Identity authentication methods, devices, equipment, media and procedures products

By acquiring user interface operation data during financial transactions, extracting key and touch swipe feature vectors, and performing fusion score calculations, the problem of continuity and reliability of identity authentication in existing technologies is solved, and the continuity of identity authentication and security in complex scenarios are improved.

CN123116235APending Publication Date: 2026-11-17INDUSTRIAL AND COMMERCIAL BANK OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611027233.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-10
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing identity authentication methods in the financial sector lack continuous identity verification during transactions. Fixed liveness detection and single biometric identification have limited adaptability to forgery attacks and environmental changes, affecting authentication reliability and transaction security.

Method used

By acquiring operation data during user interface operation, key feature vectors and touch swipe feature vectors are extracted, and dynamic weight allocation is performed using the DS evidence theory algorithm to generate a fusion score to determine the identity authentication result.

Benefits of technology

It enhances the continuity and authenticity of identity authentication, as well as the reliability of authentication in complex scenarios, and can continuously identify the authenticity and consistency of user identity during user interaction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116235A_ABST
    Figure CN123116235A_ABST
Patent Text Reader

Abstract

This application provides an identity authentication method, apparatus, device, medium, and program product, relating to the field of information security. The method acquires user action data generated during interface operations; extracts action features based on the action data to obtain key feature vectors and touch / slide feature vectors. The key feature vectors characterize the user's key presses and touch / slides on the interface; a fusion score is determined based on the key and touch / slide feature vectors, representing the confidence level that the user is a correctly identified user; and an identity authentication result is generated based on the fusion score. By jointly judging continuous actions during interface operations, the continuity, accuracy, and anti-forgery capabilities of identity authentication can be improved, enhancing the security and adaptability to complex scenarios.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of information security, and in particular to an identity authentication method, apparatus, device, medium, and program product. Background Technology

[0002] Taking the financial sector as an example, current identity authentication in financial transactions typically employs methods such as password input, fixed liveness detection, or single biometric identification. In the financial sector, this approach can be applied to various scenarios including bank counters, automated teller machines (ATMs), online banking, and mobile banking.

[0003] However, these solutions are mostly one-time static authentication methods, lacking continuous identity verification during the transaction process. Furthermore, fixed liveness commands and single-modal recognition have limited adaptability to forgery attacks, environmental changes, and the verification of the authenticity of abnormal behavior, which can easily affect authentication reliability and transaction security.

[0004] Therefore, how to improve the continuity of identity authentication, the ability to judge authenticity, and the reliability in complex scenarios during business operations has become a technical problem that needs to be solved. Summary of the Invention

[0005] This application provides an identity authentication method, apparatus, device, medium, and program product to solve the aforementioned technical problems. This solution addresses the user's identity verification needs during interface operations. By combining multi-dimensional operational behavior information generated during user interaction, it comprehensively judges the user's identity and outputs the corresponding authentication result. Based on this, it can enhance the continuity of identity authentication during business operations, the ability to determine authenticity, and the reliability of authentication in complex scenarios.

[0006] In a first aspect, embodiments of this application provide an identity authentication method, the method comprising:

[0007] Acquire user action data generated during interface operation;

[0008] Based on the operation action data, operation action features are extracted to obtain key feature vectors and touch swipe feature vectors. Key feature vectors are used to characterize the operation action features of users when they press keys in the interface, and touch swipe feature vectors are used to characterize the operation action features of users when they touch swipe in the interface.

[0009] The fusion score is determined based on the key feature vector and the touch swipe feature vector. The fusion score is used to characterize the degree of confidence that the user is the correct user.

[0010] The system generates user authentication results by integrating scores.

[0011] In one possible implementation, determining the fusion score based on the button feature vector and the touch swipe feature vector includes:

[0012] Dynamically assign weights to the button feature vectors and touch swipe feature vectors;

[0013] The fusion score is obtained by summing the weights based on the dynamic weight allocation results and the button feature vector and touch swipe feature vector.

[0014] In one possible implementation, the dynamic weight allocation is a weight allocation based on a preset baseline confidence level and a preset quality assessment factor; the dynamic weight allocation for the button feature vector and the touch swipe feature vector includes:

[0015] Based on the basic confidence and quality assessment factors of the button feature vector and the touch sliding feature vector respectively, the DS evidence theory algorithm is used to calculate the first dynamic weight of the button feature vector and the second dynamic weight of the touch sliding feature vector respectively.

[0016] The fusion score is obtained by summing the weights based on the dynamic weight allocation results and the button feature vector and touch swipe feature vector, including:

[0017] The button feature vector is weighted based on the first dynamic weight, and the touch swipe feature vector is weighted based on the second dynamic weight. The weighted results are then summed to obtain the fusion score.

[0018] In one possible implementation, operation action features are extracted based on operation action data to obtain a key feature vector, including:

[0019] Based on the operation action data, a five-channel keystroke dynamic image model is used to extract operation action features to obtain operation action feature information. Among them, the five-channel keystroke dynamic image model is a model used to extract operation action features under multiple feature dimensions. The multiple feature dimensions include at least two of the following feature dimensions: key press duration, interval between key release and the next key press, delay between adjacent keys, and virtual pressure sensitivity fluctuation.

[0020] Matrix analysis is performed based on the operation action feature information to obtain the key feature vector.

[0021] In one possible implementation, operation action features are extracted based on operation action data to obtain a touch swipe feature vector, including:

[0022] A realistic trajectory model is generated based on the user's hand motion data. The realistic trajectory model includes a trajectory sequence representing the user's touch swiping motion at multiple sampling points of the hand.

[0023] The simulated trajectory model is encoded using a feature vector encoding model to obtain the touch swipe feature vector.

[0024] In one possible implementation, the method further includes:

[0025] Acquire multimodal biometric data of users, which includes at least two types of data representing user biometrics;

[0026] Biometric feature extraction is performed based on multimodal biometric data to obtain multimodal biometric feature vectors that represent multiple biometric features of users;

[0027] The fusion score is determined based on the button feature vector and the touch swipe feature vector, including:

[0028] The fusion score is determined based on multimodal biometric feature vectors, button feature vectors, and touch swipe feature vectors.

[0029] In one possible implementation, a fusion score is determined based on multimodal biometric vectors, button feature vectors, and touch swipe feature vectors, including:

[0030] Dynamic weight allocation is performed on multimodal biometric feature vectors, button feature vectors, and touch swipe feature vectors;

[0031] The fusion score is obtained by summing the weights based on the dynamic weight allocation results and the multimodal biometric feature vectors, button feature vectors, and touch swipe feature vectors.

[0032] In one possible implementation, generating a user authentication result by fusing scores includes:

[0033] The fusion score is matched with the fusion score range of multiple preset decision intervals;

[0034] If a decision interval is successfully matched as the target decision interval among multiple decision intervals, the authentication result corresponding to the target decision interval will be determined as the user's authentication result.

[0035] Secondly, embodiments of this application provide an identity authentication device, the device comprising:

[0036] The acquisition module is used to acquire user action data generated during interface operations;

[0037] The feature extraction module is used to extract operation action features based on operation action data, and obtain key feature vectors and touch swipe feature vectors. Key feature vectors are used to characterize the operation action features of users when they press keys in the interface, and touch swipe feature vectors are used to characterize the operation action features of users when they touch swipe in the interface.

[0038] The determination module is used to determine the fusion score based on the button feature vector and the touch swipe feature vector. The fusion score is used to characterize the degree of confidence that the user is the correct user.

[0039] The generation module is used to generate user-specific authentication results by fusing scores.

[0040] In one possible implementation, the determining module is specifically used for:

[0041] Dynamically assign weights to the button feature vectors and touch swipe feature vectors;

[0042] The fusion score is obtained by summing the weights based on the dynamic weight allocation results and the button feature vector and touch swipe feature vector.

[0043] In one possible implementation, the dynamic weight allocation is a weight allocation based on a preset baseline confidence level and a preset quality assessment factor; the determination module is specifically used for:

[0044] Based on the basic confidence and quality assessment factors of the button feature vector and the touch sliding feature vector respectively, the DS evidence theory algorithm is used to calculate the first dynamic weight of the button feature vector and the second dynamic weight of the touch sliding feature vector respectively.

[0045] The button feature vector is weighted based on the first dynamic weight, and the touch swipe feature vector is weighted based on the second dynamic weight. The weighted results are then summed to obtain the fusion score.

[0046] In one possible implementation, the feature extraction module is specifically used for:

[0047] Based on the operation action data, a five-channel keystroke dynamic image model is used to extract operation action features to obtain operation action feature information. Among them, the five-channel keystroke dynamic image model is a model used to extract operation action features under multiple feature dimensions. The multiple feature dimensions include at least two of the following feature dimensions: key press duration, interval between key release and the next key press, delay between adjacent keys, and virtual pressure sensitivity fluctuation.

[0048] Matrix analysis is performed based on the operation action feature information to obtain the key feature vector.

[0049] In one possible implementation, the feature extraction module is specifically used for:

[0050] A realistic trajectory model is generated based on the user's hand motion data. The realistic trajectory model includes a trajectory sequence representing the user's touch swiping motion at multiple sampling points of the hand.

[0051] The simulated trajectory model is encoded using a feature vector encoding model to obtain the touch swipe feature vector.

[0052] In one possible implementation, the acquisition module is further configured to: acquire the user's multimodal biometric data, the multimodal biometric data including at least two types of data characterizing the user's biometrics;

[0053] The feature extraction module is also used to: extract biometric features based on multimodal biometric data to obtain multimodal biometric vectors that represent multiple biometric features of the user;

[0054] The determination module is specifically used to: determine the fusion score based on multimodal biometric feature vectors, button feature vectors, and touch swipe feature vectors.

[0055] In one possible implementation, the determining module is specifically used for:

[0056] Dynamic weight allocation is performed on multimodal biometric feature vectors, button feature vectors, and touch swipe feature vectors;

[0057] The fusion score is obtained by summing the weights based on the dynamic weight allocation results and the multimodal biometric feature vectors, button feature vectors, and touch swipe feature vectors.

[0058] In one possible implementation, the generation module is specifically used for:

[0059] The fusion score is matched with the fusion score range of multiple preset decision intervals;

[0060] If a decision interval is successfully matched as the target decision interval among multiple decision intervals, the authentication result corresponding to the target decision interval will be determined as the user's authentication result.

[0061] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0062] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0063] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0064] The identity authentication method, apparatus, device, medium, and program products provided in this application's embodiments acquire user action data generated during interface operations, and extract key feature vectors and touch swipe feature vectors based on this data. Combining the key and touch swipe feature vectors determines a fusion score characterizing the degree of confidence that the user is a legitimate user. Based on this, the identity status can be comprehensively characterized and judged from different operational dimensions during user interaction, thereby improving the continuity of identity authentication, the ability to identify abnormal or forged behavior, and the reliability of authentication in complex scenarios. Attached Figure Description

[0065] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0066] Figure 1 Flowchart of the identity authentication method provided in the embodiments of this application Figure 1 ;

[0067] Figure 2 Flowchart of the identity authentication method provided in the embodiments of this application Figure 2 ;

[0068] Figure 3 A flowchart illustrating the execution process of the authentication method provided in this application embodiment;

[0069] Figure 4 This is a schematic diagram of the structure of the identity authentication device provided in the embodiments of this application;

[0070] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0071] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0072] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0073] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with the relevant laws, regulations, and standards of the relevant countries and regions, have taken necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation access points for users to choose to authorize or refuse.

[0074] Furthermore, the technical solution involved in this application, which involves big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.) and the use of artificial intelligence technology for automated decision-making, and makes decisions that have a significant impact on personal rights based on the results of automated decision-making, provides users with corresponding operation entry points for users to choose to agree to or reject the results of automated decision-making; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0075] In this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0076] In the embodiments of this application, the use of terms such as "first" and "second" is to distinguish between identical or similar items that have essentially the same function and effect. For example, "first electronic device" and "second electronic device" are merely used to distinguish different electronic devices and do not limit their order of execution. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that "first" and "second" do not necessarily imply that they are different.

[0077] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following associated objects have an "or" relationship.

[0078] It should be noted that the identity authentication methods, devices, equipment, media, and program products provided in the embodiments of this application can be used in the fields of information security technology and finance, as well as in any field other than information security and finance. The application fields of the identity authentication methods, devices, equipment, media, and program products provided in the embodiments of this application are not limited.

[0079] The identity authentication method described in this application can be applied to security authentication in financial business scenarios. For example, it is applicable to business scenarios requiring users to perform key input and touch swipe operations on an interface, such as bank counter terminals, ATMs, self-service interactive devices, online banking, and mobile banking. In such scenarios, the user identity confirmation and subsequent business processing are typically completed collaboratively by the terminal interface, identity verification module, and transaction processing module.

[0080] Existing authentication methods typically involve a one-time identity verification process before a transaction begins, using methods such as password input, fixed liveness detection, or single biometric identification. The basic idea is to collect authentication information when the user initiates the action and then provide a pass or fail authentication result based on this information.

[0081] However, the above solutions mainly focus on static authentication before the transaction, lacking further analysis of the user's subsequent continuous operations on the interface, and are difficult to cover changes in identity status during the transaction process. Once authentication is completed, the terminal may be taken over by others, manipulated by malicious programs, or the user may continue to operate in an abnormal state. Business systems often cannot accurately identify risks based solely on the original one-time authentication result.

[0082] Furthermore, fixed liveness detection commands or single-feature recognition have limited adaptability to complex scenarios and are easily affected by environmental changes, forgery attacks, and abnormal operating behaviors. For example, during interface interaction, relying solely on passwords or single biometric features makes it difficult to combine the user's actual keystroke habits and touch swipe behaviors to determine whether the operator remains consistent, thus affecting authentication reliability and transaction security.

[0083] In view of this, improving the continuity of identity authentication and the ability to judge the true user behavior during business operations has become an urgent technical problem to be solved. To address these problems, embodiments of this application provide an identity authentication method. This method acquires the user's action data when the user performs interface operations, and extracts key feature vectors and touch swipe feature vectors based on the action data. A fusion score is determined based on the key feature vectors and touch swipe feature vectors, and finally, an identity authentication result for the user is generated based on the fusion score.

[0084] For example, in some financial terminals or online interactive systems, relevant interface operation data can be collaboratively acquired and processed by the terminal interface, identity verification module, and transaction processing module. This allows for continuous analysis of user key presses and touch swipes within the existing business architecture. By fusing and judging multiple action features during the operation process, the ability to identify the authenticity and consistency of user identity can be enhanced, improving the reliability of authentication in complex business scenarios.

[0085] The technical solutions of this application will be described in detail below with reference to specific embodiments. The specific embodiments described below can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0086] Figure 1 Flowchart of the identity authentication method provided in the embodiments of this application Figure 1 The execution subject of this method can be an electronic device with corresponding data storage and computing capabilities, such as a computer, server, or server cluster. Figure 1 As shown, the method includes:

[0087] S101, Obtain the operation data generated by the user during the interface operation.

[0088] For example, the action data can be raw data used to carry out subsequent feature extraction, such as key operation information and touch swipe operation information generated by the user during interface interaction. This step can be performed by the terminal-side data collection module, or it can be performed collaboratively by the terminal and the background authentication service.

[0089] For example, in the specific implementation, relevant operation events are collected during the interface interaction process, and data related to button operation and touch swipe operation are recorded to form operation action data that can be processed later.

[0090] Based on the above analysis, this step uses the user's actual actions on the interface as input, providing data for subsequent extraction of identity-related features from the two dimensions of button presses and touch swipes. It should be understood that the above example is merely illustrative and not limiting.

[0091] S102, based on the operation action data, operation action features are extracted to obtain key feature vector and touch sliding feature vector. The key feature vector is used to characterize the operation action features when the user presses a key in the interface, and the touch sliding feature vector is used to characterize the operation action features when the user touches and slides in the interface.

[0092] For example, a key press feature vector can be understood as a tool to convert user key presses into a computable and comparable feature representation. A touch swipe feature vector can be understood as a tool to convert user swipe trajectories and their dynamic changes into feature representations that can be used for authentication decisions.

[0093] For example, in practical implementation, the key event sequence and touch swipe sequence are first separated from the operation action data, and preprocessing is performed. For the key event sequence, key operation features can be extracted based on information such as press duration, key interval, rhythm changes, and pressure sensitivity changes. The extracted key operation features are then encoded to obtain a key feature vector. This key feature vector can reflect the temporal pattern of the user's key operations in the interface, the characteristics of press duration, and the temporal relationship between adjacent keys.

[0094] For touch swipe sequences, touch swipe operation features can be extracted based on information such as trajectory changes, speed changes, acceleration changes, direction changes, and pressure sensitivity changes. The extracted touch swipe operation features can be encoded to obtain a touch swipe feature vector. This touch swipe feature vector reflects the trajectory changes and dynamic operation characteristics of the user's touch swipe operation on the interface. To ensure that the two vectors can directly participate in subsequent fusion, the button feature vector and the touch swipe feature vector can be uniformly mapped to the same numerical space.

[0095] Based on the above analysis, this step decomposes and encodes the original operation data into two complementary behavioral feature vectors. The keystroke feature vector characterizes discrete keystroke behavior, while the touch swipe feature vector characterizes continuous touch behavior. This allows the system to simultaneously characterize the current operator using both types of operational habits during business operations, enhancing the ability of subsequent identity determination to identify consistent behavior and abnormal takeover attempts. It should be understood that the above example is for illustrative purposes only and is not limiting.

[0096] S103, determine the fusion score based on the button feature vector and the touch swipe feature vector. The fusion score is used to characterize the degree of confidence that the user is the correct user.

[0097] For example, the fusion score is a numerical result obtained by fusing the key feature vector and the touch swipe feature vector. The fusion score can be used to represent the degree of consistency between the current user and the correctly identified user.

[0098] For example, in a practical implementation, the electronic device executing the task can obtain the corresponding feature matching results based on the button feature vector and the touch swipe feature vector respectively, and perform fusion processing based on the two types of feature matching results to obtain a fusion score.

[0099] For example, matching results can be obtained through vector distance, similarity calculation, or classification model output. For instance, a higher fusion score indicates a higher confidence level that the current user is the correct user; a lower fusion score indicates a greater deviation between the current action and the behavior of a known correct user.

[0100] Based on the above analysis, this step, by uniformly quantifying the two types of behavioral features—button presses and touch swipes—forms an intermediate authentication measure that connects feature extraction and authentication decision-making. Compared to judgment methods that rely solely on a single behavioral signal, this embodiment can more stably reflect the consistency between the current operating entity and the target user during actual operation. It should be understood that the above example is merely illustrative and not limiting.

[0101] S104 generates user authentication results by fusing scores.

[0102] For example, the authentication result can be understood as the authentication conclusion output by the user. The authentication result can serve as the basis for the terminal interface, the identity verification module, and the transaction processing module to execute subsequent responses.

[0103] For example, in practical implementation, after an electronic device obtains a fusion score, it can generate a corresponding identity authentication result based on the fusion score. The identity authentication result may include, for example, authentication successful, authentication unsuccessful, or triggering further identity verification.

[0104] For example, the authentication result can be written to the current session context and fed back to the interface control logic and / or transaction processing logic so that corresponding processing can be performed based on the authentication result.

[0105] Based on the above analysis, this step can convert the fusion score obtained from continuous behavior analysis into an executable authentication conclusion, so that identity verification is no longer a one-time judgment before the transaction begins, but is integrated throughout the interface operation process. Furthermore, the identity authentication result can serve as a direct constraint on the execution conditions of subsequent business actions, thereby outputting corresponding control results when the terminal is taken over, subjected to abnormal manipulation, or when the user's operation state deviates significantly. It should be understood that the above example is for demonstration purposes only and is not limiting.

[0106] The authentication method provided in this application includes acquiring user action data generated during interface operations. Based on the action data, action features are extracted to obtain key feature vectors and touch / swipe feature vectors. The key feature vectors characterize the user's key presses on the interface, and the touch / swipe feature vectors characterize the user's touch / swipe operations on the interface. A fusion score is determined based on the key and touch / swipe feature vectors, representing the confidence level that the user is a correctly identified user. An authentication result for the user is generated based on the fusion score.

[0107] In this embodiment, by continuously collecting, encoding, and fusing user button and touch swipe actions during transactions, the originally static one-time identity verification is expanded into a continuous identity determination throughout the entire process. This allows the authentication result to update with changes in interface interaction and drives corresponding authentication processing based on the consistency of current behavior. This enhances the continuity and authenticity of identity authentication during business operations, as well as the reliability of authentication in complex scenarios. This method is applicable to continuous identity verification in financial business scenarios. It should be understood that the above example is merely illustrative and not limiting.

[0108] Figure 2 Flowchart of the identity authentication method provided in the embodiments of this application Figure 2 ,like Figure 2 As shown, in one possible implementation, the fusion score is determined based on the button feature vector and the touch sliding feature vector, including: S201, dynamically assigning weights to the button feature vector and the touch sliding feature vector; S202, calculating the fusion score by summing the weights based on the result of the dynamic weight assignment and the button feature vector and the touch sliding feature vector.

[0109] For example, the dynamic weight allocation result may include a first dynamic weight for the button feature vector and a second dynamic weight for the touch swipe feature vector. The sum of the first dynamic weight and the second dynamic weight is a preset total weight.

[0110] The electronic device executing the task can generate initial weights based on the product of a base confidence level and a quality assessment factor. These initial weights are then normalized to obtain a first dynamic weight and a second dynamic weight. Subsequently, the button feature vector is weighted based on the first dynamic weight, and the touch swipe feature vector is weighted based on the second dynamic weight. The weighted results are then summed to obtain a fusion score used for identity determination.

[0111] For example, the weighted summation calculation can be implemented using vector dot product or scalar weighted accumulation. Specifically, the formula can be written as: the fusion score equals the product of the first dynamic weight and the score corresponding to the button feature vector, plus the product of the second dynamic weight and the score corresponding to the touch swipe feature vector. In practical applications, the weight allocation module can also use other equivalent algorithms, which are not limited in this application.

[0112] In this embodiment, by adaptively weighting the two types of feature vectors, the fusion score can be adjusted according to the credibility changes of different operational behaviors, thereby ensuring that the authentication result remains consistent with the current interface operation state. Since the effectiveness of button features and touch swipe features varies in different scenarios, the fusion score, after dynamic weight allocation, can reflect the relative contribution of the two types of features, thus providing a more stable basis for the authentication result.

[0113] In one possible implementation, the dynamic weight allocation is a weight allocation based on a preset base confidence level and a preset quality assessment factor; the dynamic weight allocation for the button feature vector and the touch sliding feature vector includes: based on the base confidence level and quality assessment factor of the button feature vector and the touch sliding feature vector respectively, using the DS evidence theory algorithm to calculate the first dynamic weight of the button feature vector and the second dynamic weight of the touch sliding feature vector respectively.

[0114] The fusion score is obtained by summing the weights based on the dynamic weight allocation results and the button feature vector and touch swipe feature vector. This includes: weighting the button feature vector based on the first dynamic weight, weighting the touch swipe feature vector based on the second dynamic weight, and summing the weighted results to obtain the fusion score.

[0115] For example, after extracting the features of button operations and touch swipe operations, a basic confidence level is calculated for each type of feature vector. The basic confidence level can be determined by historical authentication consistency, feature stability, and prior matching degree. Furthermore, a quality assessment factor can be calculated for the currently collected feature data. The quality assessment factor can be determined by sampling completeness, noise level, and the proportion of valid segments.

[0116] Subsequently, the corresponding baseline confidence level and quality assessment factor can be input into the Dempster-Shafer Evidence Theory (DS) algorithm to obtain the first dynamic weight of the button feature vector and the second dynamic weight of the touch swipe feature vector, respectively. Furthermore, the button feature vector can be multiplied by the first dynamic weight, and the touch swipe feature vector can be multiplied by the second dynamic weight. The weighted results of the two can then be summed to form a fusion score.

[0117] In this embodiment, the button feature vector and the touch swipe feature vector are not fused in a fixed ratio, but rather their weights are dynamically allocated based on their respective baseline confidence levels and quality assessment factors. For example, when a feature vector has a high level of confidence and good data quality, its corresponding dynamic weight increases; when its quality is poor, its corresponding dynamic weight decreases. Therefore, the fusion score can more accurately reflect the consistency status of the current operating entity and provide a stable basis for subsequent identity authentication result generation.

[0118] In one possible implementation, operation action features are extracted based on operation action data to obtain key feature vectors. This includes: extracting operation action features using a five-channel keystroke dynamic image model based on the operation action data to obtain operation action feature information; wherein, the five-channel keystroke dynamic image model is a model used for operation action feature extraction under multiple feature dimensions, and the multiple feature dimensions include at least two of the following feature dimensions: key press duration, interval between key release and next key press, delay between adjacent keys, and virtual pressure sensitivity fluctuation; matrix analysis is performed based on the operation action feature information to obtain key feature vectors.

[0119] For example, action data can be collected by the terminal when the user performs keyboard input, numeric keypad input, or touch key input. The collected action data includes at least the key press time, key release time, adjacent key switching time, and sampled values ​​related to pressure mapping.

[0120] After constructing a keystroke feature sequence based on the operation action data as the raw data, each sequence can be aligned and normalized according to a preset time window, forming feature inputs for multiple channels. The five-channel keystroke dynamic image model can be implemented using, for example, a convolutional feature extraction structure or a temporal feature encoding structure. The operation action feature information is output after fusing the feature responses of each channel through the five-channel keystroke dynamic image model. In practical applications, other model structures can also be selected for the five-channel keystroke dynamic image model; this application does not limit this choice.

[0121] Matrix analysis operations can be used to convert action feature information into fixed-dimensional keystroke feature vectors. For example, action feature information can be arranged in chronological order to form a feature matrix. Dimensionality reduction and recombination can be performed through matrix multiplication, feature projection, principal component mapping, or tensor expansion, allowing keystroke features of different dimensions to be expressed in a unified vector space. The keystroke feature vectors obtained through matrix analysis operations can quantitatively represent the user's keystroke rhythm, input habits, and operational stability, and can serve as input for subsequent identity authentication fusion calculations.

[0122] In this embodiment, the duration of key press, the interval between key release and the next key press, the delay between adjacent keys, and virtual pressure sensitivity fluctuations can be uniformly incorporated into the feature extraction process. Furthermore, matrix analysis can generate a structured key feature vector, preserving subtle differences in key behavior and using it for identity verification, thereby improving the ability to distinguish user key operation features and the consistency of authentication.

[0123] In one possible implementation, operation action features are extracted based on operation action data to obtain touch sliding feature vectors, including: generating a simulated trajectory model based on operation action data, the simulated trajectory model including a trajectory sequence representing the touch sliding operation actions of multiple sampling points of the user's hand; and encoding the simulated trajectory model with feature vectors through a feature vector encoding model to obtain touch sliding feature vectors.

[0124] For example, operation data can be collected by the touch terminal during the user's swipe. Operation data includes at least the sampling timestamp, touch point coordinates, pressure sensitivity value, and swipe direction information.

[0125] For example, based on these operation action data as raw data, adjacent sampling points (e.g., 60-120 sampling points) are first connected in chronological order to construct a simulated trajectory model reflecting the continuity of sliding. Then, the simulated trajectory model is input into a pre-trained feature vector encoding model, which outputs a touch sliding feature vector. The feature vector encoding model can employ at least one of recurrent neural networks, temporal convolutional networks, or self-attention networks to model the temporal dependencies in the trajectory sequence, thereby outputting the touch sliding feature vector. Other model structures can also be selected for the feature vector encoding model in practical applications; this application does not limit this choice.

[0126] For example, during operation, the discrete sampled data generated by touch swipes can be reconstructed to form a realistic trajectory model with continuous temporal characteristics from the original operation. The behavioral differences in the trajectory are then converted into a computable vector representation through an encoding model, providing input for the subsequent calculation of the fusion score in identity authentication.

[0127] In this embodiment, touch swiping behavior can be converted from raw sampling points into structured feature vectors, preserving trajectory changes, speed changes, and pressure sensitivity fluctuations. The resulting touch swiping feature vectors have a stronger ability to distinguish the swiping habits of different users and can improve the accuracy of subsequent identity authentication in representing the consistency of continuous operation subjects.

[0128] In one possible implementation, the method further includes: acquiring multimodal biometric data of the user, the multimodal biometric data including at least two types of data representing the user's biometric features; performing biometric feature extraction based on the multimodal biometric data to obtain a multimodal biometric vector representing multiple biometric features of the user; and determining a fusion score based on the button feature vector and the touch swipe feature vector, including: determining the fusion score based on the multimodal biometric vector, the button feature vector, and the touch swipe feature vector.

[0129] For example, after the user authorizes the data collection, the electronic device can acquire corresponding modal data through a microphone, camera, fingerprint acquisition module, or other biometric acquisition unit. The acquired multimodal biometric data can then be input into a feature extraction module to extract multimodal biometric vectors.

[0130] For example, the feature extraction module can use convolutional networks, recurrent networks, or feature encoders to encode data from different modalities, and then concatenate or map the features of each modality to the same feature space to form a multimodal biometric vector. This multimodal biometric vector, along with button feature vectors and touch swipe feature vectors, can be input into a fusion calculation module for fusion calculation to obtain a fusion score. For example, the fusion calculation module can generate a fusion score based on the similarity, confidence level, or weighting rules between vectors.

[0131] In this implementation, the multimodal biometric vector can be normalized to the same dimension as the button feature vector and the touch swipe feature vector, and then weighted and summed. Alternatively, it can be first concatenated at the feature level and then mapped to a unified confidence value. The fusion score can be used to reflect the consistency between biometric features and operation action features. For example, when the multimodal biometric features and the current operation behavior both point to the same user, the fusion score remains at a high level; when any modality is inconsistent with the action feature, the fusion score decreases accordingly.

[0132] In this embodiment, by introducing multimodal biometric data into the fusion calculation, additional identity evidence can be obtained beyond button features and touch swipe features, and the fusion score can simultaneously represent the degree of matching between biometric features and operation action features. This improves the completeness and continuity of the identity authentication results.

[0133] In one possible implementation, the fusion score is determined based on the multimodal biometric vector, button feature vector, and touch swipe feature vector, including: dynamically assigning weights to the multimodal biometric vector, button feature vector, and touch swipe feature vector; and calculating the fusion score by summing the weights based on the dynamic weight assignment results and the multimodal biometric vector, button feature vector, and touch swipe feature vector.

[0134] For example, after acquiring multimodal biometric feature vectors, button feature vectors, and touch swipe feature vectors, their quality assessment factors can be calculated separately. These quality assessment factors can be determined by factors such as data acquisition completeness, noise level, feature stability, and temporal consistency. Subsequently, dynamic weights for each vector can be generated by combining preset baseline confidence levels. These dynamic weights can be normalized to weight coefficients that sum to one.

[0135] The multimodal biometric vector, button feature vector, and touch swipe feature vector are weighted according to weighting coefficients, and the weighted results are summed to obtain a fusion score for identity authentication. The weight summation calculation can be performed locally on the terminal side of the electronic device, or it can be performed by the identity verification module in a secure computing environment. The resulting fusion score can be used as a unified indicator output for continuous authentication.

[0136] For example, during implementation, when the confidence of a certain type of vector decreases due to reduced acquisition quality, its dynamic weight decreases accordingly, and the weights of other vectors are automatically compensated, thus keeping the fusion score stable and consistent with the current operating state. This fusion mechanism enables multimodal biometrics, button behavior, and touch swipe behavior to jointly participate in identity determination, forming a unified authentication confidence output.

[0137] In this embodiment, the fusion score comprehensively reflects the credibility of multimodal biometrics, key features, and touch swipe features under the current acquisition conditions. Based on this, the impact of a single feature anomaly on the authentication result can be reduced, improving the continuity and stability of identity authentication in financial interaction scenarios.

[0138] In one possible implementation, generating a user's identity authentication result through a fusion score includes: matching the fusion score with a range of fusion scores from multiple preset decision intervals; and if a decision interval is successfully matched as a target decision interval among the multiple decision intervals, determining the identity authentication result corresponding to the target decision interval as the user's identity authentication result.

[0139] For example, after obtaining the fusion score, multiple decision intervals pre-stored in the parameter library can be read. Each pre-stored decision interval is defined with a start value and an end value, and adjacent intervals can be configured as closed intervals, half-open intervals, or open intervals to ensure that the fusion score matches only one valid interval.

[0140] The fusion scores are compared within a certain range. When a fusion score falls within a certain range, the decision interval corresponding to that range is determined as the target decision interval. The identity authentication result bound to this target decision interval can be extracted as the final output.

[0141] For example, if the fusion score is in the high confidence range, the authentication result is output as passed. If it is in the middle confidence range, the enhanced authentication result is output as passed. If it is in the low confidence range, the interception or rejection result is output as passed as passed. Taking a banking system as an example, for ease of deployment, the decision range and its corresponding identity authentication result can be centrally configured by the server and distributed to bank counter terminals, ATMs, self-service interactive devices, or mobile banking clients, etc. In practical applications, the generation module used to generate authentication results can also be other models, and this application does not limit this.

[0142] For example, after interval matching, the fusion score is discretized into a clear authentication conclusion. The generation module then transmits the result to the transaction processing module or the interface control module to determine whether to continue subsequent business processing. Since the same fusion score corresponds to only one target decision interval, the electronic device can stably output a unique authentication conclusion, thus forming a closed loop between continuous identity verification and transaction control.

[0143] In this embodiment, the fusion score is mapped to a preset discrete authentication result, and the output rules are clear and easy for the terminal to execute. By binding different score ranges with different authentication conclusions, identity judgment can be directly linked to subsequent business control, improving the consistency and feasibility of identity authentication results.

[0144] Figure 3 The execution flowchart of the identity authentication method provided in the embodiments of this application is as follows: Figure 3 As shown, the electronic device executing the task can collect (e.g., synchronously sample) multimodal biometric data of face, voiceprint, and palm vein through sensors for face recognition, voiceprint recognition, and palm vein recognition. Subsequently, each recognition module can buffer and distribute data through a lightweight message queue (such as Kafka) to achieve low-latency communication and millisecond-level response.

[0145] The system for performing identity authentication and business response can include a data acquisition layer, an analysis layer, an evaluation layer, and a response layer. The data acquisition layer can simultaneously sample data at 5Hz from infrared cameras, voiceprint microphones, and palm vein sensors. The raw data is then timestamped and encapsulated into JavaScript Object Notation (JSON) format (including device ID, timestamp, and modality type) and pushed to a topic in the message queue.

[0146] The analysis layer can be understood as a behavior analysis module, which can realize real-time behavior recognition, including key presses and touch trajectories. The behavior feature extraction module can output structured feature vectors (such as dynamic images of keyboard keystrokes and acceleration spectra of touch trajectories), and can call the fusion engine for fusion score calculation via remote program invocation to achieve multimodal fusion decision-making.

[0147] The evaluation layer can dynamically score and output a fusion score (0–100) in real time. The response layer can make decisions and provide feedback in real time, enabling business system responses. Authentication results can be pushed to the business system in the response layer in real time via WebSockets to trigger interception or release commands. For example, the data stream protocol follows the ISO / IEC 30107-3 biometric data exchange standard, ensuring cross-platform compatibility. During data stream transmission, multimodal authentication remains active and pushes authentication results to the transaction.

[0148] For example, the identity authentication method provided in this application can be applied to the scenario of contactless continuous identity authentication in bank ATMs. In financial transaction scenarios, users need to complete identity verification when performing operations such as withdrawals and transfers through ATMs. Existing identity authentication methods rely on users entering passwords or executing fixed liveness commands, such as "shaking their head and opening their mouth," but these methods are susceptible to security risks such as artificial intelligence (AI) face-swapping attacks and 3D mask bypassing.

[0149] Based on the method of this application embodiment, in a withdrawal example, when a user approaches the ATM, the system automatically initiates a multimodal authentication process. An infrared camera captures the user's facial features in real time, a voiceprint microphone collects the user's voice commands, such as "withdraw 500 yuan," and a palm vein sensor scans the blood vessels in the user's palm.

[0150] In addition, the system records the user's key presses on the ATM touchscreen during input, clicks, swipes, or switching operations on the terminal interface. This includes inputting account information and swiping movements, such as selecting a withdrawal amount. Throughout the process, the user does not need to actively enter a password or execute a liveness detection command. The system continuously generates an authentication score during the operation and triggers either release or blocking commands based on the score.

[0151] In one possible implementation, when a user enters account information, such as a bank card number, on the ATM touchscreen and slides to select the withdrawal amount, the system collects the user's operational behavior data and biometric data by synchronously sampling at 5Hz through an infrared camera, voiceprint microphone, and palm vein sensor.

[0152] The system includes an infrared camera to record the user's facial expressions and blink frequency; a voiceprint microphone to capture the pitch fluctuations of the user's voice commands; a palm vein sensor to scan the clarity of the blood vessels in the user's palm; and a touchscreen sensor to record the duration of button presses, the delay between adjacent buttons, and the acceleration changes of swipe trajectories. The operation data includes timing data related to button operations and trajectory data related to touch swipes.

[0153] All collected data can be timestamped and encapsulated into JSON format, and then pushed to a message queue, such as a Kafka queue, so that the acceleration changes of the touch trajectory and the tone fluctuations of the voiceprint command are kept synchronized when the user slides to select the amount.

[0154] When extracting operation action features from operation action data, a five-channel keystroke dynamic image model can be used to extract operation action feature information for the key operation part. The five-channel keystroke dynamic image model represents key behavior under multiple feature dimensions. These multiple feature dimensions include at least two of the following: key press duration, interval between key release and the next key press, delay between adjacent keys, and virtual pressure sensitivity fluctuation. The operation rhythm entropy value is then extracted by combining the four-dimensional features in the scene with information such as key press duration and delay between adjacent keys.

[0155] For example, when a user enters a bank card number, the regularity of the key press interval, such as pressing once every 300ms on average, is converted into operation action feature information. Matrix analysis is performed on the obtained operation action feature information to output key feature vectors, which represent the characteristics of the user's key operations on the interface.

[0156] For the touch-sensitive swiping portion, touch point position, movement direction, displacement changes, and sampling timing information can be extracted from the operation data. This information, combined with the nonlinear acceleration changes of the touch trajectory and micro-jitter (Gaussian noise, σ=0.08), generates a simulated trajectory model. This simulated trajectory model may include, for example, a trajectory sequence representing the touch-sensitive swiping operation actions of multiple sampling points on the user's hand.

[0157] In the example of the self-service teller machine, the simulated trajectory model can be generated based on a graphical user interface (GUI) automated control AI framework. For instance, when a user slides to select "500 yuan," a sudden change in the trajectory's acceleration, such as a sudden deceleration, is incorporated into the simulated trajectory model. The simulated trajectory model is then encoded using a feature vector encoding model, such as a Long Short-Term Memory (LSTM) autoencoder, to obtain the touch-swipe feature vector. This touch-swipe feature vector characterizes the user's touch-swipe actions on the interface and is used to identify abnormal patterns corresponding to the robot's uniform sliding motion.

[0158] The terminal or authentication server can also acquire the user's multimodal biometric data. Multimodal biometric data includes at least two types of data representing the user's biometric characteristics. In banking transaction scenarios, this includes, for example, facial features, voiceprints, and palm veins. Based on the multimodal biometric data, biometric feature extraction can be performed to obtain a multimodal biometric vector representing multiple biometric characteristics of the user.

[0159] The facial data acquired by the infrared camera can include facial features, facial expressions, and blink frequency. The voiceprint data acquired by the voiceprint microphone includes the tone fluctuations of the voice command "withdraw 500 yuan". The palm vein data acquired by the palm vein sensor can include the palm veins and their clarity. Feature extraction and fusion calculations can be completed at the ATM terminal, forming an edge computing architecture, with the data remaining within the device.

[0160] During the fusion determination phase, the button feature vector and the touch swipe feature vector are input into the fusion calculation module. When the system simultaneously acquires multimodal biometric vectors, these vectors are input together with the button and touch swipe feature vectors into the fusion calculation module. The fusion calculation module first performs dynamic weight allocation on each vector, based on a preset baseline confidence level and a preset quality assessment factor.

[0161] Among them, the button feature vector and the touch swipe feature vector each correspond to a basic confidence level and a quality assessment factor. The DS evidence theory algorithm is used to calculate the first dynamic weight of the button feature vector and the second dynamic weight of the touch swipe feature vector. When multimodal biometric vectors are involved in fusion, the multimodal biometric vectors are also dynamically weighted according to their corresponding basic confidence levels and quality assessment factors.

[0162] In the example scenario of ATM operation, the system can adjust the weights by combining the basic confidence level and quality assessment factors of each modality.

[0163] For example, the baseline confidence levels were set as follows: face recognition 0.85, voiceprint recognition 0.78, palm vein recognition 0.92, key press behavior 0.81, and touch trajectory 0.79. Quality assessment factors were set as follows: for face recognition, factors included illumination and occlusion rate; for voiceprint recognition, factors included background noise; for palm vein recognition, factors included vessel clarity; for key press behavior, factors included input stability; and for touch trajectory, factors included acceleration consistency.

[0164] For example, when the ambient light is dim, the weight of face recognition can be reduced, such as adjusting the dynamic weight Wf of face recognition to 0.7; for another example, when the clarity of blood vessels is high, the weight of palm vein recognition can be increased, such as adjusting the dynamic weight Wp of palm vein recognition to 0.95; for yet another example, in a high background noise environment, the dynamic weight Ws of voiceprint recognition can be reduced to 0.6.

[0165] After dynamic weight allocation is completed, the fusion calculation module performs weighted summation calculation based on the allocation results. The button feature vector is weighted based on the first dynamic weight, and the touch swipe feature vector is weighted based on the second dynamic weight. When multimodal biometric vectors exist, corresponding weights are applied to each multimodal biometric vector, and the weighted results are summed to obtain the fusion score. The fusion score can be used to characterize the degree of confidence that the user is the correct user.

[0166] In this scenario, the weighted feature vectors are summed to generate a fusion score within the range of 0–100. This fusion score is then matched against the fusion score ranges of multiple preset decision intervals. When a decision interval is successfully matched as the target decision interval, the authentication result corresponding to the target decision interval is directly determined as the user's authentication result.

[0167] For example, the formula for calculating the fusion score can be expressed as: , where Score represents the fusion score; Wi represents the dynamic weight of the i-th modality; and Ci represents the base confidence of the i-th modality.

[0168] For example, when the fusion score is greater than 90, the authentication is successful, and the ATM can dispense the corresponding amount of cash and print a receipt. When 75 ≤ Score ≤ 90, enhanced authentication is triggered, such as a secondary biometric verification prompt appearing on the ATM screen, requiring the user to blink for verification. When the score is less than 75, the transaction is immediately interrupted, triggering a risk alarm, such as the ATM screen displaying an alarm message like "Identity abnormal, please contact staff."

[0169] For example, when a user's blinking frequency is abnormal due to fatigue, resulting in a Score of 78, the system matches this fusion score to the target decision range corresponding to 75 ≤ Score ≤ 90 and triggers enhanced authentication without directly interrupting the transaction. Throughout the transaction process, the system continuously collects and analyzes multimodal information such as palm veins, touch trajectories, facial features, and voiceprints, and cross-validates AI face-swapping attacks, 3D mask bypassing, deepfakes, and behaviors that only forge a single modality.

[0170] The method based on the embodiments of this application can break through the existing one-time authentication mode and achieve end-to-end trust. Users do not need to enter passwords or execute liveness commands during the business process, which can significantly improve business processing efficiency and customer experience. Multimodal dynamic fusion significantly improves system robustness and meets the information protection requirements of localized data processing and data not leaving the terminal.

[0171] Figure 4 This is a schematic diagram of the structure of the identity authentication device provided in the embodiments of this application, such as... Figure 4As shown in the figure, this application embodiment provides an identity authentication device, which includes:

[0172] The acquisition module 401 is used to acquire operation action data generated by the user during the interface operation;

[0173] The feature extraction module 402 is used to extract operation action features based on operation action data to obtain key feature vectors and touch sliding feature vectors. The key feature vector is used to characterize the operation action features when the user presses a key in the interface, and the touch sliding feature vector is used to characterize the operation action features when the user touches and slides in the interface.

[0174] The determination module 403 is used to determine the fusion score based on the button feature vector and the touch swipe feature vector. The fusion score is used to characterize the degree of confidence that the user is the correct user.

[0175] Module 404 is used to generate user authentication results by merging scores.

[0176] In one possible implementation, the determining module 403 is specifically used for:

[0177] Dynamically assign weights to the button feature vectors and touch swipe feature vectors;

[0178] The fusion score is obtained by summing the weights based on the dynamic weight allocation results and the button feature vector and touch swipe feature vector.

[0179] In one possible implementation, the dynamic weight allocation is a weight allocation based on a preset baseline confidence level and a preset quality assessment factor; the determining module 403 is specifically used for:

[0180] Based on the basic confidence and quality assessment factors of the button feature vector and the touch sliding feature vector respectively, the DS evidence theory algorithm is used to calculate the first dynamic weight of the button feature vector and the second dynamic weight of the touch sliding feature vector respectively.

[0181] The button feature vector is weighted based on the first dynamic weight, and the touch swipe feature vector is weighted based on the second dynamic weight. The weighted results are then summed to obtain the fusion score.

[0182] In one possible implementation, the feature extraction module 402 is specifically used for:

[0183] Based on the operation action data, a five-channel keystroke dynamic image model is used to extract operation action features to obtain operation action feature information. Among them, the five-channel keystroke dynamic image model is a model used to extract operation action features under multiple feature dimensions. The multiple feature dimensions include at least two of the following feature dimensions: key press duration, interval between key release and the next key press, delay between adjacent keys, and virtual pressure sensitivity fluctuation.

[0184] Matrix analysis is performed based on the operation action feature information to obtain the key feature vector.

[0185] In one possible implementation, the feature extraction module 402 is specifically used for:

[0186] A realistic trajectory model is generated based on the user's hand motion data. The realistic trajectory model includes a trajectory sequence representing the user's touch swiping motion at multiple sampling points of the hand.

[0187] The simulated trajectory model is encoded using a feature vector encoding model to obtain the touch swipe feature vector.

[0188] In one possible implementation, the acquisition module 401 is further configured to: acquire the user's multimodal biometric data, the multimodal biometric data including at least two types of data characterizing the user's biometrics;

[0189] The feature extraction module 402 is also used to: extract biometric features based on multimodal biometric data to obtain a multimodal biometric vector representing multiple biometric features of the user;

[0190] The determination module 403 is specifically used to: determine the fusion score based on the multimodal biometric feature vector, button feature vector, and touch swipe feature vector.

[0191] In one possible implementation, the determining module 403 is specifically used for:

[0192] Dynamic weight allocation is performed on multimodal biometric feature vectors, button feature vectors, and touch swipe feature vectors;

[0193] The fusion score is obtained by summing the weights based on the dynamic weight allocation results and the multimodal biometric feature vectors, button feature vectors, and touch swipe feature vectors.

[0194] In one possible implementation, the generation module 404 is specifically used for:

[0195] The fusion score is matched with the fusion score range of multiple preset decision intervals;

[0196] If a decision interval is successfully matched as the target decision interval among multiple decision intervals, the authentication result corresponding to the target decision interval will be determined as the user's authentication result.

[0197] The acquisition module continuously acquires user action data generated during operations at bank counters, ATMs, self-service devices, online banking, or mobile banking interfaces, enabling authentication to go beyond a single check before a transaction begins. The feature extraction module generates keystroke feature vectors and touch swipe feature vectors, reflecting both user keystroke habits and touch swipe behaviors.

[0198] The determination module fuses two types of features to determine a fusion score, enabling the assessment of the user's identity based on multi-dimensional operational behavior. The generation module then outputs the identity authentication result based on the fusion score, continuously verifying the consistency of the operator's identity during the transaction process. Therefore, this device enhances the ability to detect abnormal takeovers, malicious manipulations, and abnormal operational states, improving the continuity, accuracy, and security of identity authentication in financial business scenarios.

[0199] The identity authentication device provided in this application embodiment can be used to execute the technical solution of the identity authentication method in any of the above embodiments of this application. Its implementation principle and technical effect are similar, and will not be described again in this embodiment.

[0200] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application, such as... Figure 5 As shown, the electronic device of this embodiment may include: at least one processor 501; and a memory 502 communicatively connected to the at least one processor; wherein the memory 502 stores instructions executable by the at least one processor 501, the instructions being executed by the at least one processor 501 to cause the electronic device to perform the method as described in any of the above embodiments.

[0201] Optionally, the memory 502 can be either standalone or integrated with the processor 501.

[0202] This electronic device can be deployed in bank counter terminals, ATMs, self-service interactive devices, online banking, or mobile banking. The memory stores program instructions used to acquire interface operation data, extract key feature vectors and touch swipe feature vectors, and determine the fusion score. After the processor calls and executes the corresponding instructions, it can continuously analyze the user's key press habits and touch swipe behavior during the transaction process, thus moving beyond the limitations of one-time static authentication before the transaction begins.

[0203] By associating and fusing multiple operational characteristics within the same device, it becomes possible to more accurately identify whether the operator remains consistent. When a terminal is taken over by another party, exhibits abnormal operations, or is subjected to forgery attacks, corresponding identity authentication results are generated promptly, thus improving the continuous authentication capabilities of financial services, the accuracy of identity verification, and transaction security.

[0204] The implementation principle and technical effects of the electronic device provided in this application can be found in the foregoing embodiments, and will not be repeated here.

[0205] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the method of any of the foregoing embodiments.

[0206] This computer-readable storage medium can be deployed in bank counter terminals, ATMs, self-service interactive devices, online banking servers, or mobile banking terminals. When the processor executes computer instructions, it can continuously acquire operation data during user interface key input and touch swipe operations. Based on this, key input feature vectors and touch swipe feature vectors are extracted, and the two types of features are then fused to generate an authentication result, thereby extending one-time static authentication to continuous authentication during the transaction process.

[0207] Because the instructions are embedded in the storage medium in program form, existing financial business architectures can easily integrate this authentication capability. By combining the user's actual operating habits to identify whether the operating entity is consistent, the ability to identify risks such as terminal takeover, abnormal operations, and forgery attacks can be improved, thereby enhancing the authentication reliability and transaction security during financial business transactions.

[0208] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the method of any of the foregoing embodiments.

[0209] This computer program can be stored on a readable medium in bank counter terminals, ATMs, self-service interactive devices, online banking servers, or mobile banking clients. After being invoked by the processor, it controls the collaborative operation of the terminal's interactive interface, identity verification module, and transaction processing module. It continuously acquires operation data when the user performs key input and touch swipes, extracting key input feature vectors and touch swipe feature vectors. Based on these two data points, a fusion score is determined, and an identity authentication result is generated.

[0210] Based on this, the existing one-time static authentication is extended to continuous authentication during the transaction process. This enables the system to determine whether the current operating entity is consistent by combining the user's actual operating habits. In this way, risks can be identified more promptly in scenarios such as terminal takeover, abnormal operation, or forgery attacks, thereby improving the authentication reliability and transaction security during the financial business process.

[0211] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed.

[0212] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application.

[0213] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU) or other general-purpose processors. The processor can also be a Digital Signal Processor (DSP) or an Application Specific Integrated Circuit (ASIC), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0214] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device, and may also be various media that can store program code, such as USB flash drives, portable hard drives, read-only memory (ROM), disks or optical discs.

[0215] The aforementioned storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof. Examples of storage media include Static Random-Access Memory (SRAM) or Electrically Erasable Programmable Read Only Memory (EEPROM).

[0216] Storage media can be, for example, erasable programmable read-only memory (EPROM) or programmable read-only memory (PROM). Storage media can also be read-only memory (ROM), magnetic storage, flash memory, magnetic disks, or optical disks. Storage media can be any available medium accessible to general-purpose or special-purpose computers.

[0217] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. The processor and storage medium can reside within an application-specific integrated circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components within an electronic device or host device.

[0218] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0219] The sequence numbers of the embodiments in this application are merely for description and do not represent the superiority or inferiority of the embodiments. Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0220] Based on this understanding, the technical solution of this application, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0221] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

[0222] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0223] It should be further noted that although the steps in the flowchart are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise explicitly stated in this document, there is no strict order requirement for the execution of these steps, and they can be executed in other orders.

[0224] Furthermore, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but may be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but may be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0225] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0226] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0227] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. An identity authentication method, characterized in that, The method includes: Acquire user action data generated during interface operation; Based on the operation action data, operation action features are extracted to obtain key feature vectors and touch swipe feature vectors. The key feature vectors are used to characterize the operation action features of the user when pressing keys in the interface, and the touch swipe feature vectors are used to characterize the operation action features of the user when touching swipe in the interface. A fusion score is determined based on the button feature vector and the touch swipe feature vector, and the fusion score is used to characterize the degree of confidence that the user is the correct user. The user's identity authentication result is generated using the fusion score.

2. The method according to claim 1, characterized in that, The step of determining the fusion score based on the button feature vector and the touch swipe feature vector includes: Dynamically assign weights to the button feature vector and the touch swipe feature vector; The fusion score is obtained by summing the weights based on the dynamic weight allocation result, the button feature vector, and the touch swipe feature vector.

3. The method according to claim 2, characterized in that, The dynamic weight allocation is a weight allocation based on a preset base confidence level and a preset quality assessment factor; the dynamic weight allocation of the button feature vector and the touch swipe feature vector includes: Based on the basic confidence and quality assessment factors of the button feature vector and the touch sliding feature vector respectively, the DS evidence theory algorithm is used to calculate the first dynamic weight of the button feature vector and the second dynamic weight of the touch sliding feature vector respectively. The step of calculating the fusion score by summing the weights based on the allocation result of the dynamic weight allocation and the button feature vector and the touch swipe feature vector includes: The button feature vector is weighted based on the first dynamic weight, and the touch swipe feature vector is weighted based on the second dynamic weight. The weighted results are then summed to obtain the fusion score.

4. The method according to claim 1, characterized in that, The step of extracting operation action features based on the operation action data to obtain key feature vectors includes: Based on the operation action data, a five-channel keystroke dynamic image model is used to extract operation action features to obtain operation action feature information; wherein, the five-channel keystroke dynamic image model is a model used to extract operation action features under multiple feature dimensions, and the multiple feature dimensions include at least two of the following feature dimensions: key press duration, interval between key release and the next key press, delay between adjacent keys, and virtual pressure sensitivity fluctuation; The key feature vector is obtained by performing matrix analysis based on the operation action feature information.

5. The method according to claim 1, characterized in that, The step of extracting operation action features based on the operation action data to obtain a touch swipe feature vector includes: A realistic trajectory model is generated based on the operation data. The realistic trajectory model includes a trajectory sequence representing the touch sliding operation of multiple sampling points of the user's hand. The simulated trajectory model is encoded using a feature vector encoding model to obtain the touch swipe feature vector.

6. The method according to any one of claims 1-5, characterized in that, The method further includes: The user's multimodal biometric data is acquired, and the multimodal biometric data includes at least two types of data characterizing the user's biometric features; Based on the multimodal biometric data, biometric features are extracted to obtain a multimodal biometric vector representing multiple biometric features of the user. The step of determining the fusion score based on the button feature vector and the touch swipe feature vector includes: The fusion score is determined based on the multimodal biometric vector, the button feature vector, and the touch swipe feature vector.

7. The method according to claim 6, characterized in that, The step of determining the fusion score based on the multimodal biometric vector, the button feature vector, and the touch swipe feature vector includes: Dynamic weight allocation is performed on the multimodal biometric feature vector, the button feature vector, and the touch swipe feature vector; The fusion score is obtained by summing the weights based on the dynamic weight allocation result and the multimodal biometric vector, the button feature vector, and the touch swipe feature vector.

8. The method according to any one of claims 1-5, characterized in that, The step of generating an identity authentication result for the user using the fusion score includes: The fusion score is matched with the fusion score range of multiple preset decision intervals; If a decision interval is successfully matched as the target decision interval among the multiple decision intervals, the identity authentication result corresponding to the target decision interval is determined as the user's identity authentication result.

9. An identity authentication device, characterized in that, The device includes: The acquisition module is used to acquire user action data generated during interface operations; The feature extraction module is used to extract operation action features based on the operation action data to obtain key feature vectors and touch sliding feature vectors. The key feature vectors are used to characterize the operation action features of the user when pressing keys in the interface, and the touch sliding feature vectors are used to characterize the operation action features of the user when touching and sliding in the interface. The determination module is used to determine a fusion score based on the button feature vector and the touch swipe feature vector, wherein the fusion score is used to characterize the degree of confidence that the user is the correct user. The generation module is used to generate an identity authentication result for the user based on the fusion score.

10. An electronic device, characterized in that, include: Memory and processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-8.

12. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method as described in any one of claims 1-8.