Identity authentication method based on dynamic graph network and related device
Through the dynamic graph network authentication model, the interactive data of touch screen sliding events is processed using graph convolutional networks and long-term memory networks, and the problem of poor authentication in the existing technology in diverse scenarios is solved, and more accurate and more generalized identity authentication is achieved, suitable for multiple scenarios and protecting user privacy.
Patent Information
- Application Number
- CN202510583128.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-07
- Publication Date
- 2025-08-19
AI Technical Summary
Existing touch screen-based behavioral biometric technologies are inconsistent in a variety of usage scenarios, making it difficult to capture complex spatiotemporal relationships and sequence dependencies, resulting in poor authentication results and risks of privacy leakage and attacks.
The dynamic graph network authentication model is adopted to capture interactive data of user touch screen sliding events, build a sliding graph sequence, and use graph convolution network and long and short-term memory network to extract feature vectors for identity classification, and build a graph structure to capture the user's time, space and behavior dependencies.
It realizes more accurate and generalized identity authentication, adapts to multiple usage scenarios, does not involve user privacy data, and the user experience is not affected. It can be extended to tasks such as analyzing user gender and identifying user age.
Smart Images
Figure CN120509022A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of identity authentication, and relates to an identity authentication method based on a dynamic graph network and related devices. Background Art
[0002] Currently, authentication methods widely used on mobile devices include passwords, pattern locks, and biometric technologies (such as fingerprint and facial recognition). Although these methods improve convenience, they lack consistent protection and are vulnerable to attacks. For example, passwords and pattern locks are vulnerable to shoulder surfing attacks, while biometric technologies face the risk of privacy breaches and identity theft. Studies have found that these systems can be deceived by highly imitated fingerprints or by means such as photos, videos, and 3D-printed masks. In contrast, behavioral biometrics verifies identity by analyzing a user's unique behavioral patterns, eliminating the need to remember passwords while also enhancing security. It effectively protects against common attack methods and supports continuous authentication, ensuring that only authorized users have continuous access to the device.
[0003] Currently, the challenge facing behavioral biometrics research is that touchscreen-based methods are often limited to specific application scenarios and rely on artificially designed features, which limits their performance in more natural usage environments. While sensor-based and multimodal methods can provide more comprehensive data, they rely on multiple sensors operating at high sampling rates, which not only increases hardware requirements and energy consumption but also fails to adapt to the diversity of user environments, resulting in inconsistent performance across different usage scenarios. These factors all limit the scalability and effectiveness of behavioral biometrics in real-world applications.
[0004] Specifically, existing machine learning-based behavioral biometrics methods mainly rely on manually designed features or temporal feature extraction techniques, which have difficulty capturing the complex spatiotemporal relationships and sequence dependencies in behavioral data. Among them, although LSTM (Long Short-Term Memory Network) can identify nonlinear features in time series, its reliance on sequence structure and progressive information flow limits its ability to handle complex nonlinear relationships. At the same time, although CNN (Convolutional Neural Network) is good at extracting local features, it performs poorly in capturing global dependencies because it ignores overall contextual information, which leads to poor authentication results of existing methods. Summary of the Invention
[0005] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide an identity authentication method and related devices based on dynamic graph network.
[0006] In order to achieve the above object, the present invention adopts the following technical solutions:
[0007] In a first aspect, the present invention provides an identity authentication method based on a dynamic graph network, comprising: collecting interaction data of a plurality of user touch screen sliding events; constructing a single event sliding graph of each user touch screen sliding event based on the interaction data and combining them in time series to obtain a sliding graph sequence; inputting the sliding graph sequence into a pre-trained dynamic graph network authentication model to obtain a user identity authentication result; wherein the dynamic graph network authentication model includes a feature extraction part for extracting feature vectors of the sliding graph sequence and a classification authentication part for performing identity classification based on the feature vector; the feature extraction part includes a graph convolutional network and a long short-term memory network connected in sequence, and the classification authentication part is constructed based on a fully connected module.
[0008] Optionally, the collection of interaction data of several touch screen sliding events of the user includes: collecting feature data of each touch screen sliding in the user's current touch screen sliding event; eliminating incomplete data in the feature data of each touch screen sliding according to a preset elimination rule, and downsampling the feature data of each touch screen sliding by retaining only the first feature data of consecutive identical touch screen coordinates to obtain the interaction data of the user's current touch screen sliding event.
[0009] Optionally, the characteristic data of the touch screen sliding includes one or more of the following: timestamp, screen direction flag, touch screen area, touch screen coordinates, sliding action type and touch screen finger number.
[0010] Optionally, constructing a single-event sliding graph of each touch screen sliding event of the user based on the interaction data includes: constructing a single touch screen sliding graph of each touch screen sliding event of the user based on the interaction data, with the touch screen points as nodes and the sliding paths as edges; wherein the attributes of the nodes include one or more of the following: touch screen coordinates, touch screen area and touch screen finger number; the attributes of the edges include one or more of the following: sliding time, sliding distance, sliding speed, sliding acceleration, sliding angle and sliding action type; the single touch screen sliding graphs of each touch screen sliding event of the user are combined in sequence, and similar nodes are aggregated according to preset node aggregation rules to obtain a single-event sliding graph of each touch screen sliding event of the user.
[0011] Optionally, the node aggregation rules include: Rule 1: nodes in the same single-touch screen sliding graph are not aggregated; Rule 2: when the distance between the node to be combined and the nearest node among the existing nodes is less than a preset distance threshold, the node to be combined and the nearest node among the existing nodes are aggregated; otherwise, no aggregation is performed and an edge is constructed between the node to be combined and the nearest node among the existing nodes; wherein the node to be combined is a node in the current single-touch screen sliding graph to be combined, and the existing node is a node in the combined single-touch screen sliding graph; Rule 3: nodes in the existing nodes that participated in the aggregation process in the previous single-event sliding graph combination process do not participate in the node aggregation process in the current single-event sliding graph combination process.
[0012] Optionally, the graph convolutional network includes several parallel sub-graph convolutional networks, each sub-graph convolutional network including two layers of dynamic sliding graph integration convolutional layers and one layer of global average pooling layer connected in sequence; the fully connected module includes two layers of fully connected networks connected in sequence.
[0013] Optionally, the pre-trained dynamic graph network authentication model optimizes hyperparameters during the pre-training process in the following manner: during the pre-training process, the performance data of the dynamic graph network authentication model on the validation set is obtained; based on the performance data of the dynamic graph network authentication model on the validation set, grid search, random search and Bayesian optimization methods are used to optimize the hyperparameters of the dynamic graph network authentication model.
[0014] In a second aspect, the present invention provides an identity authentication system based on a dynamic graph network, comprising: an acquisition module for collecting interaction data of a plurality of user touch screen sliding events; a graph construction module for constructing a single event sliding graph of each user touch screen sliding event based on the interaction data and combining them in time series to obtain a sliding graph sequence; an authentication module for inputting the sliding graph sequence into a pre-trained dynamic graph network authentication model to obtain a user identity authentication result; wherein the dynamic graph network authentication model includes a feature extraction part for extracting a feature vector of the sliding graph sequence and a classification authentication part for performing identity classification based on the feature vector; the feature extraction part includes a graph convolutional network and a long short-term memory network connected in sequence, and the classification authentication part is constructed based on a fully connected module.
[0015] In a third aspect of the present invention, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned dynamic graph network-based identity authentication method when executing the computer program.
[0016] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned identity authentication method based on a dynamic graph network are implemented.
[0017] Compared with the prior art, the present invention has the following beneficial effects:
[0018] The present invention proposes a dynamic graph network-based identity authentication method. First, interaction data from several user touchscreen sliding events is collected. Based on this interaction data, a single-event sliding graph for each user touchscreen sliding event is constructed, and a sliding graph sequence is generated by combining them in a time series. The dynamic graph sequence is then input into a pre-trained dynamic graph network authentication model, which then outputs the user identity authentication result. This method achieves full identity authentication using only the interaction data generated by the user's touchscreen interaction. Compared to existing identity authentication methods based on behavioral features, the present invention utilizes a more advanced feature construction method, namely a dynamic graph sequence construction method, which constructs a graph structure to comprehensively capture the temporal, spatial, and behavioral dependencies of the user on the touchscreen, thereby more accurately representing the user's touchscreen behavior. This graph structure can deeply mine the complex patterns in the user interaction data, providing a more comprehensive and detailed representation of user behavior for identity authentication. Furthermore, a more advanced dynamic graph network authentication model is employed. Graph convolution and long-short-term memory networks are used to capture the temporal and spatial correlations during the user's sliding process, respectively. Deep sequential patterns are extracted from these graph sequence structures for accurate identity authentication. Ultimately, a more performant and generalizable authentication method is achieved. In addition, the method of the present invention does not involve user privacy data during the data collection process, and can realize implicit collection in the background without affecting the user experience. It is not only suitable for identity authentication scenarios, but also can be used for any tasks that rely on user touch screen interaction behavior. The full-process method can also be extended to interactive data-related tasks such as analyzing user gender and identifying user age. Only the collected labels need to be adjusted accordingly. This provides new possibilities and exploration paths for the field of identity authentication and other related research directions. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 This is a flow chart of an identity authentication method based on a dynamic graph network according to an embodiment of the present invention.
[0020] Figure 2 Schematic diagram of the principle of constructing a sliding graph sequence according to an embodiment of the present invention.
[0021] Figure 3 This is a schematic diagram of the dynamic graph network authentication model framework of an embodiment of the present invention.
[0022] Figure 4 This is a structural block diagram of an identity authentication system based on a dynamic graph network according to an embodiment of the present invention. DETAILED DESCRIPTION
[0023] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.
[0024] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0025] The present invention is described in further detail below with reference to the accompanying drawings:
[0026] See also Figure 1 In one embodiment of the present invention, an identity authentication method based on a dynamic graph network is provided, which is specifically a more flexible and low-power identity authentication solution to adapt to the complexity and dynamic characteristics of user behavior and provide reliable performance in a variety of unrestricted usage scenarios.
[0027] Specifically, the identity authentication method based on the dynamic graph network of the present invention includes the following steps:
[0028] S1: Collect interaction data of several user touch screen sliding events.
[0029] S2: constructing a single event sliding graph of each touch screen sliding event of the user according to the interaction data and combining them according to the time series to obtain a sliding graph sequence.
[0030] S3: Input the sliding image sequence into a pre-trained dynamic graph network authentication model to obtain the user identity authentication result. The dynamic graph network authentication model includes a feature extraction part for extracting feature vectors of the sliding image sequence and a classification authentication part for classifying identities based on the feature vectors. The feature extraction part includes a graph convolutional network and a long short-term memory network connected in sequence, while the classification authentication part is constructed based on a fully connected module.
[0031] The present invention proposes a dynamic graph network-based identity authentication method. First, interaction data from several user touchscreen sliding events is collected. Based on this interaction data, a single-event sliding graph for each user touchscreen sliding event is constructed, and a sliding graph sequence is generated by combining them in a time series. The dynamic graph sequence is then input into a pre-trained dynamic graph network authentication model, which then outputs the user identity authentication result. This method achieves full identity authentication using only the interaction data generated by the user's touchscreen interaction. Compared to existing identity authentication methods based on behavioral features, the present invention utilizes a more advanced feature construction method, namely a dynamic graph sequence construction method, which constructs a graph structure to comprehensively capture the temporal, spatial, and behavioral dependencies of the user on the touchscreen, thereby more accurately representing the user's touchscreen behavior. This graph structure can deeply mine the complex patterns in the user interaction data, providing a more comprehensive and detailed representation of user behavior for identity authentication. Furthermore, a more advanced dynamic graph network authentication model is employed. Graph convolution and long-short-term memory networks are used to capture the temporal and spatial correlations during the user's sliding process, respectively. Deep sequential patterns are extracted from these graph sequence structures for accurate identity authentication. Ultimately, a more performant and generalizable authentication method is achieved. In addition, the method of the present invention does not involve user privacy data during the data collection process, and can realize implicit collection in the background without affecting the user experience. It is not only suitable for identity authentication scenarios, but also can be used for any tasks that rely on user touch screen interaction behavior. The full-process method can also be extended to interactive data-related tasks such as analyzing user gender and identifying user age. Only the collected labels need to be adjusted accordingly. This provides new possibilities and exploration paths for the field of identity authentication and other related research directions.
[0032] In a possible implementation, the collection of interaction data of several touch screen sliding events of the user includes: collecting feature data of each touch screen sliding in the user's current touch screen sliding event; eliminating incomplete data in the feature data of each touch screen sliding according to a preset elimination rule, and downsampling the feature data of each touch screen sliding by retaining only the first feature data of consecutive identical touch screen coordinates to obtain the interaction data of the user's current touch screen sliding event.
[0033] Explanatory, a user data collection system is designed to collect feature data of each touch screen slide in each user touch screen slide event and downsample the collected data to remove duplicate data. In addition, it uses minimal touch screen data (slides) to perform fast user identification without relying on multiple sensors, making it easier to deploy on different mobile devices.
[0034] Exemplarily, the user's data on the display screen and the sensor data are collected, and each collected feature data includes one or more of the following: timestamp, screen direction flag, touch screen area, touch screen coordinates, sliding action type and touch screen finger number.
[0035] Explanatory preprocessing is performed on the collected data to verify its integrity and exclude data that does not meet the standards, such as data with a sliding distance of less than 100 pixels, coordinates outside the reasonable range, or incomplete information. In addition, the 120Hz sampling frequency results in a large amount of duplicate information in the data, and the same touch screen coordinates may be recorded multiple times in a single sliding. Data downsampling is performed by retaining only the data of the first touch screen coordinate of each consecutive identical touch screen coordinate to remove duplicate data.
[0036] In a possible implementation, constructing a single-event sliding graph of each touch screen sliding event of the user based on the interaction data includes: constructing a single-event sliding graph of each touch screen sliding in each touch screen sliding event of the user based on the interaction data, with the touch screen points as nodes and the sliding paths as edges; combining the single-event sliding graphs of each touch screen sliding in each touch screen sliding event of the user in sequence, and aggregating similar nodes through preset node aggregation rules to obtain a single-event sliding graph of each touch screen sliding event of the user.
[0037] Among them, the attributes of the node include one or more of the following: touch screen coordinates, touch screen area and touch screen finger number; the attributes of the edge include one or more of the following: sliding time, sliding distance, sliding speed, sliding acceleration, sliding angle and sliding action type.
[0038] Explanatory, we construct a single-touch screen sliding graph based on the collected interaction data. This graph structure captures the temporal, spatial, and behavioral dependencies of users, providing a more comprehensive representation of their touchscreen behavior. Each swipe on the device creates a sliding path on the touchscreen, forming an edge in the single-touch screen sliding graph, while the touch points serve as nodes in the graph.
[0039] For example, in this embodiment, the attributes of the nodes in the graph structure of the single touch screen sliding graph include the touch screen coordinates (X coordinate, Y coordinate) and the contact area. The X coordinate and Y coordinate are selected as the relative position relationship. At the same time, in order to eliminate the influence of different operating logics in different applications, the starting point of all sliding points is established at the origin (0,0), and the coordinates of subsequent points are adjusted accordingly.
[0040] Edge attributes include time, slide distance, slide speed, slide acceleration, and slide angle. Time is the duration of the slide, slide distance is the length of the path the user slides along on the screen, slide acceleration is determined by comparing the velocity of the current edge with the velocity of the previous edge, and the velocity of the initial edge is assumed to be zero by default. The slide angle is represented by its tan value.
[0041] See also Figure 2 A single touch screen slide event may involve multiple touch screen slides. By sequentially combining the single touch screen slide graphs of multiple touch screen slides, a more complex and comprehensive graph structure, namely the single event slide graph of the touch screen slide event, is constructed. During the combination process, similar nodes from multiple single touch screen slide graphs are aggregated using node aggregation rules to ultimately form a single event slide graph.
[0042] For example, a touch screen sliding event can be divided into time windows of a certain length, or the number of touch screen sliding events in a touch screen sliding event can be set. In this embodiment, the number of touch screen sliding events in a touch screen sliding event is set to 13 times, that is, a single event sliding graph is obtained by combining 13 single touch screen sliding graphs.
[0043] In a possible implementation, the node aggregation rule includes the following rules 1 to 3.
[0044] Rule 1: Nodes in the same single-touch screen sliding graph are not aggregated. Explanatory: In each complete sliding, all points will not be aggregated with other points in the sliding to ensure the integrity of the sliding.
[0045] Rule 2: When the distance between the node to be combined and the nearest node among the existing nodes is less than the preset distance threshold, the node to be combined and the nearest node among the existing nodes are aggregated; otherwise, no aggregation is performed and an edge is constructed between the node to be combined and the nearest node among the existing nodes; where the node to be combined is a node in the current single-touch screen sliding graph to be combined, and the existing node is a node in the combined single-touch screen sliding graph.
[0046] Rule 3: Nodes that participated in the aggregation process in the previous single-event sliding graph combination process among the existing nodes do not participate in the node aggregation process in the current single-event sliding graph combination process.
[0047] Explanatory, by integrating multiple swipes into a unified graph structure using node aggregation rules, the complex spatiotemporal relationships between touch points can be captured.
[0048] For illustrative purposes, after obtaining the single-event sliding graphs for each user's touch screen sliding event, each occurrence of the user's touch screen sliding event is defined as an independent event. That is, the single-event sliding graphs for each touch screen sliding event are sequentially stored in a sequence in chronological order to form a sliding graph sequence. It should be understood that the sliding graph sequence here is a combined sequence of a series of single-event sliding graphs.
[0049] In one possible implementation, the graph convolutional network includes several parallel sub-graph convolutional networks, each sub-graph convolutional network including two layers of dynamic sliding graph integration convolutional layers and one layer of global average pooling layer connected in sequence; the fully connected module includes two layers of fully connected networks connected in sequence.
[0050] Explanation: The dynamic graph network authentication model consists of two parts: feature extraction and classification authentication. The feature extraction part is composed of a graph convolutional network and a long short-term memory network. The graph convolutional network includes several parallel subgraph convolutional networks. The subgraph convolutional network is implemented as a two-layer dynamic sliding graph integration convolutional layer (DSGIConv) and a global average pooling layer.
[0051] Among them, the core mechanism of the DSGIConv layer is its ability to model node features h i (Central Node), h j (adjacent nodes) and edge_attr ij The complex interactions between (edge features) are as follows:
[0052]
[0053] Among them, the output of MLP It captures the deep interdependencies between node and edge features. Subsequently, during the model update process, the output of the MLP is aggregated, and the DSGIConv layer uses an averaging strategy to merge information. This is shown below:
[0054]
[0055] in, is the feature representation layer l of node i, W (l) is the weight matrix, N(i) of the nodes adjacent to the node, A ij The element of the adjacency matrix indicates whether there is an edge between nodes i and j, and σ is a nonlinear activation function.
[0056] After processing through two layers of dynamic sliding graph integration convolutional layers, the output is subjected to global average pooling, aggregating the node features of each graph into a single global feature representation. Since the model input is a graph sequence, each graph in the sequence generates a feature vector and then undergoes convolution to form a new feature vector sequence. This sequence is processed by a long short-term memory network (LSTM) to generate the final deep feature extraction. Ultimately, the deep features are fed into the classification and authentication component to complete the final prediction. The classification and authentication component is constructed based on a fully connected module, which consists of two layers of fully connected networks connected in sequence. The final output result is a two-dimensional probability, which represents the probability of the user's identity being authenticated as the corresponding preset label.
[0057] For example, see Figure 3 The feature extraction part is constructed by layering DSGIConv64, DSGIConv128, and LSTM64, with the activation function being ReLU. The structure of the DSGIConv layer is composed of fc32 and fc32 connected in sequence, with the activation function being ReLU.
[0058] In one possible implementation, the pre-trained dynamic graph network authentication model optimizes hyperparameters during the pre-training process in the following manner: during the pre-training process, performance data of the dynamic graph network authentication model on the validation set is obtained; based on the performance data of the dynamic graph network authentication model on the validation set, grid search, random search, and Bayesian optimization methods are used to optimize the hyperparameters of the dynamic graph network authentication model.
[0059] Explanatory: During the pre-training process, the dataset is divided into a training set and a validation set. The training set is used for initial model training, and the validation set is used to evaluate model performance and generalization. The performance data from the validation set is used to guide the adjustment of model hyperparameters. These hyperparameters, including the learning rate, number of network layers, and number of hidden units, have a significant impact on the model's learning effect and final performance. To find the optimal hyperparameter combination, grid search, random search, and Bayesian optimization methods are used. This method systematically explores the vast hyperparameter space and identifies the parameter configuration that maximizes model performance. Once the optimal hyperparameter settings are determined, the model can be fine-tuned to further improve its accuracy on the training set and its generalization ability on new data, effectively enhancing the model's adaptability to new situations and its predictive accuracy.
[0060] For example, the specific training process is as follows:
[0061] Step 1: Determine a preliminary set of hyperparameters, including learning rate, number of network layers, and number of hidden units, set the training batch size to 64, and input training data to train the entire dynamic graph network authentication model.
[0062] Step 2: Use the validation set performance data to guide hyperparameter tuning. Using a grid search method, systematically traverse the predefined list of candidate hyperparameter values, train and validate the model for each hyperparameter combination, and record the model performance for each parameter combination.
[0063] Step 3: Building on the grid search, we further explore the vast hyperparameter space using random search and Bayesian optimization. Random search selects hyperparameter combinations through random sampling, while Bayesian optimization intelligently selects the next set of hyperparameters based on historical model performance data to maximize model performance.
[0064] Step 4: Through the above steps, identify the parameter configuration that maximizes model performance. After determining the optimal hyperparameter settings, carefully tune the model and train it thoroughly on the training set. This will improve the model's accuracy and enhance its ability to generalize to new data, effectively increasing the model's adaptability to new situations and its predictive accuracy, ultimately training a complete model.
[0065] The adaptive training process of this invention can be applied to the training of neural networks for multiple other classification tasks, showing strong universality. The designed neural network structure has strong generalization and practicality. By using the adaptive training method, model parameters can be systematically optimized, thereby improving the model's accuracy and generalization ability, ensuring that the model performs well on unseen data, while avoiding overfitting and enhancing the model's robustness and reliability in practical applications.
[0066] The following are device embodiments of the present invention, which can be used to perform the method embodiments of the present invention. For details not disclosed in the device embodiments, please refer to the method embodiments of the present invention.
[0067] See also Figure 4 In another embodiment of the present invention, an identity authentication system based on a dynamic graph network is provided, which can be used to implement the above-mentioned identity authentication method based on a dynamic graph network. Specifically, the identity authentication system based on a dynamic graph network includes an acquisition module, a graph construction module and an authentication module.
[0068] Among them, the acquisition module is used to collect the interaction data of several touch screen sliding events of the user; the graph construction module is used to construct a single event sliding graph of each touch screen sliding event of the user according to the interaction data and combine it according to the time series to obtain a sliding graph sequence; the authentication module is used to input the sliding graph sequence into a pre-trained dynamic graph network authentication model to obtain the user identity authentication result; wherein, the dynamic graph network authentication model includes a feature extraction part for extracting the feature vector of the sliding graph sequence and a classification authentication part for performing identity classification according to the feature vector; the feature extraction part includes a graph convolutional network and a long short-term memory network connected in sequence, and the classification authentication part is constructed based on a fully connected module.
[0069] All relevant contents of each step involved in the embodiment of the aforementioned dynamic graph network-based identity authentication method can be referred to the functional description of the functional modules corresponding to the dynamic graph network-based identity authentication system in the embodiment of the present invention, and will not be repeated here.
[0070] The module division in the embodiments of the present invention is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in various embodiments of the present invention may be integrated into a single processor, exist physically as separate modules, or two or more modules may be integrated into a single module. The integrated modules may be implemented in either hardware or software functional modules.
[0071] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the identity authentication method based on the dynamic graph network.
[0072] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the identity authentication method based on the dynamic graph network in the above embodiment.
[0073] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0074] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0075] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.
[0076] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.
[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.
Claims
1. An identity authentication method based on a dynamic graph network, characterized in that: include: Collect interaction data of several user touch screen sliding events; Constructing a single event sliding graph of each user's touch screen sliding event based on the interaction data and combining them in time series to obtain a sliding graph sequence; Input the sliding graph sequence into the pre-trained dynamic graph network authentication model to obtain the user identity authentication result; Among them, the dynamic graph network authentication model includes a feature extraction part for extracting feature vectors of the sliding graph sequence and a classification authentication part for classifying identities according to the feature vectors; the feature extraction part includes a graph convolutional network and a long short-term memory network connected in sequence, and the classification authentication part is constructed based on a fully connected module.
2. The identity authentication method based on dynamic graph network according to claim 1, characterized in that: The interactive data collected from the user's touch screen sliding events includes: Collect feature data of each touch screen slide in the user's current touch screen slide event; According to the preset elimination rules, incomplete data in the feature data of each touch screen slide is eliminated, and the feature data of each touch screen slide is downsampled by retaining only the first feature data of consecutive identical touch screen coordinates to obtain the interaction data of the user's current touch screen slide event.
3. The identity authentication method based on dynamic graph network according to claim 2, characterized in that: The characteristic data of the touch screen sliding includes one or more of the following: timestamp, screen direction flag, touch screen area, touch screen coordinates, sliding action type and touch screen finger number.
4. The identity authentication method based on dynamic graph network according to claim 1, characterized in that: The constructing of a single event sliding graph of each touch screen sliding event of the user according to the interaction data includes: Based on the interaction data, construct a single touch screen slide graph for each touch screen slide in each touch screen slide event of the user, with the touch screen points as nodes and the slide paths as edges; wherein the attributes of the nodes include one or more of the following: touch screen coordinates, touch screen area, and touching finger number; and the attributes of the edges include one or more of the following: slide time, slide distance, slide speed, slide acceleration, slide angle, and slide action type; The single touch screen sliding graphs of each touch screen sliding event of the user are sequentially combined, and similar nodes are aggregated according to a preset node aggregation rule to obtain the single event sliding graph of each touch screen sliding event of the user.
5. The identity authentication method based on dynamic graph network according to claim 4, characterized in that: The node aggregation rules include: Rule 1: Nodes in the same single-touch screen sliding graph are not aggregated; Rule 2: When the distance between the node to be combined and the closest existing node is less than a preset distance threshold, the node to be combined and the closest existing node are aggregated; otherwise, no aggregation is performed and an edge is constructed between the node to be combined and the closest existing node; the node to be combined is a node in the current single-touch screen sliding graph to be combined, and the existing node is a node in the combined single-touch screen sliding graph; Rule 3: Nodes that participated in the aggregation process in the previous single-event sliding graph combination process among the existing nodes do not participate in the node aggregation process in the current single-event sliding graph combination process.
6. The identity authentication method based on dynamic graph network according to claim 1, characterized in that: The graph convolutional network includes several parallel sub-graph convolutional networks, each of which includes two layers of dynamic sliding graph integration convolutional layers and one layer of global average pooling layer connected in sequence; the fully connected module includes two layers of fully connected networks connected in sequence.
7. The identity authentication method based on dynamic graph network according to claim 1, characterized in that: During the pre-training process, the pre-trained dynamic graph network authentication model optimizes hyperparameters in the following ways: During the pre-training process, obtain the performance data of the dynamic graph network certification model on the validation set; Based on the performance data of the dynamic graph network authentication model on the validation set, grid search, random search and Bayesian optimization methods are used to optimize the hyperparameters of the dynamic graph network authentication model.
8. An identity authentication system based on a dynamic graph network, characterized in that: include: The collection module is used to collect interaction data of several touch screen sliding events of users; A graph construction module is used to construct a single event sliding graph of each user's touch screen sliding event based on the interaction data and combine them in time series to obtain a sliding graph sequence; The authentication module is used to input the sliding image sequence into the pre-trained dynamic graph network authentication model to obtain the user identity authentication result; Among them, the dynamic graph network authentication model includes a feature extraction part for extracting feature vectors of the sliding graph sequence and a classification authentication part for classifying identities according to the feature vectors; the feature extraction part includes a graph convolutional network and a long short-term memory network connected in sequence, and the classification authentication part is constructed based on a fully connected module.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the identity authentication method based on a dynamic graph network as described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the identity authentication method based on a dynamic graph network as claimed in any one of claims 1 to 7 are implemented.