determining whether the interactive answer data is consistent with the supplementary verification information of the user. Based on the interactive answer data obtained each time, respectively performing supplementary verification processing on the user through preset verification rules to determine the supplementary verification result corresponding to the interactive answer data obtained each time, including: Input the interactive answer data obtained from each human-machine interaction and the supplementary verification information of the user into a pre-trained verification large model to determine whether the interactive answer data is consistent with the supplementary verification information of the user through the verification large model, and obtain a first judgment result corresponding to the interactive answer data obtained from each human-machine interaction; Based on the first judgment result, determine the supplementary verification result corresponding to the interactive answer data obtained from each human-machine interaction.

5. The method according to claim 3, wherein the verification rule includes a rule for judging whether the interactive answer data matches the first dialogue style corresponding to the historical dialogue data. Based on the interactive answer data obtained from each human-machine interaction, the user is respectively subjected to supplementary verification processing through a preset verification rule to determine the supplementary verification result corresponding to the interactive answer data obtained from each human-machine interaction, including: Respectively extract the dialogue style features of the interactive answer data obtained from each human-machine interaction and the historical dialogue data to obtain a first dialogue style feature corresponding to the interactive answer data obtained from each human-machine interaction and a second dialogue style feature corresponding to the historical dialogue data; Based on the similarity between the first dialogue style feature and the second dialogue style feature, judge whether the interactive answer data matches the first dialogue style corresponding to the historical dialogue data, and obtain a second judgment result corresponding to the interactive answer data obtained from each human-machine interaction; Based on the second judgment result, determine the supplementary verification result corresponding to the interactive answer data obtained from each human-machine interaction.

6. The method according to claim 3, wherein the verification rule includes a rule for judging whether the interactive answer data matches the second dialogue style corresponding to the preset generated dialogue data. Based on the interactive answer data obtained from each human-machine interaction, the user is respectively subjected to supplementary verification processing through a preset verification rule to determine the supplementary verification result corresponding to the interactive answer data obtained from each human-machine interaction, including: Extract the dialogue style features of the interactive answer data obtained from each human-machine interaction to obtain a third dialogue style feature corresponding to the interactive answer data obtained from each human-machine interaction; Based on the similarity between the third dialogue style feature and the fourth dialogue style feature corresponding to the generated dialogue data, judge whether the interactive answer data matches the second dialogue style corresponding to the preset generated dialogue data, and obtain a third judgment result corresponding to the interactive answer data obtained from each human-machine interaction; Based on the third judgment result, determine the supplementary verification result corresponding to the interactive answer data obtained from each human-machine interaction.

7. The method according to claim 4, wherein the method further includes: Obtain the first historical dialogue data of the first user in the human-machine interaction and the supplementary verification information for the first user; Input the first historical dialogue data and supplementary verification information for the first user into a large language model, so as to guide the large language model to generate multiple first interactive question data through the first historical dialogue data and the supplementary verification information of the first user. The first interactive question data can guide the first user to provide dialogue content matching the dialogue style of the first user and is embedded with information related to the first user; Obtain the first interactive answer data provided by the first user for each of the first interactive question data, and input each first interactive answer data and the supplementary verification information of the first user into a verification large model, so as to judge the dialogue style situation corresponding to each first interactive question data and the embedding situation of the supplementary verification information of the first user through the verification large model, and obtain a fourth judgment result corresponding to each first interactive answer data; Based on the fourth judgment result corresponding to each first interactive answer data, train the corresponding reward model to obtain a trained reward model corresponding to each fourth judgment result; Based on the trained reward model corresponding to each fourth judgment result, jointly train the large language model and the verification large model through reinforcement learning to obtain a trained large language model and a trained verification large model.

8. The method according to any one of claims 1-7, wherein the identity verification result of the user's identity verification is the identity verification result of verifying the user's identity by video.

9. An identity verification device, the device comprising: A data acquisition module, when detecting that a user performs identity verification, acquires the historical dialogue data of the user in human-computer interaction and supplementary verification information for the user, where the supplementary verification information includes information related to the attributes and / or user behaviors of the user; An interactive question generation module, which inputs the historical dialogue data and the supplementary verification information into a large language model, so as to guide the large language model to generate interactive question data through the historical dialogue data and the supplementary verification information. The interactive question data can guide the user to provide dialogue content matching the dialogue style of the user and is embedded with information related to the supplementary verification information; A supplementary verification module, which acquires the interactive answer data provided by the user for the interactive question data. If it is determined based on the interactive question data and the interactive answer data that the user needs to be subjected to supplementary verification processing again, then based on the historical dialogue data of the user in human-computer interaction and the supplementary verification information for the user, the large language model is used to perform human-computer interaction with the user again until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity reaches a preset threshold; An identity verification module, which acquires the identity verification result of the user's identity verification, and determines whether the user's identity verification is passed based on the identity verification result and the identity supplementary verification result.

10. An identity verification device, the identity verification device comprising: A processor; and a memory arranged to store computer-executable instructions that, when executed, cause the processor to: When detecting that a user performs identity verification, obtain the user's historical conversation data in human-computer interaction and supplementary verification information for the user, where the supplementary verification information includes information related to the user's attributes and / or user behavior; Input the historical conversation data and the supplementary verification information into a large language model to guide the large language model to generate interactive question data through the historical conversation data and the supplementary verification information, where the interactive question data can guide the user to provide conversation content matching the user's conversation style and is embedded with information related to the supplementary verification information; Obtain the interactive answer data provided by the user for the interactive question data. If it is determined based on the interactive question data and the interactive answer data that the user needs to be subjected to supplementary verification processing again, then based on the user's historical conversation data in human-computer interaction and the supplementary verification information for the user, perform human-computer interaction with the user again through the large language model until the identity supplementary verification result for the user obtained based on the human-computer interaction data is used to determine that the accuracy of the user's identity reaches a level exceeding a preset threshold; Obtain the identity verification result of the user's identity verification, and determine whether the user's identity verification is passed based on the identity verification result and the identity supplementary verification result.

Images