E-commerce platform security authentication system and method

Abstract

The invention relates to the technical field of safety certification, in particular to an electronic commerce platform safety certification system and method. The method comprises the following steps: acquiring platform login identity information, performing field migration and recombination, constructing an identity information tracing chain, performing multi-domain splitting on the tracing chain, generating a domain segmentation identity chain, performing identity risk assessment based on the multi-domain splitting, generating risk parameters, and performing parallel fusion authentication according to the risk parameters and the domain segmentation identity chain. The method comprises the steps of generating an initial security authentication result, logging in through a disguise platform, collecting login environment change data, verifying environment authenticity and performing security judgment to obtain login environment security data, performing platform login re-authentication by utilizing the security data and the initial authentication result, updating a login state and establishing a repeated authentication mechanism. According to the invention, a set of more comprehensive security protection system is formed through a multi-level security authentication strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of security authentication, and particularly to a security authentication system and method for an e-commerce platform. Background Art

[0002] Traditional security authentication methods for e-commerce platforms often rely on static authentication and simple password protection, which leads to frequent security vulnerabilities. Especially in the face of complex network attacks, the response speed and adaptability of the system are insufficient. Existing technologies lack real-time monitoring capabilities in dynamic environments and cannot identify and respond to environmental changes in a timely manner. Especially when users log in, the impact of environmental factors on security is not fully considered, resulting in many potential risks not being discovered in a timely manner. Traditional authentication mechanisms often cannot adapt to diverse login scenarios, resulting in poor user experience and lack of security guarantee. The management of identity information and risk assessment lack systematicness, increasing security risks. Summary of the Invention

[0003] Based on this, it is necessary to provide a security authentication system and method for an e-commerce platform to solve at least one of the above technical problems.

[0004] To achieve the above object, a security authentication method for an e-commerce platform includes the following steps:

[0005] Step S1: Obtain the platform login identity information; perform field migration and recombination on the platform login identity information to obtain migrated field data; construct an identity information traceability chain based on the migrated field data;

[0006] Step S2: Perform multi-domain splitting on the identity information traceability chain to obtain a domain-segmented identity chain; perform identity risk assessment based on the domain-segmented identity chain to generate identity information risk parameters;

[0007] Step S3: Perform parallel fusion authentication based on the identity information risk parameters and the domain-segmented identity chain to generate an initial security authentication result;

[0008] Step S4: Perform platform login disguise based on the initial security authentication result, and continuously collect login environment change data; verify the authenticity of the environment based on the login environment change data, and perform security determination based on the environment authenticity verification data to obtain login environment security data;

[0009] Step S5: Perform platform login re-authentication based on the login environment security data and the initial security authentication result to obtain secure repeated authentication data; update the platform login status based on the secure repeated authentication data, and construct a repeated authentication mechanism.

[0010] Through obtaining the platform login identity information and performing field migration and recombination, the system can construct a complete identity information traceability chain, laying a foundation for subsequent security authentication. The generation of the domain-segmented identity chain makes the risk assessment for different domains more refined, thereby improving the security and accuracy of identity information. The implementation of identity risk assessment can timely identify potential security hazards, thus generating identity information risk parameters, providing a scientific basis for subsequent authentication. The parallel fusion authentication based on the risk parameters and the domain-segmented identity chain ensures the reliability of the initial security authentication result and enhances the overall security level of the platform. The strategy of disguising platform login effectively prevents the malicious acquisition of identity information. At the same time, by continuously collecting data on changes in the login environment, the system can monitor the security of the environment in real time. The process of verifying the authenticity of the environment ensures the legality of the login request. The generated login environment security data provides strong support for subsequent security determination. Based on these security data and the initial authentication result, login re-authentication is performed, establishing a dynamic security authentication mechanism that can respond to environmental changes in a timely manner and update the login status of the platform. The implementation of constructing a repeated authentication mechanism provides stronger security protection for the e-commerce platform, preventing identity theft and illegal acts. The overall system forms a comprehensive security protection system through multi-level security authentication strategies, greatly enhancing the protection of user information, enhancing users' trust in the platform, and promoting the healthy development of e-commerce. With the continuous progress of technology, this method can also be continuously optimized in practical applications to adapt to new security challenges, providing continuous guarantee for the security and stability of the e-commerce environment, and finally forming an efficient and intelligent e-commerce security authentication system, creating a more secure trading environment for users and promoting the sustainable development of the digital economy.

[0011] The present invention also provides an e-commerce platform security authentication system for implementing the e-commerce platform security authentication method as described above. The e-commerce platform security authentication system includes:

[0012] An identity traceability module, configured to obtain the platform login identity information; perform field migration and recombination on the platform login identity information to obtain migrated field data; construct an identity information traceability chain based on the migrated field data;

[0013] A risk assessment module, configured to perform multi-domain splitting on the identity information traceability chain to obtain a domain-segmented identity chain; perform identity risk assessment based on the domain-segmented identity chain to generate identity information risk parameters;

[0014] A fusion authentication module, configured to perform parallel fusion authentication according to the identity information risk parameters and the domain-segmented identity chain to generate an initial security authentication result;

[0015] The login disguise module is used to perform platform login disguise based on the initial security authentication result, and continuously collect data on changes in the login environment; verify the authenticity of the environment based on the data on changes in the login environment, and perform security determination based on the data on environmental authenticity verification to obtain login environment security data;

[0016] The re-authentication module is used to perform platform login re-authentication based on the login environment security data and the initial security authentication result to obtain secure repeated authentication data; update the platform login status based on the secure repeated authentication data, and construct a repeated authentication mechanism.

[0017] Through the application of the identity traceability module in the present invention, the system can accurately obtain and reconstruct platform login identity information, form a complete identity information traceability chain, providing a solid foundation for subsequent authentication. The risk assessment module can refine risk identification and enhance the security of identity information by performing multi-domain splitting on the identity information traceability chain. The generated identity information risk parameters provide a scientific basis for authentication decisions. The introduction of the fusion authentication module ensures the effective combination of the identity information risk parameters and the domain-segmented identity chain, and can generate a reliable initial security authentication result. The login disguise module effectively prevents identity information from being maliciously obtained through disguise strategies, and continuously collects data on changes in the login environment to ensure real-time monitoring of environmental security. Verifying authenticity based on the data on environmental changes can promptly identify illegal login requests, and the generated login environment security data provides strong support for security determination. The re-authentication module performs dynamic re-authentication based on the login environment security data and the initial security authentication result, ensuring timely update of the login status. The constructed repeated authentication mechanism enhances the security protection ability of the system, preventing identity theft and other illegal acts. The overall system forms a comprehensive security protection system through multi-level security strategies, greatly enhancing the protection of user information, enhancing users' trust in the platform, and promoting the healthy development of e-commerce. With the continuous progress of technology, the system can adapt to new security challenges, provide continuous security guarantees, and ultimately form an efficient and intelligent e-commerce security authentication system, creating a safer trading environment for users, promoting the sustainable growth of the digital economy, laying a solid foundation for future development, enhancing the overall security of the platform and the user experience, and ensuring a high level of security protection in a complex network environment. Description of the Drawings

[0018] Figure 1 It is a schematic diagram of the step flow of a security authentication method for an e-commerce platform;

[0019] Figure 2 It is a schematic diagram of the detailed implementation step flow of step S3;

[0020] The realization, functional features, and advantages of the object of the present invention will be further described with reference to the embodiments and the drawings. Detailed implementation mode

[0021] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those skilled in the art within the scope of the present invention without creative work belong to the scope of protection of the present invention.

[0022] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0023] It should be understood that although terms such as "first" and "second" may be used here to describe each unit, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit can be called the second unit, and similarly the second unit can be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed associated items.

[0024] To achieve the above object, please refer to Figures 1 to 2 , a security authentication method for an e-commerce platform, comprising the following steps:

[0025] Step S1: Obtain platform login identity information; perform field migration and recombination on the platform login identity information to obtain migrated field data; construct an identity information traceability chain based on the migrated field data;

[0026] Step S2: Perform multi-domain splitting on the identity information traceability chain to obtain a domain-segmented identity chain; perform identity risk assessment based on the domain-segmented identity chain to generate identity information risk parameters;

[0027] Step S3: Perform parallel fusion authentication according to the identity information risk parameters and the domain-segmented identity chain to generate an initial security authentication result;

[0028] Step S4: Perform platform login disguise based on the initial security authentication result, and continuously collect login environment change data; verify the authenticity of the environment based on the login environment change data, and perform security determination based on the environment authenticity verification data to obtain login environment security data;

[0029] Step S5: Perform platform login re - authentication based on the login environment security data and the initial security authentication result to obtain secure repeated authentication data; update the platform login status based on the secure repeated authentication data and construct a repeated authentication mechanism.

[0030] By obtaining the platform login identity information and performing field migration and recombination, the system of the present invention can construct a complete identity information traceability chain, laying a foundation for subsequent security authentication. The generation of the domain - segmented identity chain enables more refined risk assessment for different domains, thereby improving the security and accuracy of identity information. The implementation of identity risk assessment can timely identify potential security hazards, thus generating identity information risk parameters, providing a scientific basis for subsequent authentication. The parallel fusion authentication based on risk parameters and the domain - segmented identity chain ensures the reliability of the initial security authentication result and enhances the overall security level of the platform. The strategy of disguising platform login effectively prevents malicious acquisition of identity information. At the same time, by continuously collecting data on changes in the login environment, the system can monitor the security of the environment in real - time. The process of verifying the authenticity of the environment ensures the legality of the login request. The generated login environment security data provides strong support for subsequent security determination. Performing login re - authentication based on these security data and the initial authentication result establishes a dynamic security authentication mechanism that can respond to environmental changes in a timely manner, update the platform login status, and the implementation of constructing a repeated authentication mechanism provides stronger security protection for the e - commerce platform, preventing identity theft and illegal acts. The overall system forms a comprehensive security protection system through multi - level security authentication strategies, greatly enhancing the protection of user information, enhancing users' trust in the platform, and promoting the healthy development of e - commerce. With the continuous progress of technology, this method can be continuously optimized in practical applications to adapt to new security challenges, providing continuous guarantee for the security and stability of the e - commerce environment, and ultimately forming an efficient and intelligent e - commerce security authentication system, creating a more secure trading environment for users and promoting the sustainable development of the digital economy.

[0031] In the embodiment of the present invention, the method for security authentication of an e - commerce platform includes the following steps:

[0032] Step S1: Obtain platform login identity information; perform field migration and recombination on the platform login identity information to obtain migrated field data; construct an identity information traceability chain based on the migrated field data;

[0033] In this embodiment, by deploying an identity access collection module at the entrance of the e-commerce platform and using the HyperLogLog structure to quickly compress and extract the features of the input identity identification data, the identity identification data includes a total of 6 types of field data such as the device unique identifier (IMEI), browser User-Agent (user agent information), public network IP address, local MAC address hash value, historical order number summary, and payment channel authentication record index. Subsequently, based on the field migration and recombination rule set, cross-mapping processing is performed on these 6 types of data according to strongly correlated feature pairs. The field migration rules include establishing a belonging domain mapping between the IP and the MAC address, forming a tracking mapping between the device IMEI and the payment channel record, and constructing a browsing device behavior fusion chain between the User-Agent and the historical order number. After generating the set of migrated field data, the set is input into the multi-level traceability graph construction engine, and the generation of the identity information traceability chain is completed by constructing a dynamic traceability node graph. The traceability graph takes the central node as the current login ID and spreads 3 layers of historical nodes outward. The maximum branch width is limited to 10 concurrent identity segments, and each segment contains three field structures: a tracking path, an identity source, and a historical interaction label. After generation, it is written into the authentication cache pool.

[0034] Step S2: Perform multi-domain splitting on the identity information traceability chain to obtain a domain-segmented identity chain; perform identity risk assessment based on the domain-segmented identity chain to generate an identity information risk parameter;

[0035] In this embodiment, all tracking nodes in the identity information traceability chain are read, and category clustering processing is performed on the "interaction label" field in the nodes. The K-Modes classification algorithm is used to aggregate the node labels in the non-numerical domain. The initial clustering centers are set to 5 types of label centers, namely transaction label, browsing label, payment label, exception label, and registration label. After clustering, the traceability chain is split into 5 domain-segmented identity chain subsets according to the domain to which the label belongs. Each subset is limited to a node interval of 10 to 50. Subsequently, a behavior frequency vector group is constructed based on the identity behavior sequence, and each sub-chain is converted into a feature vector with a fixed length of 128 dimensions and input into a pre-trained risk recognition model for evaluation. The model uses a two-tower structure Transformer framework. The left tower inputs the identity behavior vector, and the right tower inputs the historical attack template features. The output is a set of risk matching score intervals. The risk score range is divided into four categories, namely L0 (low) corresponding to the score interval [0, 0.25], L1 (medium-low) for (0.25, 0.5], L2 (medium-high) for (0.5, 0.75], and L3 (high) for (0.75, 1.0]. Finally, the risk level results matched by each domain-segmented identity chain are output as an identity information risk parameter data set.

[0036] Step S3: Perform parallel fusion authentication according to the identity information risk parameter and the domain-segmented identity chain to generate an initial security authentication result;

[0037] In this embodiment, the risk parameter dataset and the domain-segmented identity chain are loaded, and all identity nodes with a risk level above L1 are extracted, grouped by domain and split into no more than 5 groups of parallel processing paths. The maximum node hop count for each group of paths is limited to no more than 3, and the time window is limited to a minimum time interval of 1.2 seconds for each interaction. Subsequently, according to the weight calibration rule, the nodes in all paths are weighted and calibrated according to the risk level. The weight of L0 is 0.25, L1 is 0.5, L2 is 0.75, and L3 is 1.0. The weighted calibration result is written into the fusion matrix, and the matrix dimension is set to 5 rows and 128 columns, representing the 128-dimensional identity behavior embedding of each of the 5 paths. The fusion operation adopts a weighted average fusion method to generate a weighted identity fusion vector by stacking according to the dimension, and is input into the authentication decision model for credibility inference. This model is a combined model of a CNN structure and a GRU sequence verification structure, and the output initial security authentication result is trusted (credibility ≥ 0.7) or to be verified (credibility < 0.7), and the result is input into the subsequent authentication branch.

[0038] Step S4: Based on the initial security authentication result, perform platform login disguise, and at the same time continuously collect data on changes in the login environment; verify the authenticity of the environment based on the data on changes in the login environment, and perform security determination based on the data on the authenticity of the environment to obtain login environment security data;

[0039] In this embodiment, if the initial security authentication result is in a to-be-verified state, the system automatically activates the platform login disguise module. Through the permission control gateway, the platform only opens the home page display interface and the historical order query interface, prohibits operations such as commodity purchase, payment, account modification, and binding. At the same time, it records all user operation behaviors and activates the login environment collection module. The collected content includes the current public network IP, browser fingerprint Canvas Hash, operating system type, screen resolution, time zone offset, mouse trajectory curve, page stay time series, window focus switching event. The collection frequency is once every 3 seconds for 120 seconds. Subsequently, the above collected data is compared with the historical environment trajectory data when the user passed authentication in the past. The comparison algorithm uses the DTW (Dynamic Time Warping) variant structure matching method to generate environment trajectory traceability data. Extract the multi-source node conflict events in it, that is, the event points where the same account appears in different network regions at the same time. Further extract the mutation rate that appears in the conflict nodes, that is, the behavior points where the mouse event interval fluctuation exceeds the threshold of 0.7. Determine whether there is an environment disguise behavior. If the mutation rate continues to exceed the critical value in more than 5 behavior segments, the environmental authenticity level is judged to be abnormal, and environmental authenticity verification data is generated. Subsequently, based on the trust evaluation module, this data is input into the dynamic trust scoring model for trust grading. The scoring rule generates a trust index by weighting the historical trust stability factor and the current mutation level. An index greater than 0.6 is high trust, and less than 0.3 is low trust. Finally, it is compared with the built-in trust threshold of 0.45 of the platform to determine whether the environmental security condition is met. If it passes, login environment security data is generated to continue the subsequent steps.

[0040] Step S5: Based on the login environment security data and the initial security authentication result, perform platform login re-authentication to obtain secure repeated authentication data; update the platform login status based on the secure repeated authentication data and construct a repeated authentication mechanism.

[0041] In this embodiment, the environmental security data and the initial security authentication result are written into the re-authentication determination engine as input parameters. The re-authentication determination engine is a stacked XGBoost regression model plus a logistic regression discriminant model. The model input features are 36 dimensions in total, including 16 dimensions of identity embedding features, 8 dimensions of risk level mapping features, and 12 dimensions of environmental behavior similarity indicators. The output is the final security repeated authentication data. If the data value is greater than 0.85, it is considered to be a trusted user login. The system removes the disguise restriction, restores all operation permissions and updates the login status, binds the current login identifier to the current environment fingerprint set, and constructs a repeated authentication mechanism. That is, when the user logs in again in the future, the system will extract the bound environment fingerprint and perform fuzzy matching with the current environment. The matching method is the local sensitive hash (LSH) method embedded in the authentication pipeline control. If the similarity drops by more than 0.3, it will automatically enter the fast re-authentication path. The repeated authentication mechanism data is stored in the authentication status update table and the authentication behavior history table for subsequent calls.

[0042] Preferably, step S1 comprises the following steps:

[0043] Step S11: obtaining platform login identity information; extracting field semantic features in the platform login identity information;

[0044] Step S12: reconstructing field mapping of field semantic features to obtain migration field data;

[0045] Step S13: segment the migrated field data and perform source tracing and nesting to generate segmented source tracing data, wherein the segmentation standard is that the login time interval is greater than 15 minutes, and the nesting method is two-level nesting, the first level is the user ID, the second level is the time segment sequence, and each segment contains 1-10 groups of field change records;

[0046] Step S14: Chain tracing of the segmented traceability data is performed, and an identity information traceability chain is constructed, wherein the traceability chain length is set in the range of 3-128 nodes.

[0047] In this embodiment, an access data listener based on the Nginx reverse proxy structure is deployed at the platform authentication entry to obtain all inbound request data from the user browser and the client in real time. The extracted fields include 10 types of original fields: browser User-Agent string, device unique identifier IMEI, public network IP address, local timestamp, language preference field, operating system type, UUID token, MAC address hash value, resolution parameter, and port protocol feature. Subsequently, a field semantic model built on the BERT-base structure is used to perform feature expression extraction operations on the above fields. The model input is in the form of field text and the current platform behavior context sequence, and the output is the semantic embedding vector of each field. The embedding dimension is fixed at 64 dimensions. By calculating the cosine similarity in the vector space, the implicit association categories between fields are confirmed, and the fields are classified into identity categories (such as IMEI, MAC, UUID), behavior categories (such as User-Agent, language preference), and environment categories (such as IP, timestamp, resolution) according to the clustering center grouping. The classification results are written into the field semantic index table. The classified fields in the field semantic index table are read, the identity category fields are mapped to core identification fields, the behavior category fields are mapped to variable feature fields, and the environment category fields are mapped to scenario auxiliary fields. A mapping reconstructor based on the rule tree structure is used to combine the same type of fields in the order of priority to generate a migration field structure. Each structure format is in the form of a triple, namely "core identification + variable feature + scenario auxiliary". For example, IMEI + User-Agent + IP constitutes a set of migration field data. The priority is set according to the field stability ranking, and the stability score is calculated based on the frequency and change rate of the field in the historical interaction records. The frequency is based on the number of occurrences in the interaction within the most recent 30 days, and the change rate is based on the proportion of unique values appearing daily. Fields with a field stability score greater than 0.7 can be used as the starting fields for combination. After reconstruction, no more than 10 sets of migration field data are generated for each login behavior. The field group numbering rule constructs a unique identifier in the way of adding a hash truncation to the current timestamp. The structure format is "TS_H(ID)", that is, the timestamp plus the hash value. All the generated migration field data are temporarily stored in the migration field buffer table and marked as pending traceability status. The full records within the past 72 hours in the migration field buffer table are read, and each record is initially grouped according to the unique identifier of the user. Within each user group, the records are sorted in ascending order according to the timestamp field of the field record. The time interval between adjacent records is calculated. If the interval is greater than 15 minutes, it is divided into independent time periods as the second-level nested time series unit. The sliding window method is used to compare the field changes, and the change points of the fields in each segment are marked. The change point determination criterion is that the change rate of the User-Agent string is greater than 0.2 or the IP address spans across provincial network segments. The format of the field change record is "field name: old value → new value". Each segment allows 1 to 10 groups of field change records to be recorded. If the field change frequency exceeds 2 times, this segment is marked as a high-frequency change segment, and the field "HF" is added for the priority processing strategy in subsequent chained processing. The nested result of each segment constructs a JSON nested structure with nested keys. The first-level key is the user unique identifier, and the second-level key is the time period number, with the number format of "D + sequence number", such as D1, D2, D3. Load all the time period nested data corresponding to the user identifiers from the nested structure library, disassemble the field change record fields into node units, construct a node structure for each change record, and the node contains five contents: field name, value before change, value after change, change timestamp, and paragraph number. Connect the nodes in the order of time from front to back according to the time period to form a chained structure. The minimum length of each chain is 3 nodes, and the maximum length is limited to 128 nodes. If the number of recorded nodes exceeds 128, the most recent 128 nodes are used as the truncation range for chain construction. The chained connection method constructs an adjacency list based on a two-way dictionary structure, and each node records the upstream and downstream change logic pointers for quickly tracing the path backtracking and multi-segment merging analysis. The naming rule for the chain identifier is "UID_CHAIN_HASH", where UID is the user identifier and HASH is the field summary fingerprint. The fingerprint calculation method is to generate a SHA-256 hash value after connecting the field names and old and new values of each node in sequence. All the constructed identity information traceability chains are written into the traceability chain database, and at the same time, the chain active time window, field change density, and abnormal change frequency are recorded as one of the risk input variables in the subsequent authentication process.

[0048] Preferably, the multi-domain splitting of the identity information traceability chain in step S2 includes:

[0049] Perform multi-domain label annotation on the identity information traceability chain to obtain the attributes of each node in the traceability chain;

[0050] Divide the attributes of each node in the traceability chain into multiple business domains to generate a multi-domain labeled traceability chain;

[0051] Identify the connection relationships between nodes in the multi-domain labeled traceability chain;

[0052] Discriminate the domain data to which the connection relationships between nodes belong, and extract the intra-domain interaction nodes based on the domain data to which the relationships belong;

[0053] Identify the cross-domain information flow breakpoints in the connection relationships between nodes according to the intra-domain interaction nodes;

[0054] Locate the boundary domain data of the identity information traceability chain through the information flow breakpoints;

[0055] Distinguish multiple business domains based on the boundary domain data to obtain the domain-segmented identity chain.

[0056] In this embodiment, each identity information traceability chain is extracted from the identity behavior database. The field name, field value, operation type, and timestamp in each node are used as inputs, and a label generation engine based on the combination of regular expressions and logical rules is called for processing. Four types of identifiers, namely behavior labels, geographical labels, device labels, and identity labels, are matched for the field name and field value. The labels are generated by comparing the field value format recognition with the field context source table, and the labels are written into the tag field in the node structure. Behavior labels such as login, payment, and jump are identified by the operation type field. Geographical labels such as the IP field match the country code. Device labels such as the UA field match the terminal model regular library. Identity labels such as the UUID field match the uniqueness rule to determine whether there is a correlation. All nodes need to complete the label marking of all fields. Fields that fail to be marked are treated as unrecognized and do not participate in subsequent domain partitioning. Read the label attributes of all nodes, encode the label set of the nodes into a 128-dimensional vector, and generate the average vector of the label words through the trained FastText model. The DBSCAN algorithm is used to cluster all nodes, with the density radius set to 0.45 and the minimum number of core points set to 5. Each cluster corresponds to a business domain, and the nodes within the business domain are areas with similar label densities. After clustering, each node is assigned one or more business domain numbers, and the business domain numbers are added as fields to the node attributes. When there are multiple label distributions in a node, it can belong to multiple domains, and nodes belonging to multiple domains will be determined as cross-domain boundary points during subsequent connection relationship processing. Establish connection edges for each node in the order of the timestamp in the traceability chain. The connection method is to form a directed edge between the current node and the next time node. Each edge is associated with the label domain attributes of the source node and the target node. The connection edges are saved in the memory structure for subsequent flow relationship analysis. The connection relationship is established only on the premise that the time is continuous and the source accounts are the same. If there is a node pair with a time jump exceeding the set threshold of 10 minutes, no connection will be established. Judge whether the domain attributes of all established node pairs are consistent. If the domain numbers of the source node and the target node are exactly the same, mark this node pair as an "intra-domain interaction node pair". If there are multiple cross-domains between the two nodes but no exactly the same domain, mark this node pair as a "multi-domain connection". If the two nodes do not belong to the same business domain at all, mark it as a "cross-domain connection". Group all "intra-domain interaction node pairs" by domain number to obtain the set of interaction nodes within each domain. Screen out the connection edges with changed field labels from the "cross-domain connection" node pairs, and give priority to matching the change of behavior labels as the core judgment basis. For example, connecting from a login behavior label to a payment behavior label, or changing from an identity device label to an IP label, is regarded as an information flow break point. If the UA field value is "iPhone13,If "4" changes to "PC-Windows10" and the behavior label changes from browsing to payment, then the connection edge is marked as a breakpoint. The breakpoint position records the node number and the changed field. All breakpoints are aggregated to form a breakpoint node set. The breakpoint set is used to determine the splitting positions between different business domains in the traceability chain. Analyze the domain numbers to which each node in the breakpoint node set belongs, and count the domain numbers to which the breakpoint nodes in each traceability chain belong. If a domain contains three or more breakpoint nodes or two consecutive breakpoints fall within this domain, then this domain is marked as a boundary domain. The boundary domain records the mapping relationship between the traceability chain number and the corresponding domain number, which is used as the basis for breakpoint partitioning in subsequent chain splitting operations. The breakpoint domain is marked as an internal boundary or an external boundary based on the connection direction. If the source node is in this domain, it is an internal boundary; if the target node is in this domain, it is an external boundary. Extract the positions of the breakpoint nodes from the boundary domain mapping, and split each identity information traceability chain in chronological order. Each split starts and ends with a breakpoint node. The internal nodes of the split sub-chain segments must all come from the same business domain or its sub-domains, and cannot be combined across domains. All split chain segments are re-numbered, and the relationship between the original chain number and the chain segment number is retained. If a chain segment contains only two nodes and there is no continuous label behavior change, it is discarded. The remaining chain segments are used as domain-separated identity chains in subsequent identity verification or risk identification steps. Each chain segment retains the original node fields and label information unchanged, and is used as an independent authentication unit in subsequent behavior pattern analysis and security judgment.,

[0057] Preferably, the identity risk assessment based on the domain-separated identity chain in step S2 includes:

[0058] Extract the chain structure features in the domain-separated identity chain;

[0059] Perform integrity verification on the domain-separated identity chain according to the chain structure features, so as to obtain integrity verification data;

[0060] Identify the cross-domain interaction behavior data in the domain-separated identity chain;

[0061] Perform abnormal cross-domain behavior monitoring on the cross-domain interaction behavior data based on the preset cross-domain behavior baseline data, so as to obtain cross-domain behavior monitoring data;

[0062] Quantify the identity credibility of the domain-separated identity chain through the integrity verification data and the cross-domain behavior monitoring data, so as to generate identity information risk parameters.

[0063] In this embodiment, the segment node information is read one by one from the domain-segmented identity chain, and the timestamp, behavior label, connection edge between nodes, and cross-domain mark of each node are parsed. The structural features such as the number of nodes in the segment, average time interval, behavior label diversity index, number of cross-domain edges, and maximum length of consecutive single-domain nodes are calculated. A directed graph model of the segment is constructed using a graph network analysis tool, and the node correlation degree and structural importance distribution are calculated using a Graph Convolutional Network (GCN). The structural features are output in the form of a multi-dimensional vector, and the dimensions include the number of nodes, number of edges, average node degree, average shortest path between nodes, node label entropy, etc. All feature values are directly calculated from the original segment data without using any inference or completion. Based on the extracted chain structural features, an integrity index system is set up, including the number of nodes threshold, time continuity score, cross-domain edge ratio limit, behavior label coverage rate, etc. The threshold rule engine is used to perform range verification on each index in the structural features. The number of nodes shall not be less than 5, the average time interval shall not exceed 15 minutes, the cross-domain edge ratio shall not exceed 20%, and the behavior label coverage rate shall not be less than 70%. The segments that exceed the index range are marked as integrity anomalies. The integrity verification data is stored as a combination of the verification score of each segment and the boolean value of the anomaly mark. The verification score is calculated using the weighted sum of the normalized structural features, and the weight coefficients are adjusted according to the risk model in the training data to 0.3 for the number of nodes, 0.25 for time continuity, 0.25 for the cross-domain edge ratio, and 0.2. The sum of all weights is 1. Traverse all the connecting edges within the segment, filter out the node pairs where the domain numbers of the two ends of the connecting edges are different, extract the behavior tags, operation times, interaction frequencies, and interaction durations of these cross-domain connections, and construct a cross-domain interaction behavior data table. The field structure of the table includes the source node ID, target node ID, cross-domain type (such as login-payment, payment-jump), interaction time difference, behavior tag category. Use a rule-based behavior pattern matching algorithm to classify the interaction behaviors. The interaction frequency is counted by time period (minute level), and the interaction duration is accumulated based on the time stamp difference. All cross-domain behavior data is multi-dimensionally aggregated in the time dimension and behavior category dimension. Use the historical real user cross-domain behavior data to train a cross-domain behavior baseline model. The baseline data includes the normal frequency interval, interaction time distribution, and behavior sequence pattern of each type of cross-domain interaction behavior. Use the Gaussian Mixture Model (GMM) to estimate the probability density of the current cross-domain interaction behavior data, and mark the behavior frequency or duration exceeding the confidence interval (99% confidence level) as an abnormal behavior. Combine the Markov Chain model to analyze the probability that the behavior sequence does not conform to the normal sequence. All abnormally marked behaviors generate abnormal behavior logs. The fields of the abnormal behavior logs include the abnormal type, abnormal degree score, and time period. The abnormal score is calculated based on the deviation from the baseline distance. The cross-domain behavior monitoring data is structured and stored. Perform multi-index fusion on the scores and abnormal marks in the integrity verification data and the abnormal scores in the cross-domain behavior monitoring. Use the weighted linear combination method. The risk parameter calculation formula is: identity risk parameter = (1 - integrity score) × 0.6 + cross-domain abnormal score × 0.4. The value range of the risk parameter is from 0 to 1, where 0 represents the lowest risk and 1 represents the highest risk. The weights are determined based on historical risk event statistics. In the fusion process, use the Min-Max normalization method to standardize the values of each index. The risk parameter is saved to the identity information risk library, associated with the corresponding domain segmentation identity chain identifier, and used as a reference for the decision threshold of the subsequent security authentication module. All calculations are completed based on real-time collected data without any manual intervention. The risk parameter update period is once per hour.

[0064] Particularly important is that step S3 includes the following steps:

[0065] Step S31: Archive the risk level of the identity information risk parameter to obtain risk-labeled identity data;

[0066] Step S32: Split the heterogeneous nodes of the domain segmentation identity chain to generate parallel node identity chains, where each parallel node is limited to having an independent identity behavior segment with no more than 3 call hops and a time interval of not less than 1.2 seconds;

[0067] Step S33: Weight calibration and recombination are performed on the parallel node identity chain data according to the risk-annotated identity data, so as to obtain weighted identity fusion data, where weighted hierarchical fusion is performed according to the fixed weight ratio of L0 = 0.25, L1 = 0.5, L2 = 0.75, and L3 = 1.0;

[0068] Step S34: Authentication determination is performed based on the weighted identity fusion data, so as to generate an initial security authentication result.

[0069] In this embodiment, based on the numerical range of the identity information risk parameter, the risk parameter is divided into four levels, namely L0, L1, L2, and L3. The specific threshold intervals are [0, 0.25), [0.25, 0.5), [0.5, 0.75), and [0.75, 1.0] respectively. The threshold comparison algorithm is used to traverse all the identity risk parameters, and each parameter value is compared with the threshold interval. After matching, the corresponding level label is attached to the identity record. The risk level annotation data structure contains the identity identifier, the risk parameter value, and the risk level label. All risk level information is saved in a structured format. The risk level archiving operation is triggered and executed based on the update of the real-time risk parameter. Traverse all nodes in the domain-segmented identity chain, and analyze the call jump number and the time stamp difference between adjacent nodes in turn. The call jump number is defined as the number of system call levels spanned in the operation process from the current node to the next node. The time interval is defined as the time stamp difference between adjacent nodes. The window sliding algorithm is used to split the chain segment starting from the head of the chain. The number of nodes in each chain segment shall not exceed 3, and the time interval between adjacent nodes shall not be less than 1.2 seconds. If there are nodes with a time interval less than 1.2 seconds in the chain, a new chain segment is formed by disconnecting. After the splitting is completed, all the chain segments are arranged in parallel to form a parallel node identity chain data set. Each parallel chain segment in the data structure saves the node sequence, the time stamp range, and the call jump number statistical information. The splitting algorithm traverses the chain nodes through depth-first search (DFS) and combines the time threshold for chain segment division. Extract the identity identifiers corresponding to each chain segment in the parallel node identity chain, associate the risk level annotation data through the identity identifier, and obtain the risk level label corresponding to each chain segment. Assign the fixed weight coefficients as 0.25 for L0, 0.5 for L1, 0.75 for L2, and 1.0 for L3. Perform weighted processing on the identity behavior characteristics of all nodes in the parallel chain segment. The calculation formula is: fused feature = ∑(weight × node feature vector), where the weight is determined by the risk level, and the node feature vector consists of the behavior label, the time stamp information, and the call jump number feature of the node. The weighted summation algorithm is used to calculate the fused feature. All the fused data of the chain segments are uniformly converted into a vector format for storage, which is used for subsequent authentication determination. The weight calibration process ensures that nodes with a high risk level contribute a greater weight. The fusion operation is completed through parallel matrix multiplication. Input the weighted identity fusion data into a pre-trained security authentication discriminant model. The model adopts a multi-layer perceptron (MLP) structure, including an input layer, two hidden layers, and an output layer. The number of nodes in the hidden layers is 128 and 64 respectively. The activation function is ReLU, and the output layer uses the Sigmoid function to calculate the authentication probability value. The input of the model is the fused feature vector, and the output is a binary classification probability value. The probability value is greater than 0.If it is determined that the authentication is passed, otherwise it is determined as authentication failure. The authentication result includes the authentication status and the corresponding probability value, and the result is saved to the secure authentication result database. The determination process combines batch normalization technology to improve the model stability. The model is trained using the labeled historical authentication data set, and the model is updated regularly after training. The determination calculation process is accelerated based on CPU multi-threading processing.

[0070] In particular, step S34 includes:

[0071] Each identity segment in the weighted identity fusion data is split into fields according to the fusion weight, so as to obtain a weighted identity feature sequence;

[0072] The feature similarity comparison is performed on the preset platform authorized identity trajectory data and the weighted identity feature sequence to generate a similarity comparison matrix;

[0073] The similarity comparison matrix is mapped to the weighted identity fusion data to generate a fusion confidence score;

[0074] The authentication determination is performed through the fusion confidence score to generate an initial secure authentication result.

[0075] In this embodiment, for each identity segment in the fusion data, its fusion weight value is extracted. The fusion weight ranges from 0.25 to 1.0. According to the fusion weight, the multi-dimensional feature fields in the identity segment are weighted and split. When splitting, the identity behavior feature vector is decomposed into several field units, and each field unit corresponds to a single behavior feature, such as a timestamp, a call jump count, and a behavior category identifier. The weighted function is used to multiply the fusion weight by the field unit feature value to form a new weighted feature field. The generated weighted identity feature sequence is a vector set arranged in an orderly manner of the weighted fields. The splitting operation is completed by combining vector decomposition technology with linear algebra operations to ensure that each weighted field accurately reflects the relationship between the corresponding identity behavior feature and its weight. In the data structure, the weighted identity feature sequence records the field index, the weighted value, and the original behavior label. The field splitting algorithm ensures a weighted accuracy of 0.001 based on the fusion weight gradient calculation method. During the processing, a batch processing mechanism is adopted to process multi-segment data in parallel, and the platform-authorized identity trajectory data is converted into a feature vector with the same dimension as the weighted identity feature sequence. The trajectory data includes user operation timestamps, operation types, device identifiers, and geographical location information. The feature normalization method is used to perform normalization processing on the trajectory data and the weighted identity feature sequence respectively. The normalization range is set from 0 to 1 to eliminate the influence of dimensions. The cosine similarity calculation algorithm is used to compare the two feature sets pairwise, and the cosine value of the cosine angle between each pair of feature vectors is calculated to obtain the feature similarity value. The calculation formula is cosine similarity = the dot product of vector A and vector B divided by the product of the norms of the two vectors. The similarity value ranges from -1 to 1, where 1 indicates exactly the same and -1 indicates exactly the opposite. All comparison results are arranged to generate a similarity comparison matrix. The rows of the matrix represent the platform-authorized trajectory feature vectors, and the columns represent the weighted identity feature sequence. The matrix elements store the similarity values of the corresponding feature pairs. The comparison operation is executed in parallel using a high-performance matrix calculation library to ensure that the matrix calculation time does not exceed 50 milliseconds. Based on the corresponding relationship between the row and column indexes of the similarity comparison matrix, several elements with the highest similarity in the matrix are extracted. The threshold is set to 0.75, and the feature pairs with similarity values greater than the threshold are selected. The corresponding weighted identity fusion data is weighted and summarized, and the weight is linearly distributed according to the similarity value. The fusion confidence score is calculated as the weighted average. The fusion confidence score range is set from 0 to 1, and the scoring formula is score = ∑(similarity × weight) / ∑weight, where the weight is normalized according to the similarity value. The mapping operation is implemented through an index mapping table, which is composed of the identity segment index and the feature field index. The hash mapping algorithm is used to quickly locate the corresponding data. The fusion confidence score data structure includes the identity segment identifier, the confidence score value, and the corresponding feature field set. The calculation process uses vectorized operations to improve the calculation efficiency. The fusion confidence score is input into the decision threshold comparison module, and the preset decision threshold is 0.7. If the score is higher than the threshold, the authentication is passed; if it is lower than the threshold, the authentication fails. The authentication result is stored in binary identification, accompanied by the fused confidence score value as a reference index. The determination module is implemented by fast comparison operations. The authentication result includes an authentication status field and a corresponding fused confidence score field. The result is written into the secure authentication log database. The log includes a timestamp, an identity segment identifier, and the authentication result. The determination process combines multi-threaded asynchronous processing technology to ensure that the response time is controlled within 100 milliseconds.

[0076] Preferably, based on the initial secure authentication result, platform login disguise is performed, and at the same time, data on changes in the login environment is continuously collected, including:

[0077] Construct a disguised identity token based on the initial secure authentication result;

[0078] Perform a temporary identity mapping on the disguised identity token to obtain temporary disguised identity data;

[0079] Isolate the temporary disguised identity data in a session sandbox to generate sandbox session data;

[0080] Construct a virtual login environment based on the sandbox session data;

[0081] Based on the virtual login environment, continuously collect login IP data and login device parameters to obtain a multi-point login IP dataset and a collection of login device parameters;

[0082] Perform geographical location drift detection on the multi-point login IP dataset to obtain geographical location drift data;

[0083] Evaluate the device change rate based on the collection of login device parameters;

[0084] Integrate the geographical location drift data and the device change rate to generate data on changes in the login environment.

[0085] In this embodiment, the binary authentication status and the fused confidence score data of the initial security authentication result are used as inputs, and the authentication result is encrypted using a hash encryption algorithm. The SHA-256 hash algorithm is used to generate a 256-bit fixed-length disguised identity token, which contains a unique identity identifier and timestamp information. The timestamp records the generation time in the UTC (Coordinated Universal Time) format. The token generation process is completed based on the hardware acceleration of the encryption module, ensuring that the generation speed is less than 10 milliseconds. The token format follows the JSON Web Token (JWT) specification and consists of three parts: a header, a payload, and a signature. The hash value of the authentication result data is embedded in the payload, and the signature part is digitally signed using a private key to prevent tampering. The token structure is stored in a high-performance cache system, and the cache expiration time is set to 30 minutes. A temporary identity mapping operation is performed on the generated disguised identity token. The mapping engine module is called, and the mapping engine adopts a rule-based mapping strategy. The rule contains the mapping relationship between the token hash value and the internal temporary identity identifier of the platform. The mapping rules are stored in a relational database, and SQL query statements are used for quick matching. After decoding and verifying the signature part of the disguised identity token through the mapping algorithm, the authentication information in the token payload is extracted, and the corresponding temporary identity identifier is queried from the mapping rule library based on the authentication information. The temporary identity identifier is associated with the token to generate temporary disguised identity data, which includes a temporary identity ID, a session start time, an identity status identifier, and permission tags. The data structure is designed in a key-value pair format. The sandbox session management module is started, and the temporary disguised identity data is used as session initialization parameters. Virtualization technology is called to create an independent and isolated running environment. The Docker container technology is used to build an isolated sandbox. The container image is pre-installed with a virtual operating system and a platform authentication environment. A unique session ID is assigned, and the temporary disguised identity data is written into the container in-memory database. The sandbox network isolation policy is set to restrict the sandbox from accessing the external network. A resource restriction mechanism is enabled to control the allocation of resources such as CPU, memory, and storage. The sandbox session data includes a session ID, a container instance ID, a running status, and resource allocation details. The process status inside the sandbox is monitored in real time, and the sandbox session data is synchronously updated to the main database at regular intervals. The sandbox startup time is controlled within 100 milliseconds. Using the session ID and the container instance ID in the sandbox session data, the virtual environment construction interface is called through the session management service to dynamically generate a user virtual login environment, which includes a virtual desktop, a virtual browser, and an operation agent that simulates user interaction. The virtual environment parameters include the operating system type and version number, virtual hardware configuration information, and a user access log recording module. The simulator engine is based on the VNC (Virtual Network Computing) protocol to implement remote virtual desktop interaction. The virtual login environment supports multi-process concurrent operations. During the construction process, a configuration management tool is used to automatically deploy the required components. The virtual environment construction time does not exceed 150 milliseconds. After completion, the environment status information is written into the environment status management database.The environmental status includes the login status, activity timestamp, and security policy identifier. A network monitoring module is embedded in the virtual login environment to listen for and record login IP address information in real time. The login IP data includes IPv4 and IPv6 addresses, port numbers, and connection timestamps. The data collection frequency is set to sample once per second. The device parameter collection module obtains the hardware parameters of the logged-in device by calling the operating system API interface, including the device model, operating system version, browser version, CPU architecture, memory capacity, and security patch level. The collected data is stored in a structured JSON format. All login IP and device parameter data are aggregated to form a multi-point login IP dataset and a collection of logged-in device parameters. The data is stored in a time series database, which supports high-concurrency writing and querying. Data collection continues in real time, with an update frequency of up to real time and a latency of no more than 100 milliseconds. The IP address information in the multi-point login IP dataset is input into the geographical location parsing module. The module calls the IP geographical location database (GeoIP database) to perform IP location parsing, calculates the change in geographical coordinates corresponding to the IP address between consecutive time points, and uses the geographical distance calculation formula (based on the spherical distance calculation method) to calculate the distance between two points in meters. If the distance exceeds the set threshold of 5000 meters, it is recorded as a geographical location drift event. The drift detection algorithm combines timestamps and continuity judgments to generate geographical location drift data. The data format includes the start time, end time, start location coordinates, end location coordinates, and drift distance. Drift events are stored in a relational database table, and the trigger mechanism is used to achieve real-time data updates. The collection of logged-in device parameters is arranged in a time series. Key hardware indicators such as device model and operating system version are extracted, and the sliding window technique is used to compare changes in device parameters within a time window length of 24 hours. If the device model or operating system version changes, it is counted as a device change event. The device change rate is calculated as the ratio of the number of device change events within a unit time to the time window length. The device change rate is represented as a floating point number, accurate to three decimal places. The evaluation process is based on the device parameter change detection algorithm, which determines changes by comparing changes in device parameter hash values. All detection results are stored in the device change log database. The log includes timestamps, change types, and parameter details before and after the change. The data integration engine is called, and the geographical location drift data and device change rate are used as inputs. The time series alignment algorithm is used to synchronize the time axes of the two types of data, with the alignment error controlled within 5 seconds. The integration method uses the weighted fusion algorithm, with the drift event weight set to 0.6 and the device change weight set to 0.4. The comprehensive score of the login environment change is calculated. The score range is set from 0 to 1, and the calculation formula is comprehensive score = 0.6 × normalized drift index + 0.4 × normalized device change rate. Normalization uses the min-max normalization method. The integration result forms a structured dataset, including the comprehensive score, timestamp, and source data summary. The integrated data is written into the environmental change database, which supports high-concurrency querying and data writing.The integrated process is automatically executed, and the completion time is controlled within 200 milliseconds.

[0086] Preferably, verifying the authenticity of the environment based on the login environment change data in step S4 and making a security determination based on the environment authenticity verification data includes:

[0087] Tracing the environmental historical trajectory of the login environment change data to obtain environmental trajectory tracing data;

[0088] Locating multi-source node conflicts in the environmental trajectory tracing data and extracting suspicious node data;

[0089] Identifying the time series mutation rate in the suspicious node data;

[0090] Judging the environmental disguise of the login environment change data based on the time series mutation rate to obtain environmental authenticity verification data;

[0091] Performing dynamic trust grading based on the environmental authenticity verification data to generate trust grading environment data;

[0092] Making an environmental security determination on the login environment change data based on a preset trust threshold and the trust grading environment data to obtain login environment security data.

[0093] In this embodiment, the login environment change data is input into the trajectory analysis module. The trajectory analysis module retrieves the historical trajectory corresponding to the login environment change data in chronological order by calling the historical environment change records in the time series database and using the timestamp indexing function. It adopts a trajectory tracing algorithm based on a graph database to map the environment change data into a temporal graph structure of nodes and edges. The nodes represent the environmental states at different time points, and the edges represent the chronological order and association strength of the state changes. The nodes in the graph structure contain information such as the login IP address, geographical location coordinates, and device parameter snapshots. The depth-first search (DFS) algorithm is used to traverse the graph structure to trace the evolution path of the environmental state. During the tracing process, a hash table is used to store the visited nodes to prevent loops. The environmental trajectory tracing data consists of a trajectory path set and a node attribute set. The environmental trajectory tracing data is input into the conflict detection module. The conflict detection module compares the multi-source information at the same time node in the trajectory graph to determine whether there is data inconsistency among the multi-source nodes. Data inconsistency includes cases where the login IP ownership does not match, there are differences in device parameter versions, and the geographical location coordinates differ by more than 1000 meters. The heterogeneous data fusion technology is used to fuse and process the multi-source node information, and the weighted voting method is used to determine the mainstream state of the nodes. The heterogeneous data fusion weights are set according to the reliability of the data sources, with the weight range from 0 to 1, and the mainstream state weight threshold is set to 0.7. Nodes below this threshold are marked as conflict nodes, and all conflict nodes form a set of suspicious node data. The conflict detection result output includes the conflict node ID, conflict type, data fields involved, and timestamp. The conflict location process uses the Apache Flink streaming computing framework for real-time processing, with a processing delay less than 100 milliseconds. The suspicious node data is input into the mutation detection module. The module calculates the temporal mutation rate based on the time series analysis method, adopts the sliding window technology, with the window size set to 60 minutes and the window step size to 5 minutes, and calculates the coefficient of variation (CV) of the environmental change indicators within the window. The coefficient of variation is equal to the ratio of the standard deviation to the mean of the data within the window, which is used to quantify the degree of change fluctuation. The mutation rate is defined as the absolute value of the difference in the coefficient of variation between two consecutive time windows. If the mutation rate exceeds the preset threshold 0.3, it is determined to be an abnormal mutation. The mutation rate result is output in a time series data format, including timestamp, mutation rate value, and corresponding suspicious node ID. The mutation detection process is implemented based on Python data analysis library Pandas and NumPy. The computing efficiency reaches 10,000 data per second. The mutation identification process takes 150 milliseconds. The mutation rate data is used as input to call the camouflage detection model, which is based on machine learning classification algorithm and adopts support vector machine (SVM). Machine) performs binary classification on the time series mutation rate. The training data set contains two types of samples: normal environment change rate and camouflage environment change rate. During the training process, the kernel function is adjusted to radial basis function (RBF), and the penalty parameter C is adjusted to 1.0. The model outputs environmental authenticity verification data, including environmental authenticity score and judgment label. The score range is 0 to 1. The closer the score is to 1, the more realistic the environment is. The judgment label is 0 for camouflage environment and 1 for real environment. The model is deployed on the GPU accelerated computing platform, and the inference time is controlled within 50 milliseconds. The format of environmental authenticity verification data is structured JSON. The field contains statistical information of score, label and input mutation rate. The environmental authenticity verification data is input into the trust grading module. The module performs dynamic grading based on the grading rule engine. The grading rule includes multi-layer trust level division. The level interval is defined as 0 to 1. The specific levels are divided into low trust interval [0.0, 0.3), medium trust interval [0.3, 0.7], and high trust interval [0.7, 1.0]. The rule engine adopts Based on the Drools rule management system, the rule engine calls the environment authenticity score field for conditional judgment, matches the corresponding level according to the score, and dynamically adjusts the level threshold based on the historical trust level change trend. The trust grading result includes the trust level label, level score and grading timestamp. The result data structure uses XML format. The trust grading process takes less than 100 milliseconds. The trust grading environment data is compared with the preset trust threshold. The preset threshold uses a floating point number of 0.5 as the judgment basis. The threshold is stored in the security policy configuration file. The system calls the configuration management module to load the threshold. The comparison process uses a high-performance Boolean judgment operation. If the trust level score is greater than or equal to the threshold, the environmental security judgment result is safe, otherwise it is unsafe. The judgment result is recorded in the security status log. The log field includes the judgment time, trust level score, judgment result and related environmental change data snapshot. The security judgment process is recorded through a transactional database to ensure data consistency and integrity. The response time of the entire judgment process is less than 80 milliseconds. .

[0094] It is particularly important to judge the environment camouflage based on the time series mutation rate of the login environment change data, including:

[0095] Construct a sequence change trajectory model based on the time series mutation rate;

[0096] Perform change time aggregation on each node window of the sequence change trajectory model and identify the mutation window range;

[0097] Calculate the behavior rate curve of the mutation window range;

[0098] Identify the environmental camouflage patterns in the mutation window range according to the behavior rate curve, and perform formaldehyde scoring on all environmental camouflage patterns based on a preset environmental camouflage template, so as to obtain the camouflage matching score;

[0099] Evaluate the authenticity of the logged-in environment change data based on the camouflage matching score and a preset matching score threshold, so as to obtain the environmental authenticity verification data.

[0100] In this embodiment, the time-series mutation rates extracted from the login environment change data are arranged in chronological order to form an initial time series. Normalization processing is applied to this series to calculate the normalized mutation rate at each time point. The normalization uses the min-max scaling method, that is, each rate value is subtracted by the minimum value of the entire series and then divided by the difference between the maximum value and the minimum value. The normalized data is used to eliminate the data scale differences of different users and time periods. Subsequently, the sliding window technique is used, with the window length set to 1800 seconds (30 minutes) and the step size set to 300 seconds (5 minutes), to segment the normalized time series. The mean, variance, and kurtosis indicators are calculated for all mutation rate values within the sliding window. The mean reflects the overall change level, the variance reflects the fluctuation amplitude, and the kurtosis reflects the sharpness of the mutation. A node is generated for each window, and directed edges are established between the nodes in chronological order. The edge weight is defined as the absolute value of the difference in the means of two adjacent nodes. This graph structure is used to represent the time trajectory of the mutation rate change. The trajectory model consists of a node set and an edge set. The node attributes include the mean, variance, and kurtosis, and the edge attribute is the weight. In the sequence change trajectory model, for the mutation rate values within the time period covered by each node, a time-weighted average is performed. The weight is determined according to the reciprocal of the distance from the central time point of the node. The closer the distance, the greater the weight. The specific weight calculation is weight = 1 / (|time point - central time| + 1 second). The mutation rates at all time points within the window are multiplied by the corresponding weights, summed, and then divided by the sum of the weights to obtain the weighted average change intensity. The threshold is set to 0.6. Nodes with a change intensity exceeding this threshold are marked as mutant nodes. Then, all mutant nodes are traversed, and the interval merging algorithm is used to merge the time intervals of adjacent or overlapping mutant nodes to obtain continuous mutant window ranges. Specifically, when implementing, the timestamps of all mutant nodes are sorted, and the time intervals between adjacent nodes are compared in turn to see if they are less than the set merging interval of 5 minutes. If satisfied, they are merged; otherwise, a new interval is started. Eventually, several mutant time periods are formed. After extracting the time series mutation rate data of all time points within the mutant window range, the cubic spline interpolation method is applied to fit the discrete time point data into a continuous curve. During the interpolation process, a cubic polynomial function is constructed to ensure that the function values and the first and second derivatives of the fitting curve are continuous at each data point, ensuring a smooth curve. The abnormal peaks are removed from the interpolated curve. An abnormal peak is defined as a point where the local value exceeds 3 times the average value of its neighborhood window (10 seconds before and after). After removing the abnormal peaks, interpolation fitting is performed again. Then, the first derivative of the curve at each time point is calculated. The derivative is calculated using the central difference method, i.e., the derivative value = (the rate difference between adjacent points before and after) divided by the time interval. The derivative accuracy is controlled within one ten-thousandth of a second. The derivative curve reveals the rate change trend. The entire process is completed with high-precision floating-point operations. The output behavior rate curve data includes timestamps, rate values, and derivative values. The rate value sequence of the behavior rate curve is input into the hidden Markov model for state decoding. The hidden Markov model contains 5 hidden states, corresponding to environmental changes of different intensities. Using the trained transition probability matrix and emission probability matrix, the optimal state sequence is calculated through the Viterbi algorithm. The abnormal camouflage state in the state sequence indicates that there is a camouflage behavior in the environment. The time period corresponding to this state is marked as the environmental camouflage mode. Subsequently, the rate change characteristics of all camouflage modes are extracted, such as frequency, amplitude, and duration. The similarity between these characteristics and the typical camouflage behavior characteristics stored in the environmental camouflage template library is calculated. The cosine similarity is used as the metric. Cosine similarity = the dot product of two feature vectors divided by the product of their norms. All template matching scores are weighted and summed to form the formaldehyde score. The weights are set according to the typicality and occurrence frequency of the camouflage behavior. The scoring range is from 0 to 100. The higher the score, the more obvious the camouflage characteristics. The scoring calculation process involves matrix operations and vector normalization. After obtaining the formaldehyde score, it is numerically compared with the preset matching score threshold. The threshold is set to 40 and is stored in the security policy configuration file. The threshold is read through the configuration management module. The authenticity assessment is completed by judging whether the score is higher than the threshold. If the score is lower than 40, the environment is determined to be in the real state; otherwise, it is determined to be in the camouflage state. The authenticity assessment result is represented in the form of a label, and the label value is 0 or 1. The value 0 represents camouflage, and 1 represents real. The assessment result, the score value, and the corresponding timestamp are encapsulated into a structured data format and sent to the subsequent security decision-making module. The entire assessment process is based on fast access to in-memory data and is implemented using boolean judgment and integer comparison operations to ensure that the response time of the assessment process is within the millisecond range.

[0101] Preferably, the platform login re-authentication based on the login environment security data and the initial security authentication result in step S5 includes:

[0102] Performing fusion authentication based on the login environment security data and the initial security authentication result to obtain security re-authentication data. When the initial security authentication result is that the security authentication fails and the login environment security data indicates that the login environment is insecure, the security re-authentication data is defined as authentication insecure;

[0103] When the initial security authentication result is that the security authentication passes and the login environment security data indicates that the login environment is insecure, the security re-authentication data is defined as authentication ambiguous;

[0104] When the initial security authentication result is that the security authentication passes and the login environment security data indicates that the login environment is secure, the security re-authentication data is defined as authentication secure.

[0105] In this embodiment, the login environment security data and the initial security authentication result are respectively represented in the form of binary codes. The encoding of the login environment security data where 1 represents environmental security and 0 represents environmental insecurity, and the encoding of the initial security authentication result where 1 represents authentication passed and 0 represents authentication failed. Both are input into the multi-input logic determination module, and a state mapping table is used for fusion determination. The design of the state mapping table includes three mapping relationships. The rows of the mapping table represent the initial security authentication result, the columns represent the login environment security data, and the fused authentication status code is stored in the table. During the fusion determination process, according to the input binary combination, the corresponding fused status value in the mapping table is directly searched. The specific content of the mapping table is: input (0,0) outputs 0 indicating insecure authentication, input (1,0) outputs 2 indicating ambiguous authentication, input (1,1) outputs 1 indicating secure authentication. The mapping table is loaded through a configuration file during system initialization. The mapping table structure is implemented using a two-dimensional array. The determination module quickly matches the mapping relationship based on Boolean logic to ensure that the fusion determination result is generated within microseconds. The fusion authentication module performs a state assignment operation according to the status code output by the mapping table. Define the secure repeated authentication data as a structure containing a status code, a status description, and a timestamp. The status code corresponds to the output of the mapping table. The status description field is predefined as a string variable. Status code 0 corresponds to "insecure authentication", status code 1 corresponds to "secure authentication", and status code 2 corresponds to "ambiguous authentication". The timestamp field obtains the current Unix timestamp by calling the system time acquisition function, and the timestamp is accurate to the second. During the generation of the entire structure, dynamic allocation of the memory heap space is used to ensure the minimization of the data structure occupancy. The state assignment is implemented through a conditional judgment structure. After determining the input status code, the corresponding description string is assigned to the structure field, and the assignment process uses a memory copy function to prevent string buffer overflow. The generated secure repeated authentication data structure is passed to subsequent modules for processing and storage through a pointer. When the encoding of the input initial security authentication result is 0 and the encoding of the login environment security data is 0, the fusion authentication module returns the status code 0 according to the mapping table, and then calls the state assignment sub-module to generate the secure repeated authentication data. The data status description field is set to "insecure authentication". At this time, the system records this secure repeated authentication data to trigger the subsequent security warning mechanism. The warning mechanism module regularly scans the status code in the secure repeated authentication data structure in a polling manner. When it finds that the status code is 0, it immediately calls the event push interface. The interface uses message queue technology to send the warning information to the security management center. The message queue adopts the first-in-first-out strategy to ensure the timeliness of the warning information. The timestamp field of the secure repeated authentication data is used to record the time when the authentication event occurs. The log management system persists and stores this event in the JSON format, which includes three fields: status code, status description, and timestamp. When the encoding of the input initial security authentication result is 1 and the encoding of the login environment security data is 0, the fusion authentication module returns the status code 2,The status description field of the secure repeated authentication data is assigned as "authentication ambiguous". This status is identified as an authentication risk warning. The authentication risk warning management module triggers a risk level adjustment operation based on status code 2. The risk level value is based on a floating-point number, with an initial value of 0.5. After the warning is triggered, the risk level value is increased to 0.8. The risk level is dynamically adjusted by combining historical risk data through a weighted average method. The risk level adjustment result is fed back to the authentication policy adjustment module. The adjustment module selects further identity verification policies based on the risk level, such as adding SMS verification code verification. The SMS verification code is generated and sent by calling an external SMS service API. The verification result is fed back to update the secure repeated authentication data, and the status description field is updated synchronously. The entire processing process is managed by an event-driven mechanism. The event listener asynchronously captures status change events. When the initial secure authentication result code input is 1 and the login environment security data code is 1, the fusion authentication module outputs status code 1. The status description field of the secure repeated authentication data is assigned as "authentication secure". This authentication result is stored in the authentication record database. The database uses a relational database system and inserts the authentication record using SQL statements. The authentication record includes the user unique identifier, authentication timestamp, authentication status code, and description field. The database connection uses connection pool technology to improve the writing efficiency. The authentication result is used to support subsequent business logic judgments, such as allowing normal access to the core function module of the e-commerce platform. The authentication status is pushed to the permission management module synchronously. The permission management module generates a user access token based on the authentication result. The access token uses the JSON Web Token (JWT) format and contains encrypted information of the authentication status. The token is encrypted by the symmetric encryption algorithm AES to ensure the security of data transmission. After authentication, the relevant data is sent through the HTTPS protocol to ensure the confidentiality and integrity of the data during network transmission.,

[0106] Preferably, updating the platform login status based on the secure repeated authentication data and constructing a repeated authentication mechanism in step S5 includes:

[0107] Updating the platform login status in real time through the secure repeated authentication data. When the secure repeated authentication data is "authentication insecure", cancel the login status and return to the platform login interface;

[0108] When the secure repeated authentication data is "authentication ambiguous", restrict sensitive operations of the platform login identity, send a security alert to a preset bound contact, and at the same time collect the feedback information of the bound contact and perform operation authorization based on the feedback information;

[0109] When the secure repeated authentication data is "authentication secure", call the camera module and fingerprint recognition chip for a hardware loopback verification, and perform operation authorization based on the hardware loopback verification result;

[0110] Construct a repeated authentication mechanism according to the above steps and embed the repeated authentication mechanism into the e-commerce platform.

[0111] In this embodiment, for the received secure repeated authentication data, the status code field therein is parsed. The status codes are divided into three types: authentication insecure, authentication ambiguous, and authentication secure. First, it is determined whether the status code is authentication insecure. If it is satisfied, the status revocation function in the login status management module is called to mark the current user session as an invalid state. This session state is indexed by the session ID and is queried and updated in the session management database. The status update operation adopts an atomic transaction mechanism to ensure data consistency. After the status update is completed, the user interface control module is called to trigger an interface refresh command. The interface refresh is completed through the front-end JavaScript event-driven method. The refresh operation redirects the user to the platform login interface URL. The URL link parameters contain an error code and a status description, which are used to display a login failure prompt on the front end. The entire login status cancellation operation takes less than 100 milliseconds to ensure the smoothness of the user experience. When the status code of the secure repeated authentication data is determined to be authentication ambiguous, the login status management module updates the current user session to a restricted access state. A list of predefined sensitive operation permissions is included in the restricted access state. Sensitive operations include fund transfer, account information modification, password reset, etc. The sensitive function entry is dynamically blocked through the permission control list. The entry disabling is achieved by modifying the properties of the front-end DOM (Document Object Model) nodes. The relevant buttons are set to the disabled state and are accompanied by a prompt text. The prompt text is pushed from the back-end permission management module to the front end and is automatically updated in the UI after being captured by the front-end event listener. Then, the message push module is called. The preset contact information of the user is queried through the bound contact information database. The contact information includes the phone number and the email address. The message content is formatted into a multi-field text message containing the security alert type, user identification, timestamp, and description of the current operation restrictions. The message push module sends security alerts through the SMS service interface and the SMTP server of the email transmission protocol respectively. After the sending is completed, the push module starts a feedback collection mechanism. The feedback collection is completed by listening to the preset reply channels. The SMS reply interface monitors the reply content in real time. The email reply interface pulls the emails regularly, parses the keywords in the reply content to match the authorized keywords. The set of authorized keywords is stored in the permission configuration file. After the authorized keyword matching is successful, the operation authorization module is called. The module unlocks the corresponding sensitive operation permissions for the current session by calling the permission management API. The permission adjustment is reflected in the form of an update to the role permission table. After the permission adjustment is completed, the status is synchronized back to the secure repeated authentication data structure to ensure the consistency of the status data. When the status code of the secure repeated authentication data is determined to be authentication secure, the login status management module marks the current user session as a fully authorized state. The fully authorized state maps to the corresponding highest permission level in the permission management system. This permission level includes all operation permissions and unrestricted access to all business function modules. The permission management system is based on role-based access control (RBAC,Based on the Role - Based Access Control model, the permission set corresponding to this state in the permission table is loaded into the user session token. The user session token uses the JSON Web Token format. The token embeds a permission claim field, and the token is encrypted with a symmetric key to ensure secure transmission. The user request verifies the token permission in the backend access control module. All operation requests are subject to permission verification based on the permission fields within the token. The permission verification module reduces the database access frequency through a caching mechanism in a high - concurrency environment. The cached content includes the token - permission mapping relationship. The cache update mechanism uses the TTL (Time To Live) mechanism to refresh periodically to ensure the accuracy of the permission status. After the permission verification passes, the normal business request is responded to. All operation logs are synchronously written to the log server. The log content includes the user ID, permission level, operation time, and operation type. According to the aforementioned login status update and permission control logic, a repeated authentication mechanism framework is designed. The framework includes an authentication status monitoring module, an authentication status processing module, and an operation permission management module. The authentication status monitoring module is responsible for monitoring the changes in the secure repeated authentication data in real - time. It adopts an event - driven architecture, and the monitored events are asynchronously transmitted through a message queue. The message queue selects the RabbitMQ message broker to ensure the reliability and orderliness of message transmission. After receiving the event, the authentication status processing module performs the corresponding login status update operation and permission adjustment according to the status code. The module builds a built - in status mapping table and permission mapping table to support multi - status fast switching. The operation permission management module provides interfaces for permission application, change, and revocation externally. The interfaces are exposed in the form of RESTful APIs, and the interface response time is controlled within milliseconds. The repeated authentication mechanism framework is deployed in the form of a microservice architecture. The microservices are managed through Docker containers, and the container orchestration adopts the Kubernetes platform to ensure the elastic scalability and high availability of the system. This repeated authentication mechanism is integrated into the main process of the e - commerce platform through the API gateway. The API gateway is responsible for request forwarding and load balancing. All user login requests and subsequent operation requests need to be verified through this mechanism to ensure the continuity and real - time nature of the platform login security management.

[0112] Preferably, when the secure repeated authentication data is authentication security, the camera module and the fingerprint recognition chip are called for a hardware loopback verification, and the operation authorization based on the hardware loopback verification result includes:

[0113] When the secure repeated authentication data is authentication security, obtain the hardware loopback trigger data;

[0114] Based on the hardware loopback trigger data, call the camera module to collect the live facial feature data, and obtain the facial live feature data;

[0115] Based on the hardware loopback trigger data, call the fingerprint recognition chip to collect the fingerprint features, and obtain the fingerprint multi - dimensional feature data;

[0116] Perform feature fusion comparison on facial liveness feature data and fingerprint multi-dimensional feature data to generate a hardware loopback verification result;

[0117] Perform comprehensive operation authorization based on the hardware loopback verification result.

[0118] In this embodiment, when the secure repeated authentication data is in the authentication secure state, the security control unit will enable the hardware loopback call mechanism through the authorization channel in the local authentication queue. First, activate the interrupt trigger control module in the main control chip. By reading the verification result identification field in the secure repeated authentication data, using the 16-bit CRC control code as the hardware loopback signal identification entry, write the identification code into the physical layer loopback buffer. Set the clock control period in the buffer to 20MHz, trigger the independent startup channels of the camera module and the fingerprint recognition chip to form hardware loopback trigger data. The data content includes the signal source identification, trigger timestamp, trigger module code, authentication flag bit, and buffer queue number. This data is transmitted to the edge control board through the SPI bus and written in hexadecimal form to the starting address of the verification buffer unit. Based on the hardware loopback trigger data, when the control unit calls the camera module to collect facial liveness features, confirm that the image sensor model used is IMX219 according to the module code item in the hardware loopback trigger data. Before starting, perform I2C address configuration and exposure time initialization on the module. Set the exposure time to 33.3ms, limit the image frame rate to 30fps, perform synchronous calibration on the image brightness and contrast through a preset color channel correction matrix. Fix the image acquisition resolution at 1280×720. The image is read by the local image processing engine and undergoes liveness screening based on infrared reflectance comparison by the liveness detection module. The screening basis is the dynamic features of the eyes and the color temperature change gradient of the skin color. Images that meet the threshold conditions will be stored in the module high-speed cache area and then transferred to the image vector generation unit to generate facial liveness feature data. The feature data is generated using a multi-scale feature extraction algorithm and includes a total of 512-dimensional face key point embedding vectors and 16-dimensional hotspot area coding values. When calling the fingerprint recognition chip to collect fingerprint features based on the hardware loopback trigger data, set the chip model to FPC1020, initialize through the hardware SPI interface, set the collection area to 192×192 pixels, with a resolution of 508dpi. Set the liveness pressure sensor threshold at 85g / cm 2Within the range, after triggering the sensor, a complete fingerprint ridge image is collected. The local DSP module performs third-order derivative ridge enhancement filtering, and after removing artifacts, the ridge image envelope analyzer is called for two-way closed-loop ridge analysis. The distance between each pair of ridge points shall not exceed 4 pixels, generating fingerprint multi-dimensional feature data including orientation maps, minutiae maps, and ridge curvature mappings. The feature structure uses the FVC general template standard and is stored in JSON format, containing 128 minutiae records and their orientation angles, types, position coordinates, and adjacency list structures. When performing feature fusion and comparison on facial liveness feature data and fingerprint multi-dimensional feature data, the local feature cross-graph structure fusion algorithm is used. First, the two types of feature data are respectively converted into graph structures and then node-level structural nesting is performed. Node vector transmission and normalization are carried out through the Graph Convolution Module. The spacing for each round of convolution update is 3-layer depth. Fusion is performed through the node similarity function after convolution. The function output is a sequence of similarity vectors. The maximum vector value is compared with the set threshold of 0.92. If the similarity is higher than the threshold, it is determined as a match. Finally, the hardware loopback verification result is output, which includes four items: graph fusion flag bit, authentication match success flag, graph merging loss value, and fusion confidence score. When performing comprehensive operation authorization based on the hardware loopback verification result, this result is written into the security token register. The current session ID is encrypted by RSA 2048 bits using the encryption token generator to generate an operation authorization token with a hardware identity binding identifier. At the same time, the authorized operation set is written into the platform operation command buffer in the form of a hardware address mapping. Each authorized command contains a four-bit operation code, an eight-bit resource locator, and a sixteen-bit permission factor. This command is written into the operation scheduling unit through the bus, and the operation event is mapped into the platform main business logic chain. The authorized actions include permission modules such as fund transfer, information change, identity transfer, and external service interface access. Each operation needs to bind the hash digest of the current hardware loopback verification result as the only verification basis to ensure that the operation authorization takes effect only for the identity entity confirmed by this verification.

[0119] The present invention also provides an e-commerce platform security authentication system for performing the e-commerce platform security authentication method as described above. The e-commerce platform security authentication system includes:

[0120] An identity traceability module, used to obtain platform login identity information; perform field migration and recombination on the platform login identity information to obtain migrated field data; construct an identity information traceability chain based on the migrated field data;

[0121] A risk assessment module, used to perform multi-domain splitting on the identity information traceability chain to obtain a domain-segmented identity chain; perform identity risk assessment based on the domain-segmented identity chain to generate identity information risk parameters;

[0122] A fusion authentication module, which is used to perform parallel fusion authentication according to identity information risk parameters and domain-segmented identity chains, so as to generate an initial security authentication result;

[0123] A login disguise module, which is used to perform platform login disguise based on the initial security authentication result, and continuously collect data on changes in the login environment; verify the authenticity of the environment based on the data on changes in the login environment, and perform security determination based on the data on the authenticity of the environment to obtain login environment security data;

[0124] A re-authentication module, which is used to perform platform login re-authentication based on the login environment security data and the initial security authentication result to obtain secure repeated authentication data; update the platform login status based on the secure repeated authentication data, and construct a repeated authentication mechanism.

[0125] Through the application of the identity traceability module in the present invention, the system can accurately obtain and reorganize platform login identity information, form a complete identity information traceability chain, providing a solid foundation for subsequent authentication. The risk assessment module can refine risk identification and enhance the security of identity information by performing multi-domain splitting on the identity information traceability chain. The generated identity information risk parameters provide a scientific basis for authentication decisions. The introduction of the fusion authentication module ensures the effective combination of identity information risk parameters and domain-segmented identity chains, and can generate reliable initial security authentication results. The login disguise module effectively prevents identity information from being maliciously obtained through disguise strategies, and continuously collects data on changes in the login environment to ensure real-time monitoring of environmental security. Verifying authenticity based on environmental change data can promptly identify illegal login requests. The generated login environment security data provides strong support for security determination. The re-authentication module performs dynamic re-authentication based on the login environment security data and the initial security authentication result, ensuring timely update of the login status. The constructed repeated authentication mechanism enhances the security protection ability of the system, preventing identity theft and other illegal acts. The overall system forms a comprehensive security protection system through multi-level security strategies, greatly enhancing the protection of user information, enhancing users' trust in the platform, and promoting the healthy development of e-commerce. With the continuous progress of technology, the system can adapt to new security challenges, provide continuous security guarantees, and ultimately form an efficient and intelligent e-commerce security authentication system, creating a more secure trading environment for users, promoting the sustainable growth of the digital economy, laying a solid foundation for future development, enhancing the overall security of the platform and the user experience, and ensuring a high level of security protection in a complex network environment.

[0126] Therefore, in whatever aspect, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Thus, all changes falling within the meaning and scope of the equivalent elements of the application documents are intended to be embraced by the present invention.

[0127] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather will conform to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. A security authentication method for an e-commerce platform, characterized in that, It includes the following steps: Step S1: Obtain the platform login identity information; perform field migration and recombination on the platform login identity information to obtain the migrated field data; Build an identity information traceability chain based on the migrated field data; Step S2: Perform multi-domain splitting on the identity information traceability chain to obtain a domain-segmented identity chain; perform identity risk assessment based on the domain-segmented identity chain to generate identity information risk parameters; Step S3: Perform parallel fusion authentication according to the identity information risk parameters and the domain-segmented identity chain to generate an initial security authentication result; Step S4: Perform platform login disguise based on the initial security authentication result, and continuously collect login environment change data; Verify the environmental authenticity based on the login environment change data, and perform security determination based on the environmental authenticity verification data to obtain login environment security data; Step S5: Perform platform login re-authentication based on the login environment security data and the initial security authentication result to obtain secure repeated authentication data; update the platform login status based on the secure repeated authentication data and build a repeated authentication mechanism.

2. The e-commerce platform security authentication method according to claim 1, wherein Step S1 includes the following steps: Step S11: Obtain the platform login identity information; extract the field semantic features in the platform login identity information; Step S12: Perform field mapping and reconstruction on the field semantic features to obtain the migrated field data; Step S13: Perform segmented traceability nesting on the migrated field data to generate segmented traceability data, where the segmentation standard is that the login time interval is greater than 15 minutes as the boundary, and the nesting method is secondary nesting. The first level is the user identifier, and the second level is the time segmentation order. Each segment contains 1-10 groups of field change records; Step S14: Perform chain tracking on the segmented traceability data and build an identity information traceability chain, where the length of the traceability chain is set in the range of 3-128 nodes.

3. The e-commerce platform security authentication method according to claim 1, characterized in that, The multi-domain splitting of the identity information traceability chain in Step S2 includes: Perform multi-domain label annotation on the identity information traceability chain to obtain the attributes of each node in the traceability chain; Divide the attributes of each node in the traceability chain into multiple business domains to generate a multi-domain annotated traceability chain; Identify the connection relationships between nodes in the multi-domain annotated traceability chain; Discriminate the domain data to which the connection relationships between nodes belong, and extract the intra-domain interaction nodes based on the domain data to which the relationships belong; Identify the cross-domain information flow breakpoints in the connection relationships between nodes according to the intra-domain interaction nodes; Locate the boundary domain data of the identity information traceability chain through the information flow breakpoints; Distinguish multiple business domains based on the boundary domain data to obtain a domain-segmented identity chain.

4. The e-commerce platform security authentication method according to claim 1, wherein, The identity risk assessment based on the domain-segmented identity chain in Step S2 includes: Extract the chain structure features in the domain-segmented identity chain; Perform integrity verification on the domain-segmented identity chain according to the chain structure features to obtain integrity verification data; Identify the cross-domain interaction behavior data in the domain-segmented identity chain; Perform abnormal cross-domain behavior monitoring on the cross-domain interaction behavior data based on the preset cross-domain behavior baseline data to obtain cross-domain behavior monitoring data; Quantify the identity credibility of the domain-segmented identity chain through the integrity verification data and the cross-domain behavior monitoring data to generate identity information risk parameters.

5. The e-commerce platform security authentication method according to claim 1, wherein, In step S4, perform platform login disguise based on the initial security authentication result, and continuously collect data on changes in the login environment, including: Construct a disguised identity token based on the initial security authentication result; Perform temporary identity mapping on the disguised identity token to obtain temporary disguised identity data; Isolate the temporary disguised identity data in a session sandbox to generate sandbox session data; Construct a virtual login environment based on the sandbox session data; Continuously collect login IP data and login device parameters based on the virtual login environment to obtain a multi-point login IP dataset and a collection of login device parameters; Perform geographical location drift detection on the multi-point login IP dataset to obtain geographical location drift data; Evaluate the device change rate based on the collection of login device parameters; Integrate the geographical location drift data and the device change rate to generate data on changes in the login environment.

6. The e-commerce platform security authentication method according to claim 1, characterized in that In step S4, verify the authenticity of the environment based on the data on changes in the login environment, and perform security determination based on the data on the authenticity of the environment, including: Trace the historical trajectory of the login environment for the data on changes in the login environment to obtain environment trajectory traceability data; Locate multi-source node conflicts in the environment trajectory traceability data and extract suspicious node data; Identify the time-series mutation rate in the suspicious node data; Perform environment disguise judgment on the data on changes in the login environment based on the time-series mutation rate to obtain data on the authenticity of the environment verification; Perform dynamic trust grading based on the data on the authenticity of the environment verification to generate trust-graded environment data; Perform environment security determination on the data on changes in the login environment based on a preset trust threshold and the trust-graded environment data to obtain login environment security data.

7. The e-commerce platform security authentication method according to claim 1, characterized in that, In step S5, perform platform login re-authentication based on the login environment security data and the initial security authentication result, including: Perform fusion authentication based on the login environment security data and the initial security authentication result to obtain secure repeated authentication data. When the initial security authentication result is a failed security authentication and the login environment security data indicates an insecure login environment, define the secure repeated authentication data as authentication insecure; When the initial security authentication result is a successful security authentication and the login environment security data indicates an insecure login environment, define the secure repeated authentication data as authentication ambiguous; When the initial security authentication result is a successful security authentication and the login environment security data indicates a secure login environment, define the secure repeated authentication data as authentication secure.

8. The e-commerce platform security authentication method according to claim 1, characterized in that, In step S5, update the platform login status based on the secure repeated authentication data and construct a repeated authentication mechanism, including: Real-time update the platform login status through the secure repeated authentication data. When the secure repeated authentication data is authentication insecure, cancel the login status and return to the platform login interface; When the secure repeated authentication data is authentication ambiguous, restrict sensitive operations of the platform login identity, send a security alert to a preset bound contact, and at the same time collect feedback information from the bound contact and perform operation authorization based on the feedback information; When the secure repeated authentication data is authentication secure, call the camera module and fingerprint recognition chip for a hardware loopback verification, and perform operation authorization based on the result of the hardware loopback verification; Construct a repeated authentication mechanism according to the above steps and embed the repeated authentication mechanism into the e-commerce platform.

9. The e-commerce platform security authentication method according to claim 8, characterized in that, When the secure repeated authentication data is authenticated as secure, call the camera module and the fingerprint recognition chip to perform a hardware loopback verification, and perform operation authorization based on the hardware loopback verification result, including: When the secure repeated authentication data is authenticated as secure, obtain the hardware loopback trigger data; Call the camera module based on the hardware loopback trigger data to collect live facial feature data, and obtain the facial live feature data; Call the fingerprint recognition chip based on the hardware loopback trigger data to collect fingerprint features, and obtain the fingerprint multi-dimensional feature data; Perform feature fusion comparison on the facial live feature data and the fingerprint multi-dimensional feature data to generate the hardware loopback verification result; Perform comprehensive operation authorization based on the hardware loopback verification result.

10. An e-commerce platform security authentication system, characterized in that, For implementing the e-commerce platform security authentication method as described in claim 1, the e-commerce platform security authentication system includes: An identity traceability module, configured to obtain platform login identity information; perform field migration and recombination on the platform login identity information to obtain migrated field data; construct an identity information traceability chain based on the migrated field data; A risk assessment module, configured to perform multi-domain splitting on the identity information traceability chain to obtain a domain-segmented identity chain; perform identity risk assessment based on the domain-segmented identity chain to generate an identity information risk parameter; A fusion authentication module, configured to perform parallel fusion authentication according to the identity information risk parameter and the domain-segmented identity chain to generate an initial security authentication result; A login disguise module, configured to perform platform login disguise based on the initial security authentication result, and continuously collect login environment change data; verify the environmental authenticity based on the login environment change data, and perform security determination based on the environmental authenticity verification data to obtain login environment security data; A re-authentication module, configured to perform platform login re-authentication based on the login environment security data and the initial security authentication result to obtain secure repeated authentication data; update the platform login status based on the secure repeated authentication data, and construct a repeated authentication mechanism.

Images