A data security verification method for click-type dynamic verification code

Abstract

The present invention relates to the field of computer vision technology, and in particular to a data security verification method for a click-type dynamic verification code. The method comprises the following steps: constructing a secure communication tunnel between a target user client and a preset verification code server to obtain encrypted session key data; analyzing the human-computer interaction mode of the target user client based on the encrypted session key data to obtain user interaction feature data; generating personalized verification data for the target user client based on the user interaction feature data to obtain personalized verification code data; sending the personalized verification code data to the target user client based on the encrypted session key data; capturing multi-dimensional interactive behaviors of the target user client to obtain user cross-modal interactive response data. The present invention can improve the security of the verification code system, improve user experience, enhance the recognizability of the verification code, and effectively resist the threats of automated attacks and black and gray industries. It can grasp the multi-dimensional and multi-source information of land use, and can also accurately predict the changing trends and patterns of future land use.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and in particular to a data security verification method for a click-type dynamic verification code. Background Art

[0002] There are a series of technical difficulties in the existing dynamic verification code technology. First, the progress of machine learning and image recognition technology, especially the application of convolutional neural networks, poses a threat to the security of verification codes. Secondly, complex verification codes designed to improve security often damage the user experience, especially for visually impaired users. These verification codes are usually not friendly because they cannot be recognized by screen reading software, and the recognizability of sound verification codes is also low. In addition, the existing machine recognition algorithms have limited effects when facing specially designed verification codes, which shows that more advanced solutions are needed to deal with the threats of automated attacks and black and gray industries. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a data security verification method of a click-type dynamic verification code to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a data security verification method of a click-type dynamic verification code includes the following steps:

[0005] Step S1: construct a secure communication tunnel between the target user client and the preset verification code server to obtain encrypted session key data;

[0006] Step S2: Analyze the human-computer interaction mode of the target user client based on the encrypted session key data to obtain user interaction feature data;

[0007] Step S3: Generate personalized verification data for the target user client according to the user interaction feature data to obtain personalized verification code data;

[0008] Step S4: Sending the personalized verification code data to the target user client based on the encrypted session key data; capturing the target user client's multi-dimensional interactive behavior to obtain user cross-modal interactive response data;

[0009] Step S5: Perform cognitive delay pattern recognition on the user cross-modal interaction response data to obtain a user cognitive delay index; and grant or deny access rights to the target user according to the user cognitive delay index.

[0010] The present invention ensures that the communication between the target user client and the verification code server is encrypted by constructing a secure communication tunnel, thereby preventing man-in-the-middle attacks and data leakage. Through the secure communication tunnel, the two parties successfully establish a connection and exchange key data, ensuring that the data transmission in the subsequent steps is encrypted and secure. By analyzing the human-computer interaction mode of the target user client, the difference between a real user and an automated attack can be distinguished. This helps to distinguish malicious attackers from legitimate users and improve the security of the verification code. Human-computer interaction mode analysis can be used to identify whether the user is a visually impaired user. By identifying the visually impaired user, a verification code experience adapted to their special needs can be provided, such as using an audio verification code or providing auxiliary functions to improve the user experience and accessibility of the visually impaired user. By generating personalized verification code data according to user interaction feature data, the complexity and randomness of the verification code can be increased, making it difficult to be cracked by the algorithm of the automated attack. This helps to improve the security of the verification code and prevent the threat of malicious attacks and black and gray industries. Personalized verification code data is generated according to the user's interactive feature data, and verification codes suitable for their special needs are provided for normal users and visually impaired users respectively. For example, a regular image verification code can be generated for normal users, while an audio verification code or other accessibility options can be generated for visually impaired users. This helps meet the needs of different user groups and improve user experience and accessibility. By using encrypted session key data to send personalized verification code data to the target user client, it is ensured that the data is not stolen or tampered with during transmission, and the security of the verification code is maintained. By capturing the multi-dimensional interactive behavior of the target user client, a variety of interactive response data of the user in the process of identifying the verification code, such as clicks, slides, touches, etc., can be obtained. This helps to further analyze the user's cognitive patterns and behavioral characteristics and improve the accuracy and security of the verification code. By performing cognitive delay pattern recognition on the user's cross-modal interactive response data, the user's cognitive delay in the process of completing the verification code can be analyzed. This helps to determine whether the user is a legitimate user and whether there is an automated attack behavior. By identifying the user's cognitive delay, the security of the verification code and the ability to defend against automated attacks can be further improved. According to the user's cognitive delay index, the target user can be granted or denied access rights. When the user's cognitive delay index meets the range of legitimate users, the access right is granted; otherwise, the access request is denied. This helps to effectively prevent access by automated attackers and improve the security of the system. In summary, the present invention can improve the user experience and accessibility of visually impaired users, and can also increase the complexity and randomness of the verification code, making it difficult to be cracked by the algorithm of automated attacks, thereby improving the security of the verification code and preventing malicious attacks and threats from the black and gray industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0012] Figure 1 A schematic flow chart of the steps of a data security verification method of a click-type dynamic verification code according to an embodiment is shown.

[0013] Figure 2 A detailed flowchart diagram of step S33 of an embodiment is shown.

[0014] Figure 3 A detailed flowchart diagram of step S335 of an embodiment is shown. DETAILED DESCRIPTION

[0015] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0016] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0017] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0018] To achieve this, please refer to Figures 1 to 3 The present invention provides a data security verification method of a click-type dynamic verification code, comprising the following steps:

[0019] Step S1: construct a secure communication tunnel between the target user client and the preset verification code server to obtain encrypted session key data;

[0020] Specifically, for example, a secure communication protocol may be used to perform a handshake process, including steps such as key exchange, identity authentication, and negotiation of encryption algorithms. During the handshake process, the target user client and the verification code server will negotiate and generate an encrypted session key for subsequent communication encryption and decryption. The encrypted session key may be generated using a symmetric encryption algorithm. The key negotiation information exchanged between the target user client and the verification code server during the handshake process includes the encrypted session key. The target user client saves the encrypted session key data obtained from the handshake process for subsequent use.

[0021] Step S2: Analyze the human-computer interaction mode of the target user client based on the encrypted session key data to obtain user interaction feature data;

[0022] Specifically, for example, the target user client records the user's interactive behaviors during the process of entering the verification code, such as mouse clicks, keyboard input, scrolling behaviors, etc. The target user client encrypts the collected user interaction data according to the encrypted session key data to ensure the confidentiality of the data. The encrypted user interaction data is sent to the preset verification code server for further analysis. The preset verification code server receives the encrypted user interaction data. The preset verification code server decrypts the data using the same encryption session key as the target user client. The decrypted user interaction data is analyzed and processed on the preset verification code server to extract user interaction feature data. The preset verification code server focuses on the target user's interactive operations during the process of entering the verification code, such as the user's click position, number of clicks, keyboard input speed, etc. Based on the analysis results, the preset verification code server can detect abnormal interaction patterns, identify automated scripts or robots, etc.

[0023] Step S3: Generate personalized verification data for the target user client according to the user interaction feature data to obtain personalized verification code data;

[0024] Specifically, for example, the preset verification code server uses algorithms and models, such as machine learning or deep learning models, to analyze and process the user interaction feature data. Based on the user interaction feature data, the preset verification code server generates personalized verification code data. The personalized verification code data may include a specific image, text, or other form of verification code generated according to the user interaction mode.

[0025] Step S4: Sending the personalized verification code data to the target user client based on the encrypted session key data; capturing the target user client's multi-dimensional interactive behavior to obtain user cross-modal interactive response data;

[0026] Specifically, for example, the preset verification code server uses an encrypted session key to encrypt the personalized verification code data. The encrypted personalized verification code data is sent to the target user client through a secure communication channel. The target user client receives the encrypted personalized verification code data. The target user client uses the same encrypted session key to decrypt the received data. The decrypted personalized verification code data is displayed on the target user client, which can be an image, text, or other form of verification code. The target user interacts with the personalized verification code, such as entering the verification code text, clicking a specific area in the image, etc. The target user client captures the user's interactive behavior, including mouse clicks, keyboard input, touch operations, etc., to obtain the user's cross-modal interaction response data. The target user client encrypts the user's cross-modal interaction response data and uses the encrypted session key to ensure the confidentiality of the data. The encrypted user cross-modal interaction response data is sent back to the preset verification code server. The preset verification code server receives the encrypted user cross-modal interaction response data and decrypts the data using the same encrypted session key.

[0027] Step S5: Perform cognitive delay pattern recognition on the user cross-modal interaction response data to obtain a user cognitive delay index; and grant or deny access rights to the target user according to the user cognitive delay index.

[0028] Specifically, for example, cognitive delay pattern recognition algorithms and models, such as machine learning or statistical analysis methods, can be used to analyze and process the user's cross-modal interaction response data. According to the user's cross-modal interaction response data, the user's cognitive delay pattern is identified, that is, the time difference shown by the user when crossing different interaction modes. The user's cognitive delay index is calculated by analyzing the time interval, reaction speed and other indicators in the user's cross-modal interaction response data. The cognitive delay index can represent the difference in the user's cognitive processing speed between different interaction modes, and a higher index indicates that the user's cognitive delay is larger. According to the user's cognitive delay index, a decision to grant or deny access rights is made. According to a pre-set policy or rule, such as setting a threshold, it is determined whether the user's cognitive delay index exceeds the set threshold. If the user's cognitive delay index is less than the threshold, the verification code server may deny the user's access rights, believing that the user may be an automated script or a robot. If the user's cognitive delay index is within an acceptable range, the verification code server can grant the user's access rights, believing that the user is a real legitimate user.

[0029] The present invention ensures that the communication between the target user client and the verification code server is encrypted by constructing a secure communication tunnel, thereby preventing man-in-the-middle attacks and data leakage. Through the secure communication tunnel, the two parties successfully establish a connection and exchange key data, ensuring that the data transmission in the subsequent steps is encrypted and secure. By analyzing the human-computer interaction mode of the target user client, the difference between a real user and an automated attack can be distinguished. This helps to distinguish malicious attackers from legitimate users and improve the security of the verification code. Human-computer interaction mode analysis can be used to identify whether the user is a visually impaired user. By identifying the visually impaired user, a verification code experience adapted to their special needs can be provided, such as using an audio verification code or providing auxiliary functions to improve the user experience and accessibility of the visually impaired user. By generating personalized verification code data according to user interaction feature data, the complexity and randomness of the verification code can be increased, making it difficult to be cracked by the algorithm of the automated attack. This helps to improve the security of the verification code and prevent the threat of malicious attacks and black and gray industries. Personalized verification code data is generated according to the user's interactive feature data, and verification codes suitable for their special needs are provided for normal users and visually impaired users respectively. For example, a regular image verification code can be generated for normal users, while an audio verification code or other accessibility options can be generated for visually impaired users. This helps meet the needs of different user groups and improve user experience and accessibility. By using encrypted session key data to send personalized verification code data to the target user client, it is ensured that the data is not stolen or tampered with during transmission, and the security of the verification code is maintained. By capturing the multi-dimensional interactive behavior of the target user client, a variety of interactive response data of the user in the process of identifying the verification code, such as clicks, slides, touches, etc., can be obtained. This helps to further analyze the user's cognitive patterns and behavioral characteristics and improve the accuracy and security of the verification code. By performing cognitive delay pattern recognition on the user's cross-modal interactive response data, the user's cognitive delay in the process of completing the verification code can be analyzed. This helps to determine whether the user is a legitimate user and whether there is an automated attack behavior. By identifying the user's cognitive delay, the security of the verification code and the ability to defend against automated attacks can be further improved. According to the user's cognitive delay index, the target user can be granted or denied access rights. When the user's cognitive delay index meets the range of legitimate users, the access right is granted; otherwise, the access request is denied. This helps to effectively prevent access by automated attackers and improve the security of the system. In summary, the present invention can improve the user experience and accessibility of visually impaired users, and can also increase the complexity and randomness of the verification code, making it difficult to be cracked by the algorithm of automated attacks, thereby improving the security of the verification code and preventing malicious attacks and threats from the black and gray industries.

[0030] Preferably, step S1 comprises the following steps:

[0031] Step S11: collecting identity authentication request data from the target user client to obtain original identity authentication data;

[0032] Specifically, for example, an authentication request data collection module can be introduced into the client application of the target user. Use a network packet capture tool (such as Wireshark) to monitor the communication traffic between the target user's client and the server. During the authentication request process, capture the request data packet sent by the client. Extract data related to authentication from the request data packet, such as user credentials, device information, etc. Save the extracted authentication request data as authentication raw data, ready for subsequent processing and analysis.

[0033] Step S12: Performing a validity check on the original identity authentication data to obtain identity validity judgment result data;

[0034] Specifically, for example, a server-side identity authentication system or module can be used to receive and parse the target user's original identity authentication data. The algorithm or rule for legitimacy verification is executed for the original identity authentication data. The user credentials are verified, such as checking the correctness of the user name and password, checking the validity of the token, etc. The device information is checked, such as verifying the legitimacy of the device, checking the consistency of the IP address, etc. Based on the result of the legitimacy verification, the identity legitimacy judgment result data is generated to indicate whether the identity authentication has passed or failed.

[0035] Step S13: Based on the identity legitimacy judgment result data, a preset verification code server is used to generate a random encryption seed to obtain initial encryption seed data;

[0036] Specifically, for example, a programming language (such as Python) can be used to write a code that communicates with the verification code server. Extract necessary information, such as a user ID or other identifier, from the identity legitimacy judgment result data. Send a request to the preset verification code server through a network request library (such as Requests) to request the generation of a random encrypted seed. The request includes necessary parameters such as the identity legitimacy judgment result data and the user identifier. After receiving the request, the verification code server generates a random encrypted seed based on the received data. The verification code server returns the generated initial encrypted seed data as a response.

[0037] Step S14: performing key parameter negotiation between the target user client and the preset verification code server to obtain key negotiation parameter data;

[0038] Specifically, for example, a key parameter negotiation module can be introduced in the target user client application to communicate with a preset verification code server. Use a network request library (such as Requests) to send a key parameter negotiation request to the verification code server. The request includes necessary parameters such as initial encryption seed data and user identifier. After receiving the request, the verification code server uses the received initial encryption seed data and other related parameters to negotiate key parameters. During the key parameter negotiation process, the verification code server and the target user client exchange data and calculations to generate shared key negotiation parameters. The verification code server returns the generated key negotiation parameter data to the target user client as a response.

[0039] Step S15: performing encryption seed mixing and asymmetric key pair generation on the target user client and the preset verification code server based on the initial encryption seed data and the key negotiation parameter data, respectively, to obtain mixed encryption seed data and asymmetric key pair data;

[0040] Specifically, for example, an encryption algorithm library, such as OpenSSL, can be introduced into the target user client application and the preset verification code server. Using the initial encryption seed data and the key negotiation parameter data as input, the hybrid encryption seed data is generated by the encryption algorithm. In the process of generating the hybrid encryption seed data, the initial encryption seed data can be encrypted using an encryption algorithm (such as AES) and then mixed with the key negotiation parameter data. At the same time, asymmetric key pair data, including a public key and a private key, is generated by the encryption algorithm. The generated asymmetric key pair data can be generated using an asymmetric encryption algorithm (such as RSA).

[0041] Step S16: Perform security key exchange between the target user client and the preset verification code server according to the mixed encryption seed data and the asymmetric key pair data to obtain encrypted session key data.

[0042] Specifically, for example, a key exchange protocol, such as the Diffie-Hellman protocol, can be introduced into the target user client application and the preset verification code server. The target user client and the verification code server respectively use asymmetric key pairs to perform key exchange on the public key in the data. During the key exchange process, the mixed encrypted seed data is used as input to perform the calculation steps of the key exchange protocol. After the key exchange is completed, the target user client and the verification code server respectively calculate and generate encrypted session key data. The encrypted session key data can use a hash function (such as SHA-256) to process the key exchange result to generate a secure session key.

[0043] The present invention verifies the user identity by collecting identity authentication request data, ensuring that only authenticated users can continue to access the system or perform sensitive operations. This helps to improve the security of the system and prevent unauthorized access. The accuracy and integrity of the original identity authentication data can be verified by legitimacy verification, ensuring that the input data meets the specifications and requirements of the system. This helps to prevent malicious users from illegally accessing or deceiving the system by forging or tampering with data. The security and randomness of encryption can be increased by generating random encryption seeds using identity legitimacy judgment result data, providing a solid foundation for subsequent key generation and secure exchange. This helps to prevent the encryption process from being cracked or predicted, and improves the security of the system. Through key parameter negotiation, it is ensured that the target user client and the verification code server can securely negotiate shared key parameters for subsequent encryption and decryption operations. This helps to ensure the confidentiality and integrity of communication and prevent third parties from intercepting or tampering with communication content. By mixing encryption seed data and asymmetric key pair generation, the randomness and security of the encryption seed can be ensured, and an asymmetric key pair for encryption and decryption can be generated. This helps to provide secure encrypted communication and protect the confidentiality and integrity of user data. Secure key exchange is based on hybrid encrypted seed data and asymmetric key pairs, ensuring that communications between the target user client and the verification code server are encrypted and decrypted using secure keys. This helps protect the confidentiality of the communication content, prevent eavesdropping or tampering by third parties, and ensure the secure transmission of data.

[0044] Preferably, step S2 comprises the following steps:

[0045] Step S21: collecting historical interaction behavior data of the target user client based on the encrypted session key data to obtain a historical interaction behavior data set;

[0046] Specifically, for example, a data collection module or API can be introduced in the target user client application to capture the user's historical interaction behavior data. In the data collection module, the collected historical interaction behavior data is encrypted and protected using encrypted session key data. The data collection module collects various interaction behavior data of the user in the application, such as clicks, scrolls, inputs, etc., according to the required data type and range. The collected historical interaction behavior data may include relevant information such as timestamps, operation types, trigger elements, etc., for subsequent verification code modality classification.

[0047] Step S22: classify the historical interaction behavior dataset by verification code modality to obtain a sub-modal verification code interaction dataset;

[0048] Specifically, for example, a machine learning or deep learning framework, such as Scikit-learn or TensorFlow, can be introduced for the verification code modality classification task. Prepare a labeled training data set, which contains the interaction behavior data and corresponding labels of various verification code modalities. Use the training data set to train a verification code modality classification model, and you can choose an algorithm suitable for the task, such as a support vector machine (SVM) or a convolutional neural network (CNN). Use the trained model to classify and predict the collected historical interaction behavior data set to obtain a verification code interaction data set of different modalities. During the classification prediction process, use the encrypted session key data to decrypt the historical interaction behavior data set to obtain the original data for classification.

[0049] Step S23: Perform statistical pattern mining on the sub-modal verification code interaction data set to obtain user interaction pattern distribution data;

[0050] Specifically, for example, data analysis and mining tools such as NumPy, Pandas, and Scikit-learn in Python can be introduced. The sub-modal captcha interaction dataset is loaded into the data analysis environment for further processing and analysis. A series of statistical analyses are performed on the sub-modal captcha interaction dataset, such as calculating the average number of user interactions, interaction time intervals, and distribution of interaction element types. Data analysis tools and techniques, such as cluster analysis, frequent pattern mining, or association rule mining, are used to identify and extract user interaction patterns. The analysis results will include the distribution of different interaction patterns, such as common user interaction behavior sequences, frequently occurring interaction element combinations, etc.

[0051] Step S24: performing a human-computer interaction pattern analysis on the target user client according to the user interaction pattern distribution data and the sub-modal verification code interaction data set to obtain user interaction feature data.

[0052] Specifically, for example, user interaction pattern distribution data and sub-modal verification code interaction data sets can be used in combination with data analysis tools and machine learning methods, such as Scikit-learn, XGBoost or deep learning frameworks in Python. User interaction pattern distribution data and sub-modal verification code interaction data sets are subjected to feature engineering to extract meaningful features to represent the user's interactive behavior. Features that can be considered for extraction include the number of interactions, interaction time intervals, types of interactive elements, interaction order, interaction methods, etc. Machine learning algorithms, such as classification algorithms (such as decision trees, support vector machines) or deep learning models (such as recurrent neural networks, convolutional neural networks), are used to train and model the extracted features. Using the trained model, the interactive behavior of the target user client is analyzed and predicted to obtain user interaction feature data.

[0053] The present invention can obtain the behavior patterns and habits of users in past interaction processes through the collection of historical interaction behavior data. This helps to understand the normal behavior and habits of users, and provides a data basis for subsequent interaction pattern analysis and user feature extraction. Through verification code modal classification, the interaction behaviors in the historical interaction behavior data set can be classified according to different modalities. This helps to distinguish and analyze different types of interaction behaviors, and provides a more refined data basis for subsequent pattern mining and analysis. Through statistical pattern mining, the user's interaction pattern and behavior regularity can be identified from the sub-modal verification code interaction data set. This helps to discover common interaction behavior patterns and abnormal behavior patterns, and provides a basis for subsequent user interaction pattern analysis and identity recognition. Through human-computer interaction pattern analysis, the user's interaction feature data can be extracted based on the user interaction pattern distribution data and the sub-modal verification code interaction data set. This helps to establish a model of the user's behavior pattern and interaction characteristics, and provides a basis for subsequent identity recognition and verification code personalized generation.

[0054] Preferably, step S24 includes the following steps:

[0055] Step S241: constructing an interactive behavior knowledge graph for the user interaction pattern distribution data to obtain an interactive behavior knowledge ontology graph;

[0056] Specifically, for example, knowledge graph construction tools and technologies such as Neo4j, OWL or RDF can be introduced. The user interaction pattern distribution data is loaded into the knowledge graph construction tool as the initial data of the knowledge graph. According to the interactive behavior characteristics in the interactive pattern distribution data, the types of entities (nodes) and relationships (edges) of the knowledge graph are defined. According to the relevance of the interactive behavior, the association relationship between nodes is established, and the corresponding attributes are set. The query language or API of the knowledge graph construction tool is used to verify, expand and optimize the constructed knowledge graph.

[0057] Step S242: performing interaction sequence mining on the sub-modal verification code interaction data set to obtain a typical user interaction mode sequence;

[0058] Specifically, for example, sequence mining tools and algorithms, such as sequence pattern mining or sequence clustering algorithms, can be introduced. The sub-modal verification code interaction dataset is loaded into the sequence mining tool for pattern discovery and analysis. According to information such as timestamps, the interaction dataset is sorted and organized to form the user's interaction sequence. The interaction sequence is analyzed using a sequence mining algorithm, such as the Apriori algorithm or sequence clustering algorithm based on frequent pattern mining, to discover the user's typical interaction pattern sequence. Based on the output results of the algorithm, the user's typical interaction pattern sequence and its frequency or other related metrics are determined.

[0059] Step S243: performing temporal fusion coding on the interactive behavior knowledge ontology graph and the user's typical interactive mode sequence to obtain dynamic interactive cognitive fingerprint data;

[0060] Specifically, for example, the interactive behavior knowledge ontology map and the user's typical interaction pattern sequence can be loaded into a data processing environment, such as a data processing library and algorithm in Python. According to the timeline, the user's typical interaction pattern sequence and the interactive behavior knowledge ontology map are temporally fused and encoded. In the fusion encoding process, the user's typical interaction pattern sequence is mapped to the entities and relationships in the map according to the relationships and attributes in the interactive behavior knowledge ontology map. Encoding techniques such as one-hot encoding, sequence encoding or graph embedding can be used to convert the temporally fused data into a form that can be processed by a computer. In the encoding process, other features such as time interval, frequency, etc. can be considered to be added to enrich the expressive power of dynamic interactive cognitive fingerprint data.

[0061] Step S244: Decomposing the dynamic interaction cognitive fingerprint data into heterogeneous interaction elements to obtain a user multivariate interaction feature set;

[0062] Specifically, for example, data processing and feature extraction tools such as NumPy, Pandas, and Scikit-learn in Python can be introduced. The dynamic interaction cognitive fingerprint data is loaded into the data processing environment for further decomposition and analysis. Decomposition processing is performed based on the heterogeneous interaction elements in the dynamic interaction cognitive fingerprint data, such as entities and relationships in the graph, and interaction elements in the user's typical interaction pattern sequence. For entities and relationships in the graph, graph embedding technology can be used to convert them into vector representations, such as using algorithms such as Graph2Vec or Node2Vec. For interaction elements in the user's typical interaction pattern sequence, sequence encoding technology, such as Bag-of-Words, TF-IDF (TermFrequency-Inverse Document Frequency), or sequence embedding methods (such as Word2Vec or GloVe) can be used for encoding. The decomposed feature set is integrated and merged to obtain a user multivariate interaction feature set.

[0063] Step S245: Perform heterogeneous fusion encoding on the user multivariate interaction feature set to obtain user interaction feature data.

[0064] Specifically, for example, feature fusion and encoding techniques such as feature selection, feature weighting and feature combination can be introduced. The user multivariate interaction feature set is loaded into the data processing environment for feature fusion and encoding. For heterogeneous feature sets, feature selection can be considered to select features related to the target task for retention. Feature selection algorithms such as mutual information, correlation coefficient or model-based methods can be used. For the retained features, feature weighting can be performed according to their importance to increase the influence of features with greater contributions. It is possible to consider combining different types of features to generate new feature representations. For example, the graph element representation and the sequence element encoding are connected or spliced. For the fused feature set, encoding techniques such as unique hot encoding, label encoding or normalization can be applied to convert it into a computer-processable form. Finally, the user interaction feature data is obtained, that is, the feature representation after heterogeneous fusion encoding.

[0065] The present invention converts user interaction pattern distribution data into a structured knowledge graph by constructing an interactive behavior knowledge graph. This helps to semantically model and analyze the user's interactive behavior and provides a deeper ability to understand user behavior. Through the knowledge graph, the association and dependency between different interactive behaviors can be identified, providing a basis for subsequent interactive sequence mining and cognitive fingerprint generation. Through interactive sequence mining, the user's typical interactive pattern sequence can be extracted from the sub-modal verification code interactive data set. This helps to discover the common interactive behavior patterns and habits of users and better understand the user's behavioral laws. Temporal fusion coding fuses the interactive behavior knowledge ontology graph and the user's typical interactive pattern sequence according to the timeline. This helps to capture the temporal evolution characteristics of user interactive behavior and better understand the dynamics of user interactive behavior. Through dynamic interactive cognitive fingerprint data, the user's interactive features in different time periods can be extracted, providing more comprehensive information for the subsequent generation of multivariate interactive feature sets. Heterogeneous interactive element decomposition decomposes the dynamic interactive cognitive fingerprint data to extract the heterogeneous interactive elements therein. This helps to identify and separate different types of interactive features, such as text input, mouse clicks, sliding, etc. Heterogeneous fusion coding fuses and encodes the user's multivariate interaction feature set, integrates different types of interaction features, and generates comprehensive interaction feature data for the user. This helps to unify the representation and processing of different types of interaction features and provide a comprehensive and integrated user feature description.

[0066] Preferably, step S3 comprises the following steps:

[0067] Step S31: embedding and mapping the user interaction feature data to obtain a low-dimensional interaction habit semantic vector;

[0068] Specifically, for example, feature embedding techniques such as principal component analysis (PCA), autoencoder or t-SNE can be introduced. User interaction feature data is loaded into a data processing environment for embedding mapping. For high-dimensional user interaction feature data, dimensionality reduction techniques such as principal component analysis (PCA) can be applied to map it to a low-dimensional space. PCA finds the main variance in the data through linear transformation and projects it into a new low-dimensional space. Using feature embedding methods such as autoencoders, potential feature representations can be learned to map high-dimensional features to low-dimensional space. Autoencoders are neural network models that train the network to reconstruct the input data as much as possible after the encoding and decoding process, thereby learning the potential feature representation of the data. t-SNE (t-distributed Stochastic Neighbor Embedding) is a nonlinear dimensionality reduction technique that can map high-dimensional data to a low-dimensional space while retaining the local relationship between samples. It can help discover clustering structures and relationships between categories in the data. By selecting a suitable embedding method and adjusting relevant parameters, user interaction feature data is mapped to a low-dimensional interaction habit semantic vector.

[0069] Step S32: classifying the interaction mode of the corresponding target user according to the low-dimensional interaction habit semantic vector to obtain user interaction mode label data;

[0070] Specifically, for example, a machine learning classification algorithm such as a support vector machine (SVM), a random forest, or a deep learning model can be introduced. The low-dimensional interaction habit semantic vector and the corresponding user interaction modality label data are loaded into the machine learning environment. The data set is divided into a training set and a test set for model training and evaluation. The model is trained using the training set according to the selected classification algorithm. For example, a classification model can be established using a support vector machine (SVM) to distinguish different interaction modalities by finding an optimal hyperplane. The trained model is used to predict and classify the low-dimensional interaction habit semantic vectors in the test set. The model will classify the user's interaction habit semantic vector and attribute it to the corresponding interaction modality label. To evaluate the performance of the model, indicators such as accuracy, precision, recall, and F1 score can be used for evaluation. Finally, the user interaction modality label data is obtained, that is, the result of classifying the interaction modality of the target user according to the low-dimensional interaction habit semantic vector.

[0071] Step S33: when the user interaction modality label data is normal user data, a personalized visual verification code is generated for the target user client to obtain personalized visual verification code data;

[0072] Specifically, for example, an image processing library OpenCV and a trained image generation model can be prepared. When the user interaction modality label data is obtained, the user is determined to be a normal user based on the label. The interaction habit semantic vector and other relevant information of the target user are used on the preset verification code server to generate a personalized verification code image through the image generation model. Image processing techniques are applied, such as adding distortion, noise, interference lines, etc., to increase the security and readability of the verification code.

[0073] Step S34: when the user interaction mode label data is visually impaired user data, a personalized voice verification code is generated for the target user client to obtain personalized voice verification code data;

[0074] Specifically, for example, a speech synthesis engine and a speech recognition library may be prepared. When the user interaction modality label data is obtained, the user is determined to be a visually impaired user based on the label. The speech synthesis engine and the speech recognition library are used to generate a speech verification code on a preset verification code server. The speech synthesis engine is used to generate a personalized voice verification code using the interaction habit semantic vector and other relevant information of the target user. Speech processing technologies, such as noise reduction and audio enhancement, are applied to improve the comprehensibility and recognizability of the voice verification code.

[0075] Step S35: using the personalized visual verification code data or the personalized voice verification code data as the personalized verification code data.

[0076] Specifically, for example, personalized visual verification code data or personalized voice verification code data may be used as personalized verification code data.

[0077] The present invention can convert the original user interaction feature data into a low-dimensional interaction habit semantic vector by embedding mapping. This helps to reduce the dimension of high-dimensional interaction features, reduce the complexity and redundancy of data, and extract more representative and potential semantic feature representations. Through interaction modality classification, the target user can be modally classified according to the low-dimensional interaction habit semantic vector, that is, it is determined whether the user belongs to a normal user or a visually impaired user. This helps to identify the user's interaction feature pattern, distinguish different types of users, and provide a decision basis for subsequent personalized verification code generation. For normal users, personalized visual verification code generation is performed, and a visual verification code suitable for their use is generated according to the user's interaction modality and interaction characteristics. The personalized visual verification code can be designed according to the user's preferences and habits to improve the readability and ease of use of the verification code. This helps to enhance the user's verification code recognition and verification experience, reduce the possibility of incorrect input and authentication failure, and improve security and user satisfaction. For visually impaired users, personalized voice verification code generation is performed, and a voice verification code suitable for their use is generated according to the user's interaction modality and interaction characteristics. The personalized voice verification code can convert the verification code content into voice form, which is convenient for visually impaired users to recognize and verify through hearing. This helps visually impaired users to participate in the verification code verification process, increases their user rights and accessibility, and improves security and user experience. By using personalized visual verification code data or personalized voice verification code data as personalized verification code data, a suitable verification code form can be provided according to the user's interaction mode and needs. This helps to provide a personalized verification code verification method to meet the needs and preferences of different users. The present invention can generate personalized verification code data according to the user's interaction mode and needs, improve the readability, ease of use and accessibility of the verification code, and enhance the user's security and verification experience.

[0078] Preferably, step S33 includes the following steps:

[0079] Step S331: when the user interaction modality label data is normal user data, obtaining the user interaction behavior log data and the user account risk entropy value score;

[0080] Specifically, for example, when the user interaction modality label data determines that the user is a normal user. Through appropriate network requests or database queries, the preset verification code server obtains the user interaction behavior log data related to the target user. The user's unique identifier, such as user ID or user name, can be obtained from the user account system. According to the preset risk quantification standard or expert advice, the verification code server performs a risk assessment on the user account. The user account's historical data, behavior patterns, device information and other factors can be analyzed, and risk quantification can be performed according to the preset risk quantification standard to obtain the user account risk entropy value score.

[0081] Step S332: performing verification code difficulty rating on the user account risk entropy value score according to a preset progressive complexity mapping matrix to obtain expected verification code complexity level data;

[0082] Specifically, for example, a preset progressive complexity mapping matrix may be prepared. The matrix maps different user account risk entropy scores to corresponding verification code difficulty ratings. The corresponding verification code difficulty rating is searched in the mapping matrix according to the user account risk entropy score.

[0083] Step S333: performing abnormal pattern detection and trustworthiness evaluation on the user interaction behavior log data to obtain a verification trustworthiness index;

[0084] Specifically, for example, an anomaly detection algorithm or statistical method can be used to analyze the user interaction behavior log data to find abnormal behaviors that do not conform to normal behavior patterns. According to the degree and frequency of the abnormal pattern, the user's historical behavior data and the comparison with normal user behavior are comprehensively considered to calculate the user's trust score or trust index. The trust score is converted into a verification trust index, which reflects the credibility of the user's interactive behavior. A preset mapping function or threshold can be used to map the trust score to the verification trust index.

[0085] Step S334: evaluating the situational awareness and cognitive ability of the target user according to the user interaction behavior log data to obtain cognitive preference index data;

[0086] Specifically, for example, the user's behavior patterns and preferences in the account login verification scenario can be analyzed based on the user interaction behavior log data, such as identifying the input order, interaction time interval, etc. The accuracy of the user in the process of entering the verification code, such as the number of input errors, correction behavior, etc., is analyzed to evaluate the user's understanding and accuracy of the verification code. Features can be extracted and evaluated using machine learning algorithms, rule engines, or expert-defined rules. The results of the situational awareness and cognitive ability assessment are converted into cognitive preference index data, which reflects the user's cognitive preferences in the account login verification scenario. The evaluation results can be mapped to cognitive preference index data according to a preset mapping function or threshold.

[0087] Step S335: Generate a personalized visual verification code for the target user client according to the expected verification code complexity level data, the verification trust index, and the cognitive preference index data to obtain personalized visual verification code data.

[0088] Specifically, for example, the expected verification code complexity level, such as simple, medium or complex level, can be determined according to the expected verification code complexity level data according to system requirements or user configuration. According to the user's trust evaluation or historical behavior analysis, the verification trust index is calculated, which reflects the system's trust in the target user. Select appropriate tools or methods to generate personalized visual verification codes. Image processing libraries, machine learning models, or custom algorithms can be used. Image processing libraries or deep learning models can be used to extract image features that are compatible with the expected complexity level and cognitive preference index data. Based on the extracted image features, a personalized verification code image is generated. Methods such as generative adversarial networks (GANs), style transfer algorithms, or image transformation techniques can be used. According to the user's cognitive preference index data, the generated verification code image is style customized, such as adjustments in color, font, shape, etc.

[0089] The present invention helps to understand the user's behavior pattern and account security risk by obtaining user interaction behavior log data and user account risk entropy value scoring. By analyzing the interaction behavior log data, the user's routine operations and habitual behaviors can be understood, providing a reference for the subsequent verification code generation. At the same time, the user account risk entropy value scoring can provide information about the risk level of the account, which is used for the difficulty assessment and security enhancement of personalized verification codes. By rating the difficulty of the verification code based on the user account risk entropy value scoring, a verification code of corresponding difficulty can be generated for the user according to the account risk level. The preset progressive complexity mapping matrix can associate verification code difficulties of different levels with it according to the risk entropy value scoring to provide a verification code difficulty suitable for the user's risk level. This helps to improve the security and defensiveness of the verification code, provide users with more challenging verification codes according to the account risk situation, and increase the difficulty of abnormal visitors. By performing abnormal pattern detection and trust evaluation on the user interaction behavior log data, abnormal patterns and potential risk behaviors in user behavior can be identified. This helps to improve the system's trust evaluation of users and determine the reliability and integrity of users. The verification trust index can be used as a basis for generating personalized verification codes, ensuring that the generation of verification codes is based on credible user behavior and reducing interference and misjudgment of normal users. By performing situational awareness and cognitive ability assessment on user interaction behavior log data, users' behavioral preferences and cognitive abilities in different scenarios can be understood. This helps to generate personalized verification codes and provide verification codes that are more suitable for users based on their cognitive abilities and preferences. Cognitive preference index data can be used to determine the readability, comprehensibility and user-friendliness of verification codes, and improve the usability and user satisfaction of verification codes. Based on the expected verification code complexity level data, verification trust index and cognitive preference index data, personalized visual verification code generation can generate more adaptable verification codes based on the user's risk level, trust and cognitive preference. The generated personalized visual verification code can combine the user's specific needs and abilities to improve the security, readability and user experience of the verification code. This helps to enhance the security of the system, prevent malicious attacks and illegal access, and improve user satisfaction and trust in the system.

[0090] Preferably, step S335 includes the following steps:

[0091] Step S3351: constructing a verification code generation seed library based on the expected verification code complexity level data to obtain a graded visual element seed library;

[0092] Specifically, for example, the expected verification code complexity level, such as simple, medium or complex level, can be determined according to the expected verification code complexity level data according to system requirements or user configuration. According to the expected verification code complexity level data, the visual elements are classified according to different complexity levels, such as simple elements, medium elements and complex elements. A corresponding seed library is generated for the visual elements of each complexity level. For example, a simple element seed library: a simple graphic, number or letter is selected as a seed element, and a corresponding seed library is generated.

[0093] Step S3352: Determine the interference element injection intensity and interference mode of the hierarchical visual element seed library according to the verification trust index, and obtain interference element library data and superposition strategy data;

[0094] Specifically, for example, appropriate tools or methods can be selected to inject interference elements into the verification code generation. Image processing libraries, graphics generation algorithms, or hybrid models, etc. can be used. According to the level of the verification trust index, the injection intensity of the interference elements is adjusted. For a higher trust index, fewer interference elements are injected; for a lower trust index, more interference elements are injected. Different interference patterns are selected according to different ranges of the verification trust index. For example, for users with a high trust index, a slight interference pattern is selected, such as adding slight image noise; for users with a low trust index, a more complex interference pattern is selected, such as distortion, occlusion, or adding complex image distractors.

[0095] Step S3353: constructing a user visual cognitive preference model according to the cognitive preference index data to obtain a user visual cognitive preference model;

[0096] Specifically, for example, an appropriate model type can be selected to construct a user visual cognitive preference model. Machine learning algorithms, artificial neural networks, or rule-based methods can be used. According to the feature description in the cognitive preference index data, a suitable feature extraction method is selected. For example, if the cognitive preference index involves color preference, color features can be extracted; if it involves shape preference, shape features can be extracted. The user visual cognitive preference model is trained using the extracted features and the corresponding cognitive preference index data. Model training can be performed using supervised learning or unsupervised learning methods.

[0097] Step S3354: inputting the hierarchical visual element seed library into the visual cognition preference model to generate initial verification code screen data;

[0098] Specifically, for example, an appropriate tool or method can be selected to input the hierarchical visual element seed library into the visual cognitive preference model and generate initial verification code screen data. Generative adversarial networks (GANs), image synthesis algorithms, or rule-based methods, etc. can be used. The hierarchical visual element seed library is used as input and generated through the visual cognitive preference model. The following is an example: Feature matching: Match the elements in the hierarchical visual element seed library with the features in the user's visual cognitive preference model to determine the elements that match the user's preferences. Generate initial verification code screen data: Generate initial verification code screen data based on the matched elements. According to the user's visual cognitive preferences, the position, size, color and other attributes of the element can be adjusted to generate initial verification code screen data that meets the user's preferences.

[0099] Step S3355: injecting adversarial generated noise into the initial verification code screen data according to the interference element library data to obtain adversarial visual noise layer data;

[0100] Specifically, for example, appropriate tools or methods can be selected to inject noise against the initial verification code screen data. Image processing libraries, generative adversarial networks (GANs) or hybrid models, etc. can be used. Suitable interference elements are selected from the interference element library, and can be selected according to predefined injection intensity and mode. Using the selected interference elements, the noise generated against is injected into the initial verification code screen data. Injection can include operations such as superposition, coverage, distortion, transparency adjustment, so that noise is fused with the initial verification code screen data.

[0101] Step S3356: superimpose the adversarial visual noise layer data onto the initial verification code screen data according to the superposition strategy data to obtain personalized visual verification code data.

[0102] Specifically, for example, an appropriate tool or method can be selected to superimpose the adversarial visual noise layer data onto the initial verification code screen data. An image processing library, a synthesis algorithm, or a hybrid model, etc. can be used. According to the order defined in the superposition strategy data, the adversarial visual noise layer data is superimposed layer by layer onto the initial verification code screen data. According to the method defined in the superposition strategy data, the superposition operation is performed. It can be overlay, mixing, fusion, etc. to ensure the reasonable superposition of the adversarial visual noise layer data and the initial verification code screen data.

[0103] The present invention helps to generate a verification code that meets the expected complexity level by constructing a verification code generation seed library based on the expected verification code complexity level data. The seed library contains visual elements of different levels, and appropriate element combinations can be selected as needed to generate personalized verification codes. This helps to improve the security and defensiveness of the verification code, and generates a verification code with a certain difficulty for the user according to the complexity level, increasing the difficulty of abnormal visitors. By determining the injection intensity and interference mode of the interference element in the graded visual element seed library according to the verification trust index, interference elements can be added to the verification code according to the user's trust evaluation result. The interference element library data contains interference elements of different types and strengths, and the superposition strategy data determines the superposition method of the interference elements. This helps to increase the complexity and diversity of the verification code, improve the resistance of the verification code, and prevent automated attacks and malicious cracking. According to the cognitive preference index data, a user visual cognition preference model is constructed to understand the user's personal preference in visual cognition. This helps to generate a verification code that meets the user's preference, and improves the readability and user experience of the verification code. By selecting appropriate visual elements and layouts for the verification code according to the user's visual cognition preference, the verification code can be made more in line with the user's cognitive habits and reduce the difficulty of the user's understanding of the verification code. By inputting the hierarchical visual element seed library into the visual cognitive preference model, the initial verification code screen data can be generated according to the user's preferences and cognitive model. This helps to generate more adaptable verification codes according to the personalized needs of users. The generated initial verification code screen data can provide more attractive and easy-to-recognize verification codes based on the user's preferences and cognitive abilities, and improve the usability and user satisfaction of the verification code. Injecting adversarial generated noise into the initial verification code screen data according to the interference element library data can add randomness and noise to the verification code and improve the anti-interference ability of the verification code. The adversarial visual noise layer data can increase the number and position of interference elements, making the verification code more challenging and resistant, and reducing the success rate of automated attacks and malicious cracking. The adversarial visual noise layer data is superimposed on the initial verification code screen data according to the superposition strategy data, and personalized visual verification code data can be generated. This helps to improve the uniqueness and discrimination of the verification code, making the verification code more difficult to crack and attack. The personalized visual verification code data combines the user's cognitive preferences, interference elements, and the superposition strategy of adversarial noise to provide each user with a unique verification code experience, enhancing the security and usability of the verification code.

[0104] Preferably, step S34 includes the following steps:

[0105] Step S341: when the user interaction modality label data is visually impaired user data, voice verification code interaction history data collection is performed on the target user client to obtain a voice verification code interaction history data set;

[0106] Specifically, for example, when it is determined that the user interaction modality label data is visually impaired user data, select appropriate tools or methods for collecting voice verification code interaction history data for the target user client. Specialized data collection software, mobile applications, or custom scripts can be used. Ensure the identity authentication of the target user client to obtain access rights. Monitor, record, or intercept the voice verification code interaction history data performed by the user on the target user client. This may include key information such as the playback of the voice verification code, the user's answer or input, and the verification result.

[0107] Step S342: extracting the voice segments that have successfully passed the verification from the voice verification code interaction history data set to obtain a historical successful voice verification data set;

[0108] Specifically, for example, an appropriate tool or method can be selected to extract successfully verified voice segments from the voice verification code interaction history dataset. An audio processing library, a speech recognition algorithm, or a custom script, etc. can be used. The voice segments in the voice verification code interaction history dataset are converted into text. This can be achieved using automatic speech recognition (ASR) technology. Based on the results of speech recognition, the voice segments that have successfully passed the verification are screened out. This can be done by matching predefined successful verification text or verification result labels.

[0109] Step S343: Perform voice verification code parameter analysis on the historical successful voice verification data set to obtain voice verification code parameter pattern data;

[0110] Specifically, for example, appropriate tools or methods can be selected to perform voice verification code parameter analysis on historical successful voice verification data sets. Audio processing libraries, voice feature extraction algorithms, or machine learning models, etc. can be used. Relevant features of voice verification codes are extracted from historical successful voice verification data sets. These features may include spectral features of speech, sound intensity, speech speed, stress position, etc. The extracted voice verification code features are analyzed to identify voice verification code parameter patterns for different users or in different scenarios. This can be achieved through statistical analysis, clustering algorithms, or pattern recognition techniques.

[0111] Step S344: Generate a personalized voice verification code for the target user client according to the voice verification code parameter mode data to obtain personalized voice verification code data.

[0112] Specifically, for example, an appropriate tool or method may be selected to generate a personalized voice verification code for the target user client according to the voice verification code parameter pattern data. A speech synthesis engine, a template matching algorithm, or a deep learning model, etc. may be used. According to the attributes or characteristics of the target user, the appropriate voice verification code parameter pattern data is matched. Using the selected voice verification code parameter pattern data, the speech synthesis engine is applied to generate a personalized voice verification code. This will synthesize a voice verification code audio with personalized characteristics according to the characteristics of the parameter pattern data.

[0113] The present invention can obtain the user's interactive behavior and pattern in the past voice verification code verification process by collecting voice verification code interaction history data from the client of the visually impaired user. This helps to understand the visually impaired user's usage habits and preferences for voice verification codes, and provides a basis for generating personalized voice verification codes. The collected voice verification code interaction history data set helps to analyze the user's behavior pattern, thereby providing a voice verification code verification experience that is more in line with the user's expectations and habits. By extracting the voice segments that have successfully passed the verification in the voice verification code interaction history data set, the user's effective voice samples in the past verification process can be obtained. These historical successful voice verification data sets can be used as the basis for training and model building. The extracted successful voice segments help to obtain high-quality voice features and patterns, thereby improving the accuracy and reliability of generating personalized voice verification codes. By performing voice verification code parameter analysis on the historical successful voice verification data set, common voice features and parameter patterns can be extracted. This helps to understand the voice recognition characteristics and personalized needs of visually impaired users. The voice verification code parameter pattern data can include parameters such as pitch, tone, and speech speed to build a user's voice feature model. These pattern data can be used to generate personalized voice verification codes that meet user characteristics and preferences, and improve the recognizability and user satisfaction of the verification code. Based on the voice verification code parameter pattern data, a personalized voice verification code is generated for the target user client, and a voice verification code that meets the user's characteristics and preferences can be generated. This helps to provide a personalized voice verification code verification experience and improve the comprehensibility and operability of the verification code for visually impaired users. Personalized voice verification code data can select the appropriate voice mode, tone and speaking speed according to the user's voice characteristics and preferences, making the verification code easier to identify and verify, and improving the convenience and success rate of users passing the verification.

[0114] Preferably, step S344 includes the following steps:

[0115] Step S3441: constructing a personalized voice element library according to the voice verification code parameter pattern data to obtain a hierarchical voice element library;

[0116] Specifically, for example, appropriate tools or methods can be selected to construct a personalized voice element library based on voice verification code parameter pattern data. A speech synthesis engine, audio editing software, or a custom script, etc. can be used. According to the attributes or characteristics of the target user, the appropriate voice verification code parameter pattern data is matched. Voice elements are extracted from the matched voice verification code parameter pattern data. This may include phonemes, tones, speech speed, etc. The extracted voice elements are combined and classified according to specific rules to construct a hierarchical voice element library. This will include various combinations and variants of voice elements to meet the needs of personalized voice verification codes.

[0117] Step S3442: collecting biological auditory feature data of the target user to obtain the user's auditory physiological data;

[0118] Specifically, for example, appropriate tools or methods may be selected for collecting biological auditory feature data of the target user. Specialized auditory physiological data collection equipment, biosensors, or mobile applications may be used. Ensure the identity authentication of the target user to obtain access rights. Collect the auditory physiological data of the target user using the selected tools or methods. This may include the measurement of auditory features such as auditory reaction time, auditory frequency perception range, and sound intensity perception.

[0119] Step S3443 performs personalized auditory cognition assessment on the target user according to the user's auditory physiological data to obtain a user auditory cognition assessment vector;

[0120] Specifically, for example, appropriate tools or methods can be selected for personalized auditory cognitive assessment based on the user's auditory physiological data. A specialized auditory cognitive assessment system, data analysis algorithm, or machine learning model can be used. The user's auditory physiological data is preprocessed, such as denoising, normalization, or feature extraction. Using the selected tool or method, the preprocessed auditory physiological data is compared and analyzed with the preset auditory cognitive model to obtain the user's auditory cognitive assessment results. This can include the evaluation of auditory cognitive features such as auditory preference, sound discrimination ability, and speech recognition accuracy. The proposed auditory cognitive model construction process can be referred to as follows: Collect auditory cognitive data from a large number of participants, including evaluation results in terms of auditory preference, sound discrimination ability, and speech recognition accuracy. These data can be collected through laboratory experiments, online questionnaires, or mobile applications. Preprocess the collected auditory cognitive data, such as removing outliers, normalization, feature extraction, etc. This helps to reduce the noise and uncertainty of the data and extract useful features. According to the domain knowledge and related research of auditory cognition, select appropriate features to represent auditory cognitive ability. This can include sound spectrum features, sound intensity features, time features, etc. According to the preset auditory cognition goals and data features, select the appropriate machine learning model or algorithm. Common models include support vector machine (SVM), decision tree, random forest, neural network, etc. Use the collected auditory cognition data as a training set to train and optimize the selected model. Use methods such as cross-validation to evaluate the trained auditory cognition model, such as calculating accuracy, recall, F1 value and other indicators. Based on the evaluation results, adjust and optimize the model to improve its performance and generalization ability.

[0121] Step S3444: constructing a voice verification code generation pool according to the user auditory cognitive evaluation vector and the preset voice verification code grammar rules to obtain a voice verification code prototype library;

[0122] Specifically, for example, appropriate tools or methods can be selected to construct a voice verification code generation pool according to the preset voice verification code grammar rules based on the user's auditory cognitive evaluation vector. Text processing tools, audio synthesis engines, or speech synthesis algorithms can be used. According to the preset voice verification code grammar rules, the structure, elements, and combination of the voice verification code are defined. This can include constraints on phonemes or syllables, arrangement rules for voice elements, etc. According to the user's auditory cognitive evaluation vector, suitable voice verification code elements or combinations are matched. This will select voice elements that meet the user's cognitive ability and preferences based on the user's auditory cognitive characteristics. The matched voice verification code elements or combinations are assembled and stored to build a voice verification code prototype library. This will include various voice verification code samples that conform to grammatical rules.

[0123] Step S3445: performing personalized voice distortion correction on the voice verification code prototype library according to the user's auditory physiological data to obtain a personalized voice verification code candidate set;

[0124] Specifically, for example, appropriate tools or methods can be selected to perform personalized voice distortion correction on the voice verification code prototype library according to the user's auditory physiological data. Speech signal processing algorithms, audio special effects processing tools or deep learning models can be used. Samples that match the user's auditory cognitive features are selected from the voice verification code prototype library as reference audio. These samples can be matched according to the user's auditory cognitive evaluation vector and the preset auditory cognitive model. Using the selected tools or methods, the reference audio is personalized for voice distortion correction. This can include operations such as audio spectrum adjustment, sound pitch change, noise addition or audio special effects application to adjust the sound characteristics and style of the audio to make it more in line with the user's personalized auditory cognitive features. Based on the reference audio after voice distortion correction, multiple personalized voice verification code candidate sets are generated. A series of voice verification code candidate sets that meet personalized auditory cognition can be generated by methods such as parameterized control, randomization operation or audio variant generation technology.

[0125] Step S3446: Randomly extract the personalized voice verification code candidate set to obtain personalized voice verification code data.

[0126] Specifically, for example, an appropriate tool or method may be selected to randomly extract a candidate set of personalized voice verification codes. A random number generation function, a random sampling algorithm, or a related library function in a programming language may be used. The number of candidate sets of personalized voice verification codes to be extracted is determined according to the requirements. Using the selected tool or method, the required number of samples is randomly extracted from the candidate set of personalized voice verification codes. Ensure that the extraction process is random and unbiased to ensure the representativeness and diversity of the samples.

[0127] The present invention constructs a personalized voice element library according to voice verification code parameter mode data, and can select appropriate voice elements according to the user's voice characteristics and preferences. This helps to generate a personalized voice verification code that meets the user's characteristics and preferences. The hierarchical voice element library can be combined and layered according to different voice characteristics and patterns to provide more diverse, rich and flexible voice verification code options, and enhance the user's recognition and verification experience. By collecting the biological auditory feature data of the target user, the user's auditory physiological attributes and characteristics can be obtained. This helps to understand the user's hearing ability, sensitivity and preference, and provides a basis for the generation of personalized voice verification codes. The user's auditory physiological data may include information on audio perception range, frequency resolution ability, sound intensity perception, etc. The collected data helps to customize the voice verification code to make it more suitable for the user's auditory characteristics, improve the recognizability of the verification code and the user's verification success rate. Based on the user's auditory physiological data, a personalized auditory cognitive assessment can understand the user's cognitive and perceptual ability of different sound features. This helps to determine the user's ability to understand and verify the voice verification code, and provides guidance for generating a personalized voice verification code. The user's auditory cognitive assessment vector may include the user's perception and preference of sound attributes such as pitch, tone, and speech speed. Through the analysis of the evaluation vector, the generation strategy of the personalized voice verification code can be adjusted according to the user's auditory cognitive characteristics, thereby improving the recognizability of the verification code and the user's verification experience. According to the user's auditory cognitive evaluation vector and the preset voice verification code grammar rules, the voice verification code generation pool can be constructed to generate voice verification codes that meet the user's auditory cognitive characteristics and preferences. This helps to provide personalized and diverse voice verification code options. The voice verification code prototype library is a plurality of voice verification code prototypes generated by combining the user's auditory cognitive evaluation vector and grammar rules. By constructing a generation pool, richer and more diverse voice verification code options can be provided, increasing the flexibility and accuracy of user identification and verification. Based on the user's auditory physiological data, personalized voice distortion correction of the voice verification code prototype library can adjust the sound characteristics of the voice elements to make them more consistent with the user's auditory perception and cognition. This helps to improve the recognizability of the personalized voice verification code and the user's verification experience. Personalized voice distortion correction can include adjusting sound attributes such as pitch, tone, and speech speed, as well as adding personalized sound effects and sound special effects. The corrected personalized voice verification code candidate set can better adapt to the user's auditory characteristics, improve the recognizability of the verification code and the user's verification success rate. By randomly extracting the candidate set of personalized voice verification codes, a personalized voice verification code that meets the user's needs can be selected from multiple candidates. This helps to provide a variety of voice verification code options and increase the flexibility of user identification and verification. The random extraction process ensures that each generated personalized voice verification code is unique, providing a certain degree of randomness and security.Users can select the verification code that best suits them from the extracted personalized voice verification code data for verification, thereby improving the reliability of verification and user satisfaction.

[0128] Preferably, step S5 comprises the following steps:

[0129] Step S51: extracting the interactive operation time series from the user cross-modal interactive response data to obtain the interactive operation time series;

[0130] Specifically, for example, appropriate tools or methods can be selected to extract the interactive operation time series from the user's cross-modal interactive response data. Time series processing algorithms, signal analysis methods, or related data processing library functions can be used. User cross-modal interactive response data includes operation records and corresponding timestamps between different modes. For example, the user's click operation on the touch screen, the input time of the voice command, the duration of eye gaze, etc. can be recorded. The collected interactive response data is preprocessed, such as removing outliers, data alignment, or timestamp correction. Ensure the accuracy and consistency of the data to make it suitable for subsequent time series extraction and analysis. Use the selected tool or method to extract the interactive operation time series from the preprocessed interactive response data. According to the data type and requirements of different modalities, sliding windows, event extraction algorithms, or specific pattern recognition methods can be used to convert interactive operation records into continuous time series.

[0131] Step S52: performing operation time loss consistency analysis on the target user according to the interactive operation time series to obtain a time consistency evaluation vector;

[0132] Specifically, for example, appropriate tools or methods can be selected to perform operation time loss consistency analysis on the interaction operation time series of a single target user. Time series analysis algorithms, statistical methods or related data analysis software can be used. According to specific task requirements and user experience indicators, the definition and measurement method of operation time loss consistency are clarified. For example, time consistency can be defined as whether the time interval between user cross-modal interaction operations is within a certain range. Use the selected tools or methods to perform operation time loss consistency analysis on the interaction operation time series of the target user. This can include calculating statistical indicators of operation time intervals, such as average time interval, maximum time interval, time interval change trend, etc. Based on the results of the analysis, a time consistency evaluation vector describing the target user's operation time loss consistency is generated.

[0133] Step S53: performing cognitive response time analysis on the user cross-modal interaction response data based on the interactive operation time series to obtain user cognitive response feature data;

[0134] Specifically, for example, appropriate tools or methods can be selected for cognitive response time analysis of user cross-modal interaction response data. Time series analysis algorithms, pattern recognition methods, or related data processing software, etc. can be used. Using the selected tools or methods, the user's cognitive response time is extracted from the interactive operation time series. The user's response time between different modes can be calculated based on the operation time interval, pattern recognition algorithm, or user behavior model, etc. Based on the extracted cognitive response time, data describing the user's cognitive response characteristics are generated. These characteristics may include average response time, change trend of response time, frequency of modal switching, etc.

[0135] Step S54: performing cognitive delay pattern recognition on the target user according to the time coherence evaluation vector and the user cognitive response characteristic data to obtain a user cognitive delay index;

[0136] Specifically, for example, appropriate tools or methods can be selected for cognitive delay pattern recognition of the temporal coherence evaluation vector and the user cognitive response feature data. Machine learning algorithms, pattern matching methods or related data analysis software, etc. can be used. Using the selected tools or methods, pattern recognition is performed on the temporal coherence evaluation vector and the user cognitive response feature data. The user's cognitive delay pattern can be identified through steps such as training models, feature extraction and algorithm matching. Based on the results of pattern recognition, an index describing the cognitive delay of the target user is generated. This index can reflect the user's level of cognitive delay in cross-modal interaction, such as the frequency of delay, the degree of delay, etc.

[0137] Step S55: comparing the user cognitive delay index with a preset cognitive time threshold, and when the user cognitive delay index is less than or equal to the preset cognitive time threshold, denying access to the target user and prompting the user that the account may be under robot attack;

[0138] Specifically, for example, a preset cognitive time threshold can be determined, which is used to determine whether the user's cognitive latency meets the requirements. The setting of the threshold can be based on specific security policies, user experience requirements or statistical analysis, etc. Compare the user's cognitive latency index with the preset cognitive time threshold. If the user's cognitive latency index is less than or equal to the preset cognitive time threshold, execute the next step; otherwise, execute step S56. According to security policies or system rules, the target user's access request is rejected. Measures such as preventing users from logging in, prohibiting the use of specific functions, or restricting access rights can be taken. Display a corresponding prompt message to the user to warn that the user account may be under robot attack. The prompt message may include an explanation of the cognitive latency index and the risk of robot attack, and provide further guidance or contact information to solve the problem.

[0139] Step S56: When the user cognitive delay index is greater than a preset cognitive time threshold, the target user is granted access rights.

[0140] Specifically, for example, the user's cognitive delay index can be compared with a preset cognitive time threshold. If the user's cognitive delay index is greater than the preset cognitive time threshold, the next step is executed; otherwise, the user access right is not granted. According to the security policy or system rules, the target user is granted the corresponding access right. The user can continue to log in, use functions or access related resources.

[0141] The present invention can understand the time interval and sequence of user operations between different modes by extracting the interactive operation time series of user cross-modal interactive response data. This helps to analyze the user's interactive behavior pattern and operation habits. The interactive operation time series provides the time dimension information of the user when performing multimodal interaction, which can be used for subsequent analysis and evaluation to improve the system's understanding and response capabilities to user behavior. By analyzing the interactive operation time series, the consistency of the user's operation time loss, that is, the consistency and stability of the user's operation time between different modes, can be evaluated. The time consistency evaluation vector provides a quantitative indicator of the user's operation time loss, which can be used to evaluate the user's behavioral consistency and stability. This helps to identify whether the user has abnormal operation behavior or mode, and detect possible robot attacks or fraudulent behaviors. Cognitive response time analysis based on the interactive operation time series can understand the user's cognitive response ability and speed in cross-modal interaction. Cognitive response time refers to the time interval from the user receiving the stimulus to the user making the corresponding cognitive behavior. By analyzing the user's cognitive response time, the user's perception and processing speed of different modes can be inferred. The user's cognitive response feature data provides information about the user's cognitive ability and reaction speed, which helps the subsequent cognitive delay pattern recognition and access authorization decision. By combining the temporal coherence evaluation vector and the user cognitive response feature data, the cognitive delay pattern of the target user, that is, the cognitive delay characteristics and patterns of the user between different modalities, can be identified. The cognitive delay index provides a quantitative indicator of the user's cognitive delay degree, which is used to measure the user's cognitive delay level in cross-modal interaction. This helps to identify whether the user has abnormal cognitive delay behavior and detect possible robot attacks or fraudulent behaviors. By comparing the user's cognitive delay index with the preset cognitive time threshold, it can be determined whether the user's cognitive delay level is in line with the normal cognitive response range. When the user's cognitive delay index is less than or equal to the preset cognitive time threshold, it may indicate that the user's cognitive response speed is abnormally fast, indicating that there may be a risk of robot attack. At this time, the target user is denied access and a warning is issued to the user to protect the security of the system and the user. This helps to improve the security of the system and the ability to resist robot attacks. When the user's cognitive delay index is greater than the preset cognitive time threshold, it may indicate that the user's cognitive response speed is slow, but still within the normal range. In this case, the system can grant access rights to the target user, allowing him to use system functions and resources normally. This helps to ensure user convenience and a good user experience, while avoiding unnecessary access restrictions and improving system availability and user satisfaction.

[0142] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0143] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A data security verification method for a click-type dynamic verification code, characterized in that: The following steps are involved: Step S1: construct a secure communication tunnel between the target user client and the preset verification code server to obtain encrypted session key data; Step S2: Analyze the human-computer interaction mode of the target user client based on the encrypted session key data to obtain user interaction feature data; Step S2 includes the following steps: Step S21: collecting historical interaction behavior data of the target user client based on the encrypted session key data to obtain a historical interaction behavior data set; Step S22: classify the historical interaction behavior dataset by verification code modality to obtain a sub-modal verification code interaction dataset; Step S23: Perform statistical pattern mining on the sub-modal verification code interaction data set to obtain user interaction pattern distribution data; Step S24: performing a human-computer interaction pattern analysis on the target user client according to the user interaction pattern distribution data and the sub-modal verification code interaction data set to obtain user interaction feature data; Step S3: Generate personalized verification data for the target user client according to the user interaction feature data to obtain personalized verification code data; Step S4: Sending the personalized verification code data to the target user client based on the encrypted session key data; capturing the target user client's multi-dimensional interactive behavior to obtain user cross-modal interactive response data; Step S5: performing cognitive delay pattern recognition on the user cross-modal interaction response data to obtain a user cognitive delay index; granting or denying access rights to the target user according to the user cognitive delay index; Step S5 includes the following steps: Step S51: extracting the interactive operation time series from the user cross-modal interactive response data to obtain the interactive operation time series; Step S52: performing operation time loss consistency analysis on the target user according to the interactive operation time series to obtain a time consistency evaluation vector; Step S53: performing cognitive response time analysis on the user cross-modal interaction response data based on the interactive operation time series to obtain user cognitive response feature data; Step S54: performing cognitive delay pattern recognition on the target user according to the time coherence evaluation vector and the user cognitive response characteristic data to obtain a user cognitive delay index; Step S55: comparing the user cognitive delay index with a preset cognitive time threshold, and when the user cognitive delay index is less than or equal to the preset cognitive time threshold, denying access to the target user and prompting the user account may be under robot attack; Step S56: When the user cognitive delay index is greater than a preset cognitive time threshold, the target user is granted access rights.

2. The data security verification method of the click-type dynamic verification code according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: collecting identity authentication request data from the target user client to obtain original identity authentication data; Step S12: Performing a validity check on the original identity authentication data to obtain identity validity judgment result data; Step S13: Based on the identity legitimacy judgment result data, a preset verification code server is used to generate a random encryption seed to obtain initial encryption seed data; Step S14: performing key parameter negotiation between the target user client and the preset verification code server to obtain key negotiation parameter data; Step S15: performing encryption seed mixing and asymmetric key pair generation on the target user client and the preset verification code server based on the initial encryption seed data and the key negotiation parameter data, respectively, to obtain mixed encryption seed data and asymmetric key pair data; Step S16: Perform security key exchange between the target user client and the preset verification code server according to the mixed encryption seed data and the asymmetric key pair data to obtain encrypted session key data.

3. The data security verification method of the click-type dynamic verification code according to claim 1 is characterized in that: Step S24 includes the following steps: Step S241: constructing an interactive behavior knowledge graph for the user interaction pattern distribution data to obtain an interactive behavior knowledge ontology graph; Step S242: performing interaction sequence mining on the sub-modal verification code interaction data set to obtain a typical user interaction mode sequence; Step S243: performing temporal fusion coding on the interactive behavior knowledge ontology graph and the user's typical interactive mode sequence to obtain dynamic interactive cognitive fingerprint data; Step S244: Decomposing the dynamic interaction cognitive fingerprint data into heterogeneous interaction elements to obtain a user multivariate interaction feature set; Step S245: Perform heterogeneous fusion encoding on the user multivariate interaction feature set to obtain user interaction feature data.

4. The data security verification method of the click-type dynamic verification code according to claim 1 is characterized in that: Step S3 includes the following steps: Step S31: embedding and mapping the user interaction feature data to obtain a low-dimensional interaction habit semantic vector; Step S32: classifying the interaction mode of the corresponding target user according to the low-dimensional interaction habit semantic vector to obtain user interaction mode label data; Step S33: when the user interaction modality label data is normal user data, a personalized visual verification code is generated for the target user client to obtain personalized visual verification code data; Step S34: when the user interaction mode label data is visually impaired user data, a personalized voice verification code is generated for the target user client to obtain personalized voice verification code data; Step S35: using the personalized visual verification code data or the personalized voice verification code data as the personalized verification code data.

5. The data security verification method of the click-type dynamic verification code according to claim 4 is characterized in that: Step S33 includes the following steps: Step S331: when the user interaction modality label data is normal user data, obtaining the user interaction behavior log data and the user account risk entropy value score; Step S332: performing verification code difficulty rating on the user account risk entropy value score according to a preset progressive complexity mapping matrix to obtain expected verification code complexity level data; Step S333: performing abnormal pattern detection and trustworthiness evaluation on the user interaction behavior log data to obtain a verification trustworthiness index; Step S334: evaluating the situational awareness and cognitive ability of the target user according to the user interaction behavior log data to obtain cognitive preference index data; Step S335: Generate a personalized visual verification code for the target user client according to the expected verification code complexity level data, the verification trust index, and the cognitive preference index data to obtain personalized visual verification code data.

6. The data security verification method of the click-type dynamic verification code according to claim 5 is characterized in that: Step S335 includes the following steps: Step S3351: constructing a verification code generation seed library based on the expected verification code complexity level data to obtain a graded visual element seed library; Step S3352: Determine the interference element injection intensity and interference mode of the hierarchical visual element seed library according to the verification trust index, and obtain interference element library data and superposition strategy data; Step S3353: constructing a user visual cognitive preference model according to the cognitive preference index data to obtain a user visual cognitive preference model; Step S3354: inputting the hierarchical visual element seed library into the visual cognition preference model to generate initial verification code screen data; Step S3355: injecting adversarial generated noise into the initial verification code screen data according to the interference element library data to obtain adversarial visual noise layer data; Step S3356: superimpose the adversarial visual noise layer data onto the initial verification code screen data according to the superposition strategy data to obtain personalized visual verification code data.

7. The data security verification method of the click-type dynamic verification code according to claim 4 is characterized in that: Step S34 includes the following steps: Step S341: when the user interaction modality label data is visually impaired user data, voice verification code interaction history data collection is performed on the target user client to obtain a voice verification code interaction history data set; Step S342: extracting the voice segments that have successfully passed the verification from the voice verification code interaction history data set to obtain a historical successful voice verification data set; Step S343: Perform voice verification code parameter analysis on the historical successful voice verification data set to obtain voice verification code parameter pattern data; Step S344: Generate a personalized voice verification code for the target user client according to the voice verification code parameter mode data to obtain personalized voice verification code data.

8. The data security verification method of the click-type dynamic verification code according to claim 7 is characterized in that: Step S344 includes the following steps: Step S3441: constructing a personalized voice element library according to the voice verification code parameter pattern data to obtain a hierarchical voice element library; Step S3442: collecting biological auditory feature data of the target user to obtain the user's auditory physiological data; Step S3443 performs personalized auditory cognition assessment on the target user according to the user's auditory physiological data to obtain a user auditory cognition assessment vector; Step S3444: constructing a voice verification code generation pool according to the user auditory cognitive evaluation vector and the preset voice verification code grammar rules to obtain a voice verification code prototype library; Step S3445: performing personalized voice distortion correction on the voice verification code prototype library according to the user's auditory physiological data to obtain a personalized voice verification code candidate set; Step S3446: Randomly extract the personalized voice verification code candidate set to obtain personalized voice verification code data.

Images