An e-commerce platform information security protection method and system

Abstract

The application provides an e-commerce platform information security protection method and system, relates to the technical field of information security and automatic review, and has the technical scheme as follows: a plurality of candidate expert objects and a plurality of candidate suppliers are acquired, a plurality of source event information is obtained through multi-database retrieval based on the plurality of candidate expert objects; each candidate expert object is combined with each candidate supplier to obtain a plurality of object groups, and the associated event information of each object group is determined; the association depth of the associated event information and the corresponding object group is identified, the event type of the associated event information is determined, the association strength is determined according to the association depth and the event type of the associated event information; the interest conflict level is determined according to the association strength; and the control instruction is generated according to the interest conflict level. The application aims to realize deep identification of the implicit interest association between experts and suppliers, reduce the risk of potential conflict experts obtaining sensitive data, and guarantee the fairness of review.

Description

Technical Field

[0001] This application relates to the field of information security and automated review technology, and more specifically, to a method and system for information security protection of e-commerce platforms. Background Technology

[0002] In the bidding and procurement process of e-commerce platforms, the formation of review committees typically involves automated conflict-of-interest filtering of candidate experts. This is achieved through methods such as expert-submitted information and explicit decision-making rules to identify conflicts of interest, granting review committees access to sensitive data of bidding suppliers to experts who pass the comparison process. However, complex implicit conflicts of interest may exist between experts and suppliers, such as joint participation in technology projects, joint ownership of invention patents, or holding positions in the same industry association. The filtering logic primarily targets explicit direct relationships, failing to identify deeper, implicit, weak connections. This leads to experts with potential conflicts of interest being deemed qualified and granted access to sensitive data, resulting in a loss of fairness in the review process and undermining the trust between the review body and the platform. Summary of the Invention

[0003] The purpose of this application is to provide a method and system for information security protection of e-commerce platforms, which aims to achieve in-depth identification of implicit interest relationships between experts and suppliers, reduce the risk of potential conflicting experts obtaining sensitive data, and ensure the fairness of the review process.

[0004] Firstly, this application provides a method for information security protection of an e-commerce platform, including: Multiple candidate expert objects and multiple candidate suppliers are obtained, and multi-database retrieval is performed based on the multiple candidate expert objects to obtain multi-source event information for each candidate expert object; Each candidate expert is combined with each candidate supplier to obtain multiple object groups. Multi-source event information that has a direct or indirect relationship with both the candidate expert and the candidate supplier in the object group is identified, and the associated event information of each object group is obtained. Identify the association depth between the associated event information and the corresponding object group, determine the event type of the associated event information, and determine the association strength between the associated event information and the corresponding object group based on the association depth and event type of the associated event information; Based on the correlation strength and the preset judgment threshold, the conflict of interest level between candidate expert objects and candidate suppliers in the object group corresponding to the correlation strength is determined; Based on the conflict of interest level, control instructions are generated to determine the access permissions of candidate expert objects in the corresponding object group to the bidding project data of candidate suppliers in the same object group.

[0005] Secondly, this application provides an information security protection system for an e-commerce platform, comprising: The information acquisition module is used to acquire multiple candidate expert objects and multiple candidate suppliers, and perform multi-database retrieval based on the multiple candidate expert objects to obtain multi-source event information of each candidate expert object. The associated event module is used to combine each candidate expert object with each candidate supplier to obtain multiple object groups, determine multi-source event information that has a direct or indirect relationship with both the candidate expert objects and candidate suppliers in the object group, and obtain the associated event information of each object group. The association strength module is used to identify the association depth between the associated event information and the corresponding object group, and to determine the event type of the associated event information. Based on the association depth and event type of the associated event information, the association strength between the associated event information and the corresponding object group is determined. The conflict level module is used to determine the conflict level between candidate expert objects and candidate suppliers in the object group corresponding to the association strength based on the association strength and the preset judgment threshold. The instruction control module is used to generate control instructions based on the conflict of interest level to determine the access permissions of candidate expert objects in the corresponding object group to the bidding project data of candidate suppliers in the same object group.

[0006] As can be seen from the above, the e-commerce platform information security protection method and system provided in this application can identify direct conflict of interest relationships by constructing a relationship network graph including multi-source event information. Furthermore, it can mine implicit historical interactions and weak correlation information between experts and suppliers through graph traversal algorithms. It calculates the correlation strength through multi-dimensional trade-off factors such as time decay, relationship depth, and event type, and optimizes homogeneous relationship clusters through cluster analysis to improve the accuracy of risk assessment. Moreover, by combining an anomaly feedback mechanism with refined hierarchical access control, it can effectively prevent unauthorized access to bidding sensitive data by potentially conflicting experts while protecting the legitimate rights and interests of review experts, and avoid review security defects caused by the omission of implicit interest chains. Attached Figure Description

[0007] Figure 1 This is a flowchart of an information security protection method for an e-commerce platform provided in the first specific embodiment of this application; Figure 2 This is a flowchart of an information security protection method for an e-commerce platform provided in the second specific embodiment of this application; Figure 3 This is a flowchart of an information security protection method for an e-commerce platform provided in the third specific embodiment of this application. Detailed Implementation

[0008] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0009] It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0010] Firstly, such as Figure 1 As shown, Figure 1 This is a flowchart of an information security protection method for an e-commerce platform provided in the first specific embodiment of this application. The method may include, but is not limited to, steps S110 to S150.

[0011] Step S110: Obtain multiple candidate expert objects and multiple candidate suppliers, and perform multi-database retrieval based on multiple candidate expert objects to obtain multi-source event information for each candidate expert object; Step S120: Combine each candidate expert object with each candidate supplier to obtain multiple object groups, determine multi-source event information that has a direct or indirect relationship with both the candidate expert objects and candidate suppliers in the object groups, and obtain the associated event information of each object group. Step S130: Identify the association depth between the associated event information and the corresponding object group, and determine the event type of the associated event information. Based on the association depth and event type of the associated event information, determine the association strength between the associated event information and the corresponding object group. Step S140: Based on the association strength and the preset judgment threshold, determine the conflict of interest level between candidate expert objects and candidate suppliers in the object group corresponding to the association strength. Step S150: Based on the conflict of interest level, generate control instructions to determine the access permissions of candidate expert objects in the corresponding object group to the bidding project data of candidate suppliers in the same object group.

[0012] For example, candidate experts refer to individual experts who are qualified to review projects in e-commerce platform bidding projects and have been initially screened into the candidate list of the review committee. Each candidate expert corresponds to at least one set of identity information that can be retrieved, including one or more combinations of name, former name, current employer, previous employer, unified social credit code associated unit information, and certificate identification information. Candidate suppliers refer to enterprises or organizations that submit bid documents and related qualification materials to participate in the current bidding project. In addition to the enterprise name, candidate suppliers can also be associated with unified social credit code, historical name, parent company name, branch name, legal representative, principal responsible person, controlling shareholder, and other identification information. Multi-source event information refers to data records of various historical activities, cooperative projects, or social relationships participated in by record objects extracted from databases of different dimensions. Multi-source event information can be a structured event unit, including fields such as event name, event type, participating entities, participating roles, occurrence time or time interval, source database identifier, original record link or original record number, etc.

[0013] For example, multi-source event information can be obtained by connecting to multiple external databases through pre-configured API interfaces, such as the State Intellectual Property Office Patent Database, the Ministry of Science and Technology National Science and Technology Management Information System, and the Ministry of Civil Affairs Social Organization Query Platform. Using the identity identifiers of candidate experts as search keywords, event records such as participation in patent inventions, scientific research project reviews, and industry association appointments can be retrieved to form multi-source event information. For instance, an external data source configuration table can be established first, recording the interface address, authentication method, queryable fields, returned fields, and update cycle of each external database. After receiving the list of candidate experts, the system first processes the candidate experts... The identity identifiers of the objects are standardized, for example, by unifying the name format, separating the current unit from the historical unit, and establishing alias mapping. Then, query requests are generated according to the interface requirements of different databases. After the interface returns the results, field extraction, field renaming, time format unification, unit name standardization, and duplicate record merging are performed. For example, the same patent may be found in different sources at the same time, and deduplication can be performed based on the patent application number or publication announcement number. The same project may appear on the project publicity page and the project acceptance page respectively, and can be merged based on the project number. After merging, each record is organized into standardized multi-source event information.

[0014] For example, to identify multi-source event information that has a direct or indirect relationship with both candidate experts and candidate suppliers in the object group, text matching can be used to search for events related to suppliers within the experts' multi-source event information. For instance, searching for supplier name keywords in the expert's event record text; if a match is found, the event is considered an associated event. Alternatively, the name features of candidate suppliers can be expanded before matching. Name feature expansion can include the full name of the company, historical name, commonly used abbreviations, names of subordinate research institutes, names of legal representatives, names of key personnel, etc. Subsequently, the multi-source event information of experts is scanned one by one. If any name feature of the supplier appears in the fields of participating entity, cooperating unit, research unit, applicant, or employing institution in a certain event information, it is preliminarily determined that the event has a direct relationship with the supplier. If the supplier's name does not appear directly, but an entity with a stable mapping relationship with the supplier appears, such as the supplier's wholly-owned subsidiary, holding parent company, legal representative, or entity corresponding to the unified social credit code, the event is marked as a candidate indirect relationship event.

[0015] For example, association depth is used to determine the closeness of the connection between experts and suppliers established through events, event type characterizes the nature of the event, and association strength can be obtained through a pre-built association strength assessment model. For example, a rule-based scoring matrix can be used. When associated event information is identified, the association depth of the event is determined, and the corresponding score is found in the scoring matrix in combination with the event type to obtain the association strength. Specifically, an event type classification table and a depth level classification table can be established first. The event type classification table is used to classify multi-source event information into several stable categories, such as joint patents, joint undertaking or participation in projects, project review and research relationships, joint appointments in industry associations, commercial transactions, equity control, and cross-appointments in historical positions. The depth level classification table is used to describe the hierarchy of event connections. For example, direct joint participation in the same event is classified as a higher association level, connection through one intermediary entity is classified as an intermediate association level, and connection through multiple intermediary entities is classified as a lower association level. Subsequently, in combination with the risk control requirements of bidding business, a correspondence table between event type, depth level, and association strength range can be pre-established. If there are multiple related event information for the same object group, the association strength can be determined separately, and then the overall association strength of the object group can be generated by taking the highest risk event as the priority or by superimposing multiple events.

[0016] For example, the preset judgment threshold can be set as a graded threshold. For instance, the conflict of interest level can be divided into high-risk conflict, medium-risk pending review, low-risk warning, and no obvious conflict. Correspondingly, a mapping relationship between the association strength level and the conflict of interest level can be established in advance. For example, if the object group has events such as direct joint patents, direct equity control, or direct overlapping of positions, it is mapped as a high-risk conflict. If the object group only has a historical cooperative relationship formed through an intermediate project, and the event time is early and the number of repetitions is few, it is mapped as medium-risk pending review or low-risk warning. If there are only weak connections such as appearing at the same public meeting or being ordinary members of the same association, it is mapped as low-risk warning or no obvious conflict.

[0017] For example, control instructions may include candidate expert object identifiers, candidate supplier identifiers, bidding project identifiers, permission action type, permission effective status, triggering reason explanation, and generation time. The permission action type may include prohibiting viewing, allowing viewing, allowing viewing of anonymized data, and submitting for manual review before decision. For instance, when the conflict of interest level is high-risk, a prohibition control instruction is generated and written to the review permission control table, allowing the review front-end to directly intercept the expert when opening the corresponding supplier's bid document. When the conflict of interest level is medium-risk and pending review, a temporary authorization control instruction is generated, preventing the complete data of the corresponding supplier from being disclosed to the candidate expert object before manual review is completed. When the conflict of interest level is low-risk (indicating a warning), a control instruction is generated allowing viewing but recording in the audit log, while simultaneously retaining the risk marker for the object group in the background.

[0018] If sensitive data is protected in a tiered manner, the control methods can be further refined. For example, high-risk conflict groups can be completely prohibited from accessing tender documents, quotations and qualification attachments, while medium-risk groups awaiting review can only view basic information that has been anonymized, and are not allowed to view price composition and core technical solutions.

[0019] Understandably, this application obtains multi-source event information of candidate experts through multi-database retrieval, avoiding the limitations of relying solely on expert self-reporting and single explicit rules. It broadens the dimensions of information acquisition, and by combining experts with suppliers and identifying related event information with direct or indirect relationships, it can effectively uncover complex interest chains hidden in historical interactions. By combining the depth of association and event type to comprehensively quantify the strength of association and classify the level of conflict of interest, the determination of conflict of interest becomes more objective and accurate. Finally, based on the level of conflict of interest, access control instructions are generated, which can block experts with potential conflicts of interest from obtaining sensitive data from specific suppliers, avoid the risk of data leakage caused by the failure to identify hidden interest relationships, ensure the fairness and impartiality of the bidding project review process, and improve the data security protection level of e-commerce platforms. Furthermore, multi-database retrieval enables the detection of historical collaborations, joint patents, association appointments, and other events that were not originally included in the judgment criteria set; object grouping processing ensures that access control is no longer targeted at the expert as a whole, but rather at the relationship between the expert and specific suppliers on an individual basis; joint assessment of association depth and event type enables the differentiation between direct interest ties and weak indirect contacts; mapping of conflict of interest levels to control instructions allows risk identification results to be directly transformed into access control actions; and the seamless connection of each link, from data collection, relationship identification, risk assessment to access control execution, forms a closed-loop control, thereby effectively suppressing the risk of data leakage caused by implicit conflicts of interest.

[0020] like Figure 2 As shown, Figure 2 This is a flowchart of an information security protection method for an e-commerce platform provided in the second specific embodiment of this application. Regarding the above step S120, the method may include, but is not limited to, steps S210 to S220.

[0021] Step S210: Take multiple candidate expert objects and multiple candidate suppliers as nodes, and connect the corresponding candidate expert objects and candidate suppliers through multi-source event information to generate a relationship network diagram; Step S220: In the relationship network diagram, determine the multi-source event information that directly connects the candidate expert objects and candidate suppliers in the object group or indirectly connects them through intermediate events, and obtain the associated event information of each object group. Intermediate events include events in which the candidate expert objects and candidate suppliers participate.

[0022] For example, when constructing a relationship network graph, the Neo4j graph database method can be used. For instance, experts, suppliers, and other entities can be created as nodes. If multi-source event information shows that an expert and a supplier jointly applied for a patent, an edge is created between the expert node and the supplier node, with the edge's attribute recording "joint patent." Additionally, event nodes or intermediate entity nodes can be established. Event nodes can represent patents, projects, associations, transaction contracts, employment records, etc., while intermediate entity nodes can represent project research units, affiliated companies, legal representatives, controlling shareholders, associations, etc. Thus, the connections in the graph can be represented by various structures such as expert, event supplier, event expert, intermediate entity supplier, and intermediate entity. For example, when multi-source event information shows that an expert participates in a project as a review expert, and a supplier participates in the project as one of the project research units, two connections are established in the graph: expert to project and supplier to project. When multi-source event information shows that the supplier's legal representative also serves as a director of an association, and an expert also serves as a director of an association, connections are established between the expert and the association, the supplier's legal representative and the association, and the supplier and the legal representative.

[0023] For example, after generating the relationship network graph, a path search is performed on the graph for a specific group of objects. If there is an edge in the graph that directly connects the expert and the supplier, the event corresponding to that edge is the multi-source event information of the direct connection. If the expert and the supplier are not directly connected, but the graph shows that the expert participated in the science and technology project and the supplier is the research and development unit of the science and technology project, the expert and the supplier form an indirect connection through the intermediate event science and technology project. This intermediate event is also extracted as the associated event information. For example, starting with the candidate expert object node in the object group as the starting point and the candidate supplier node as the ending point, a reachable path is searched in the relationship network graph. If a direct path is found, the edge or event node on the direct path is extracted as the associated event information. If no direct path is found, the search continues to search for indirect paths that include one or more intermediate nodes, and the key event nodes, intermediate subject nodes and connection relationships on the path are recorded.

[0024] Based on this, this application transforms discrete multi-source event information into a structured relational network graph, making the complex relationships hidden in the text data visible and topological. Path finding in the relational network graph can traverse all possible connection paths between nodes to determine direct relationships and identify indirect relationships formed through intermediate events. This avoids the limitations of text matching in handling multi-level weak associations, ensures the comprehensive identification of hidden interest chains, and improves the rigor of interest conflict identification.

[0025] In the e-commerce platform information security protection method provided in some specific embodiments of this application, the association depth includes first-order association depth and multi-order association depth. Regarding the above step S130, the method may include, but is not limited to, step S310.

[0026] Step S310: Identify the associated event information of candidate expert objects and candidate suppliers in the object group that are directly connected by a preset graph traversal algorithm, so as to determine the first-order association depth between the corresponding associated event information and the corresponding object group. Identify the connection depth level of the associated event information that is indirectly connected to candidate expert objects and candidate suppliers in the object group by the preset graph traversal algorithm, so as to determine the multi-order association depth between the corresponding associated event information and the corresponding object group according to the connection depth level.

[0027] For example, the preset graph traversal algorithm can be a breadth-first search algorithm or a depth-first search algorithm. If the goal is to identify the shortest associated path first, a breadth-first search algorithm is used. If the goal is to enumerate all reachable paths of a group of objects as much as possible, a depth-first search algorithm combined with path deduplication rules is used. Simultaneously, the types of nodes allowed to participate in the traversal can be preset, such as allowing access to project nodes, patent nodes, association nodes, enterprise nodes, and individual nodes. However, noisy nodes unrelated to conflict of interest identification are excluded from the traversal to avoid generating a large number of meaningless paths. Path truncation rules can also be preset, for example, stopping further expansion when a path has already passed through multiple intermediate levels and its association significance weakens.

[0028] In some specific embodiments, a breadth-first search algorithm is used to traverse layer by layer starting from the candidate expert object nodes in the object group. When the algorithm reaches the candidate supplier node in the first layer traversal, it indicates that there is a direct edge connection. For example, if the two parties have directly signed a technical consulting contract or jointly participated in the same patent event as the applicant, inventor, and applicant, the related event information is identified as a direct connection and determined as a first-order association depth. If the first layer traversal does not reach the candidate supplier node, the algorithm continues to expand outward. When the algorithm reaches the candidate supplier node in the Nth layer traversal (greater than or equal to 2), it indicates that an indirect connection is formed through an intermediate node. At this time, the connection depth level corresponding to this level is used as the level identifier of the indirect connection.

[0029] If there are multiple paths of different lengths for the same object group, the connection depth level corresponding to the shortest effective path can be used as the primary association depth. At the same time, other longer paths can be retained as supplementary evidence for subsequent manual review or risk interpretation.

[0030] Based on this, this application scans the relationship network graph using a pre-defined graph traversal algorithm, which can accurately calculate the shortest path level between nodes. Direct connections are defined as first-order association depths, and indirect connections are quantified into multi-order association depths according to specific levels. This enables accurate identification of the layers of complex interest-related networks, eliminates subjective biases from human judgment, and improves the efficiency of processing large amounts of data. Furthermore, by refining implicit relationships into depth indicators at different levels, it provides fine-grained dimensional parameters for subsequent calculation of association strength, enabling accurate differentiation between strong direct interests and weak indirect interests, thereby ensuring that the final generated access control instructions can prevent risks.

[0031] In some specific embodiments of the e-commerce platform information security protection method provided in this application, regarding the above step S130, the method may include, but is not limited to, steps S410 to S420.

[0032] Step S410: Determine the current time constant by the time distance between the occurrence time of the associated event information and the current time; determine the current association depth weight according to the association depth of the associated event information; extract the corresponding preset relationship type weight according to the event type of the associated event information; the current association depth weight corresponding to the first-order association depth is greater than the current association depth weight corresponding to the multi-order association depth. Step S420: Based on the product of the current time constant, the current association depth weight, and the preset relationship type weight, obtain the association strength between the associated event information and the corresponding object group.

[0033] For example, the current time constant is used to characterize the degree of decay of an event's influence over time. The occurrence time can be directly read from the associated event information record. If the original data only records the year or month, it is supplemented according to a uniform time granularity. For example, if only the year is available, the year is mapped to the median time of that year or marked as a low-precision time, and a decay level adapted to the low-precision time is used in subsequent calculations. After obtaining the time distance, the current time constant can be obtained by substituting the time distance between the occurrence time recorded in the associated event information and the current system time into a preset time decay function. Alternatively, it can be mapped to the current time constant according to a preset time decay rule. The time decay rule can be established through historical audit samples. For example, several historical object groups that have been confirmed to constitute a conflict of interest are collected, and the distance of the events from the current time is compared with the final judgment result to statistically determine the degree of retention of the event's influence in different time intervals, and then a time constant configuration table is formed.

[0034] For example, the current association depth weight can also be mapped to different weight levels by a pre-established weight mapping table, where first-order association depth and multi-order association depth are respectively mapped to different weight levels. First-order association depth represents a direct connection, so the highest level is determined in the mapping table, and the association depths of second-order, third-order and deeper levels decrease sequentially. The preset relationship type weight represents the inherent risk attributes of different event types. For example, event types such as jointly applying for invention patents, jointly undertaking high-value projects, having equity control relationships, and having commercial transaction contract relationships are configured with higher preset relationship type weights because they are more likely to form real-world interest binding. Event types such as only having records of being present at public forums or records of being present at ordinary academic conferences are configured with lower preset relationship type weights.

[0035] Based on this, this application decomposes the complex historical interactions that are originally difficult to quantify into three independent and mutually restrictive mathematical dimensions: time, depth, and type. By using multiplication operations to achieve cross-modulation of multiple dimensions, it can ensure that extremely low values ​​in any dimension will reduce the overall correlation strength, thereby truly representing the logic of interest relationships in reality. This effectively avoids the static evaluation ignoring the time decay effect, allowing for the forgetting of weak correlations that are no longer active, while identifying recent strong correlations. This improves the scientificity and accuracy of correlation strength quantification and avoids over-defense caused by historical events.

[0036] In the e-commerce platform information security protection method provided in some specific embodiments of this application, the conflict of interest level includes high-risk conflict level, medium-risk conflict level and low-risk conflict level; regarding the above step S420, the method may include but is not limited to steps S510 to S520; regarding the above step S140, the method may include but is not limited to step S530.

[0037] Step S510: Based on the product of the current time constant, the current association depth weight, and the preset relationship type weight, obtain the individual association strength between each association event information and the corresponding object group; Step S520: Combine the associated event information of the object group into event units, and obtain the comprehensive association strength between the object group and the corresponding event units based on the individual association strength of each associated event information of the object group. Step S530: If the overall correlation strength of the event unit is greater than the preset high-risk threshold, determine the high-risk conflict level between candidate expert objects and candidate suppliers in the object group corresponding to the event unit. If the overall correlation strength of an event unit is less than a preset high-risk threshold but greater than a preset low-risk threshold, determine the medium-risk conflict level between candidate expert objects and candidate suppliers in the object group corresponding to the event unit. If the overall correlation strength of an event unit is less than the preset low-risk threshold, the low-risk conflict level between candidate expert objects and candidate suppliers in the object group corresponding to the event unit is determined.

[0038] For example, all related event information belonging to the same object group is packaged into a complete event unit. By performing aggregation calculation on all individual association strengths within the event unit, a comprehensive association strength that can represent the overall interest binding degree of the object group is calculated. After obtaining the comprehensive association strength, the continuous numerical range is divided into three discrete risk levels through a preset high-risk threshold and a preset low-risk threshold judgment mechanism.

[0039] The overall correlation strength can be obtained using preset aggregation rules. In the initial stage, a direct accumulation method is used to ensure simplicity. Once enough samples have been accumulated during business operations, the method is switched to weighted summation or nonlinear fusion to improve adaptability to complex scenarios.

[0040] During the judgment process, the overall correlation strength of the event unit is compared with the preset high-risk threshold and the preset low-risk threshold to output the high-risk conflict level, medium-risk conflict level or low-risk conflict level, and the judgment result is written into the object group risk result table, while recording the judgment time, rule version number and trigger basis field.

[0041] In addition, an event unit can be a master record in a structured data table, with multiple event details attached to the master record, or it can be a collection of objects in memory, dynamically generated in the risk analysis task. The event unit can be a traceable structure, with the master record storing the object group identifier, number of events, aggregation method, comprehensive correlation strength, and current risk level, and the detail records storing the original fields and individual correlation strength of each event.

[0042] Based on this, this application achieves a shift from single-point event assessment to global relationship assessment by constructing event units. By dividing the risk into dual threshold intervals, it achieves gradient classification of risks, which can effectively avoid the omission of multiple weak correlations that are overlooked. It can identify hidden interests in single compliant but frequently interacting situations. At the same time, the division of high, medium and low conflict levels enables precise control measures to be taken for risks of different degrees, improving business adaptability and control flexibility.

[0043] like Figure 3 As shown, Figure 3 This is a flowchart of an information security protection method for an e-commerce platform provided in the third specific embodiment of this application. Regarding the above step S520, the method may include, but is not limited to, steps S610 to S640.

[0044] Step S610: Extract the occurrence scenario level and object role information of each associated event, and combine the event type, occurrence scenario level and object role information of each associated event to obtain the event characteristics of each associated event; Step S620: Combine the associated event information of the object group into event units. By comparing the clustering of event features of each associated event information within the event unit, combine the associated event information within the event unit with clustering similarity greater than the preset feature matching threshold into homogeneous relationship clusters. Mark the associated event information within the event unit with clustering similarity less than the preset feature matching threshold as independent event information. Step S630: Remove association event information that has an individual association strength less than a preset association truncation threshold and belongs to the same homogeneous relationship cluster. Extract the maximum value of the individual association strength of each association event information in the removed homogeneous relationship cluster as the calculation benchmark value. Input the calculation benchmark value and the number of association event information in the removed homogeneous relationship cluster into a preset logarithmic saturation function to obtain the cluster association strength of the removed homogeneous relationship cluster. Adjust the value of the cluster association strength that is greater than the preset low-risk threshold to the value of the preset low-risk threshold. Step S640: Based on the cluster association strength of each removed homogeneous relation cluster and the individual association strength of each independent event information, obtain the comprehensive association strength between the object group and the corresponding event unit.

[0045] It is understandable that in the process of combining multiple related events into event units and calculating the overall correlation strength, there is an objective situation of redundant superposition of homogeneous events. Homogeneous events refer to events that originate from the same core cooperation matter in essence, but are split into multiple events only due to different business stages, record sources, and registration standards. For example, in government and enterprise database records, cooperation projects with long durations will generate multiple records at different stages. Large-scale scientific research projects jointly participated in by experts and suppliers may be captured as multiple independent related event information due to multiple stages such as project application, mid-term inspection, fund disbursement, and final acceptance. If the individual correlation strengths of homogeneous events are linearly accumulated, the overall correlation strength of the object group will appear abnormally high, thus misjudging what is originally a normal single business transaction of medium to low risk as a high-risk conflict.

[0046] For example, the occurrence scenario level is used to characterize the business level or scenario weight of the event, such as national-level projects, provincial-level projects, municipal-level projects, internal corporate meetings, industry forums, procurement contracts, consulting services, equity investments, etc. The business substance and scope of impact differ in different scenarios; the object role information is used to characterize the position of candidate experts and candidate suppliers in the event, such as chief expert, project leader, review group leader, ordinary group member, consultant, contract party A, contract party B, controlling shareholder, equity participation entity, etc., which can be extracted from the original event record.

[0047] Event features are formed by combining event type, occurrence level, and participant roles. These features can be represented as structured tag groups. For example, an event feature could be described as "jointly undertaken project," "provincial key R&D program," or "review team leader / undertaking unit." Another record could also be described as "jointly undertaken project," "provincial key R&D program," or "review team leader / undertaking unit."

[0048] For example, the preset feature matching threshold can be obtained by labeling samples; the preset association truncation threshold is used to remove weakly contributing derivative edge events in homogeneous relationship clusters, such as automatically generated project reminders, check-in records, process notifications, etc., which, even if related to core business, are not at the same level as records with real cooperation substance. The preset association truncation threshold can be obtained through historical sample statistics; the preset logarithmic saturation function is used to make the cluster association strength increase as the number of effective events in the homogeneous relationship cluster increases, but the rate of increase gradually slows down. The cluster association strength can be obtained by using a table of effective event numbers and gain levels, combined with the calculation benchmark value. For example, when there is only 1 effective event in the homogeneous relationship cluster, the cluster association strength is close to the calculation benchmark value. When it grows to multiple events, the cluster association strength increases moderately. When the number continues to increase, the increase gradually decreases until it approaches saturation.

[0049] For example, clustering algorithms, such as K-Means or hierarchical clustering, can be executed to calculate the cluster similarity between the features of each event. When the similarity exceeds a preset feature matching threshold, they are grouped into the same homogeneous relationship cluster.

[0050] In addition, if the number of events is small, hierarchical clustering can be used for pairwise comparison, which is convenient for intuitive interpretation. If the number of events is large, K-Means or other clustering algorithms can be used to improve processing efficiency. Furthermore, clustering descriptions can be recorded for each clustering result. For example, events 1, 2, and 3 are highly consistent in terms of event type, occurrence scene level, and object role information, so they are merged into a homogeneous relation cluster.

[0051] For example, after obtaining the event characteristics of each related event, the event characteristics can be standardized first, including unifying event type encoding, unifying the enumeration of occurrence scenario levels, unifying object role names, cleaning up invalid spaces and aliases, and filling in missing fields that can be inferred from the context. For example, if one project record includes a provincial key R&D plan and another includes a certain provincial key R&D plan, they can both be mapped to the same occurrence scenario level through the scenario dictionary; the team leader, review team leader, and review expert group leader can all be mapped to a unified role label.

[0052] For example, after generating homogeneous relation clusters, noise reduction and dimensionality reduction processing can be performed. First, weak edge records within the cluster are filtered out by a preset association truncation threshold. Then, the maximum value of the individual association strength among the remaining events within the cluster is extracted as the calculation benchmark value representing the basic risk of the core business. The calculation benchmark value and the number of effective events within the cluster are input into a preset logarithmic saturation function. As the independent variable increases, the growth rate of the dependent variable in the logarithmic function gradually slows down and tends to level off. That is, as the number of events increases, the growth rate of the cluster association strength gradually slows down and tends to level off, thus reasonably representing the increase in marginal risk brought about by multiple homogeneous records, while preventing the value from expanding indefinitely. Finally, the upper limit of the cluster association strength is forcibly truncated at a preset low-risk threshold to ensure that no matter how many records a single core business generates, its own risk level will not cross the medium-high risk range. Then, the processed cluster association strength is aggregated with the individual association strength of independent event information to obtain the final comprehensive association strength.

[0053] In some embodiments, the event unit for experts and suppliers includes 5 related event information. Four of these events occur in a provincial key R&D program, involving the review team leader and the research unit, respectively. Their cluster similarity is 0.95, exceeding the preset feature matching threshold of 0.85. Therefore, these four events are grouped into a homogeneous cluster 1. The remaining event is a joint application for a utility model patent, with low similarity, and is marked as an independent event. Within homogeneous cluster 1, one meeting attendance record with extremely low intensity is removed. The individual association strengths of the remaining three events are 0.3, 0.35, and 0.32, respectively. The maximum value of 0.35 is extracted as the calculation baseline. The baseline value of 0.35 and the quantity 3 are input into a preset logarithmic saturation function. Since three valid records are retained within the cluster, it indicates that the project is not just a single, occasional occurrence, but rather has stable records across multiple stages. Therefore, a moderate gain is applied to the baseline of 0.35, resulting in a preliminary cluster association strength of 0.45, which exceeds the preset low-risk threshold of 0.4. Therefore, it is forcibly adjusted to 0.4. Finally, the cluster association strength of 0.4 is combined with the individual association strength of the independent event information to obtain the final comprehensive association strength.

[0054] Based on this, this application identifies the homogeneity of data through multidimensional feature clustering, simulates the marginal diminishing effect of risk through a logarithmic saturation function, and prevents the risk of a single business from exceeding the limit through threshold truncation. This can avoid the falsely high misjudgment caused by the redundant superposition of homogeneous events, and make it possible to accurately distinguish the difference in the nature of risk between multiple records of the same project and one record of multiple different projects. This improves the scientificity and anti-interference ability of the comprehensive correlation strength calculation and ensures the fairness of the high-risk conflict judgment.

[0055] In some specific embodiments of the e-commerce platform information security protection method provided in this application, before step S140 above, the method may include, but is not limited to, steps S710 to S730.

[0056] Step S710: If the difference between the comprehensive correlation strength of the event unit and the preset high-risk threshold or the difference between the comprehensive correlation strength of the event unit and the preset low-risk threshold is within the preset critical range, perform event information missing analysis and timestamp logic conflict analysis on the related event information in the corresponding event unit to obtain the analysis results. Step S720: If the analysis results indicate that there are missing event information or timestamp logic conflicts in the related event information of each event unit, mark the corresponding event unit as a feedback event unit and stop obtaining the conflict of interest level of the feedback event unit. Step S730: After obtaining the feedback event information of the corresponding event unit to be fed back, determine the feedback object group corresponding to the feedback event information, determine the feedback correlation strength between the feedback event information and the corresponding feedback object group, and determine the conflict of interest level between the candidate expert objects and candidate suppliers in the feedback object group corresponding to the feedback correlation strength based on the feedback correlation strength and the preset judgment threshold.

[0057] For example, after completing the comprehensive association strength calculation, the comprehensive association strength of the current event unit is read and compared with the preset high-risk threshold and the preset low-risk threshold respectively. If the distance to either threshold falls within the preset critical range, a data quality review is triggered. Among them, the event information missing analysis is used to check whether the associated event has the core fields required to support the judgment, such as event type, occurrence time, participating entity name, entity identifier, object role information, scene information, etc.; the timestamp logic conflict analysis is used to determine whether the event occurrence time, the existence time of the relevant entity, and the order of the event chain conform to the basic factual logic. For example, situations such as a supplier not being established but participating in the cooperation, an expert not holding a position but signing project documents in that capacity, and the project completion time being earlier than the project initiation time are all timestamp logic conflicts that need to be intercepted.

[0058] In the event information missing analysis, each related event information within an event unit can be checked to see if any key fields are missing. This can be done by using a mapping table between event type and a list of required fields. For example, for a joint application for an invention patent, required fields include the patent application date, applicant or inventor information, and patent identifier. For a technical consulting contract, required fields include the contract signing date, contract parties, and contract type. For joint employment, required fields include the employing organization, employment start date, and employment role. If an event is missing a corresponding required field, the missing item type and original record number are recorded in the analysis results.

[0059] In timestamp logic conflict analysis, time logic verification rules can be used. For example, the event occurrence time should not be earlier than the supplier's business establishment time, not earlier than the actual existence time of the patent application subject, and not later than an unreasonable time interval after the subject's cancellation. The mid-term inspection time of the project should be later than the project initiation time, the final acceptance time should be later than the mid-term inspection time, and the expert's appointment start time should be earlier than the time when they participated in the project review in that role. After verifying each item, the result of the abnormal rule hit is written into the analysis result.

[0060] For example, when missing event information or timestamp logic conflict is found, the event unit is marked as an event unit to be fed back and written into the list of teams to be fed back. At the same time, the acquisition of the conflict of interest level of the event unit in the current round is stopped, that is, the automatic judgment process is terminated and the object group is no longer output to the final risk list. Alternatively, the status can be marked as pending data for verification in the background interface until the feedback event information is returned and then the process is resumed.

[0061] The methods for obtaining feedback event information include initiating secondary retrieval or incremental queries from the original external data source, sending manual verification work orders to the platform administrator to request confirmation of the subject's identity, event time, and role information, and supplementing with higher-credibility records from authorized internal business systems. After the feedback event information arrives, it is re-categorized into the corresponding feedback object group, and the process of calculating association strength, aggregating comprehensive association strength, and determining risk level is repeated to ultimately obtain the conflict of interest level corresponding to the feedback association strength.

[0062] For example, the closed-loop repair mechanism is used to enable a traceable correspondence with the original event unit to be fed back. Therefore, a unique feedback task number can be assigned to each event unit to be fed back, and information such as the original abnormality cause, fields to be supplemented, conflict type, and associated original record number can be written into the work order or query request. After the feedback is returned, the repair result is backfilled into the original object group according to the feedback task number, ensuring that the recalculation is performed in the original analysis context.

[0063] Based on this, this application uses the critical range as a trigger to focus limited computing resources and manual review efforts on error-prone boundary data. By using circuit breaking and closed-loop feedback to prevent contamination of dirty data to downstream processes, it can effectively address the judgment distortion caused by defects in surface edge data. This allows for self-correction and dynamic repair when facing complex and ambiguous boundary situations, thereby improving the reliability and credibility of conflict of interest determination.

[0064] In some specific embodiments of the e-commerce platform information security protection method provided in this application, after the above step S140, the method may also include, but is not limited to, step S810.

[0065] Step S810: If the conflict of interest between the candidate expert and the preset number of candidate suppliers is detected to be of a high-risk or medium-risk level, the corresponding candidate expert will be marked as a high-risk candidate expert and removed.

[0066] For example, after determining the conflict of interest levels for all object groups, conflict distribution statistics are performed using the candidate expert object as the primary key. This involves iterating through all object groups associated with that candidate expert object and counting the number of candidate suppliers whose conflict of interest reaches a high-risk or medium-risk level. If this number reaches or exceeds a preset limit, it indicates that the candidate expert object has an intertwined interest with the supplier group of the current bidding project. The candidate expert object is then marked as a high-risk candidate expert object and directly removed from the candidate expert list for this bidding project. If the removal of this expert results in an insufficient number of valid candidate experts, an expert re-selection process can be triggered. New candidate expert objects are reintroduced, and multi-source information retrieval and conflict of interest determination are re-executed until a review committee meeting compliance requirements is formed.

[0067] The preset excess quantity can be an absolute value or a relative proportion to the total number of current bidding suppliers. The preset excess quantity can be obtained by the platform management side based on the sensitivity level of the project, the sufficiency of the expert database, and the strictness of risk prevention.

[0068] Based on this, this application adopts a global elimination mechanism based on conflict diffusion to avoid systemic risks caused by multiple conflicts between candidate experts and multiple suppliers in a single project. By pre-setting an excessive number of experts with common interests, the application removes experts from the expert group, ensuring the independence, impartiality, and authority of the review committee as a whole. This avoids configuring a large number of cumbersome local permission restrictions for individual experts in subsequent stages, reduces the computational load and operational complexity of the permission system, and improves the overall security control efficiency of the e-commerce platform in complex bidding scenarios.

[0069] In some specific embodiments of the e-commerce platform information security protection method provided in this application, regarding the above step S150, the method may also include, but is not limited to, step S910.

[0070] Step S910: For candidate expert objects in the object group with a low-risk conflict of interest level, generate control instructions for full access to the bidding project data of candidate suppliers in the same object group. For candidate experts in an object group with a medium-risk conflict of interest level, control instructions are generated to restrict access to bidding project data of candidate suppliers in the same object group, based on the corresponding association strength and the event type of the associated event information in the object group. For candidate experts in a high-risk conflict of interest group, a pre-set data security protection mechanism is used to generate control instructions that do not grant access to the bidding project data of candidate suppliers in the same group. Restricting access permissions is used to configure field-level or operation-level access restrictions in access control lists.

[0071] For example, a tiered permission configuration strategy can be adopted for different conflict levels. For low-risk conflict levels, if the weak association between the expert and the supplier is determined to be insufficient to affect the fairness of the review, a control instruction with full access permissions is generated, allowing the expert to normally view all tender documents, quotation details, technical solutions, and qualification materials submitted by the supplier. For high-risk conflict levels, a preset data security protection mechanism is triggered, generating a control instruction with no access permissions. At the permission level, any request from the expert to the supplier's data is completely blocked. The review system front end will cover or lock the supplier's tender entries, preventing the expert from opening files or obtaining metadata. For medium-risk conflict levels, control instructions restricting access permissions can be adaptively generated based on the strength of the association and the specific event type of the associated event information. Restricting access permissions is used to issue field-level or operation-level access control policies in the Access Control List (ACL).

[0072] Regarding field-level access restrictions, sensitive fields can be dynamically anonymized or filtered based on the type of associated event. For example, if the associated event type indicates that the expert and the supplier have a technical cooperation background in the same industry, fields such as core patent formulas and key technical parameters in the supplier's bid documents can be anonymized or hidden on the expert's interface, and only a technical summary that meets the basic review requirements can be displayed. If the associated event type is business dealings, sensitive business fields such as the total bid price, the breakdown of itemized bids, and commercial discount terms can be blocked to prevent the leakage of sensitive business information.

[0073] Regarding operational-level access restrictions, the types of operation permissions for experts can be restricted in the access control list. For example, experts can be allowed to browse the supplier's general qualifications and some solutions within the system, but are prohibited from performing operations such as local downloading, batch exporting, screen recording, printing, and text copying. At the same time, a dynamic anti-counterfeiting security watermark, including the expert's name, access time, and IP address, should be forcibly loaded on the expert's interface, and audit logging should be enabled in the background to record the entire process of the expert's access to the supplier's data.

[0074] Based on this, this application seamlessly maps conflict of interest levels to underlying access control lists, forming a multi-level security protection network that covers full access, dynamic control, and comprehensive blocking. By configuring field-level and operation-level restriction policies in medium-risk conflict scenarios, it effectively prevents the substantial leakage and improper access of sensitive core data, ensures experts' legitimate access to non-sensitive necessary review materials, avoids review interruptions caused by brute-force interception, balances data security protection and business review efficiency, and improves the flexibility and security in dealing with complex interest relationships.

[0075] Secondly, this application provides an information security protection system for an e-commerce platform, comprising: The information acquisition module is used to acquire multiple candidate expert objects and multiple candidate suppliers, and to perform multi-database retrieval based on multiple candidate expert objects to obtain multi-source event information for each candidate expert object. The associated event module is used to combine each candidate expert object with each candidate supplier to obtain multiple object groups, determine multi-source event information that has a direct or indirect relationship with both the candidate expert objects and candidate suppliers in the object group, and obtain the associated event information of each object group. The association strength module is used to identify the association depth between associated event information and the corresponding object group, and to determine the event type of the associated event information. Based on the association depth and event type of the associated event information, the association strength between the associated event information and the corresponding object group is determined. The conflict level module is used to determine the conflict level between candidate expert objects and candidate suppliers in the object group corresponding to the association strength, based on the association strength and the preset judgment threshold. The instruction control module is used to generate control instructions based on the conflict of interest level to determine the access permissions of candidate expert objects in the corresponding object group to the bidding project data of candidate suppliers in the same object group.

[0076] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for protecting information security on an e-commerce platform, characterized in that, include: Multiple candidate expert objects and multiple candidate suppliers are obtained, and multi-database retrieval is performed based on the multiple candidate expert objects to obtain multi-source event information for each candidate expert object; Each candidate expert is combined with each candidate supplier to obtain multiple object groups. Multi-source event information that has a direct or indirect relationship with both the candidate expert and the candidate supplier in the object group is identified, and the associated event information of each object group is obtained. Identify the association depth between the associated event information and the corresponding object group, determine the event type of the associated event information, and determine the association strength between the associated event information and the corresponding object group based on the association depth and event type of the associated event information; Based on the correlation strength and the preset judgment threshold, the conflict of interest level between candidate expert objects and candidate suppliers in the object group corresponding to the correlation strength is determined; Based on the conflict of interest level, control instructions are generated to determine the access permissions of candidate expert objects in the corresponding object group to the bidding project data of candidate suppliers in the same object group.

2. The information security protection method for e-commerce platforms according to claim 1, characterized in that, The determination of multi-source event information that has a direct or indirect relationship with both candidate expert objects and candidate suppliers in the object group yields the associated event information for each object group, including: The multiple candidate expert objects and multiple candidate suppliers are used as nodes, and the corresponding candidate expert objects and candidate suppliers are connected through the multi-source event information to generate a relationship network diagram; In the relationship network diagram, multi-source event information that directly connects candidate expert objects and candidate suppliers in the object group or indirectly connects them through intermediate events is determined to obtain the associated event information of each object group. The intermediate events include events in which candidate expert objects and candidate suppliers participate.

3. The information security protection method for e-commerce platforms according to claim 1, characterized in that, The association depth includes first-order association depth and multi-order association depth; The process of identifying the association depth between the associated event information and the corresponding object group includes: The algorithm identifies the associated event information of candidate expert objects and candidate suppliers in the object group that are directly connected by a preset graph traversal algorithm, so as to determine the first-order association depth between the corresponding associated event information and the corresponding object group. The algorithm also identifies the connection depth level of the associated event information that is indirectly connected to candidate expert objects and candidate suppliers in the object group, so as to determine the multi-order association depth between the corresponding associated event information and the corresponding object group based on the connection depth level.

4. The information security protection method for e-commerce platforms according to claim 3, characterized in that, The step of determining the association strength between the associated event information and the corresponding object group based on the association depth and event type of the associated event information includes: The current time constant is determined by the time distance between the occurrence time of the associated event information and the current time. The current association depth weight is determined according to the association depth of the associated event information. The corresponding preset relationship type weight is extracted according to the event type of the associated event information. The current association depth weight corresponding to the first-order association depth is greater than the current association depth weight corresponding to the multi-order association depth. The association strength between the associated event information and the corresponding object group is obtained by multiplying the current time constant, the current association depth weight, and the preset relationship type weight.

5. The information security protection method for e-commerce platforms according to claim 4, characterized in that, The step of obtaining the association strength between the associated event information and the corresponding object group based on the product of the current time constant, the current association depth weight, and the preset relationship type weight includes: The individual association strength between each association event information and the corresponding object group is obtained by multiplying the current time constant, the current association depth weight, and the preset relationship type weight. The associated event information of the object group is combined into an event unit, and the comprehensive association strength between the object group and the corresponding event unit is obtained based on the individual association strength of each associated event information of the object group. The conflict of interest levels include high-risk, medium-risk, and low-risk conflict levels. Determining the conflict of interest levels between candidate experts and candidate suppliers within the object group corresponding to the association strength, based on the association strength and a preset judgment threshold, includes: If the overall correlation strength of an event unit exceeds a preset high-risk threshold, determine the high-risk conflict level between candidate expert objects and candidate suppliers in the object group corresponding to the event unit. If the overall correlation strength of an event unit is less than a preset high-risk threshold but greater than a preset low-risk threshold, determine the medium-risk conflict level between candidate expert objects and candidate suppliers in the object group corresponding to the event unit. If the overall correlation strength of an event unit is less than the preset low-risk threshold, the low-risk conflict level between candidate expert objects and candidate suppliers in the object group corresponding to the event unit is determined.

6. The information security protection method for e-commerce platforms according to claim 5, characterized in that, The step of combining the associated event information of the object group into event units, and obtaining the comprehensive association strength between the object group and the corresponding event units based on the individual association strength of each associated event information of the object group, includes: Extract the occurrence scenario level and object role information of each related event, and combine the event type, occurrence scenario level and object role information of each related event to obtain the event characteristics of each related event; The associated event information of the object group is combined into an event unit. By comparing the clustering of event features of the associated event information within the event unit, the associated event information within the event unit with a clustering similarity greater than a preset feature matching threshold is combined into a homogeneous relationship cluster, and the associated event information within the event unit with a clustering similarity less than a preset feature matching threshold is marked as independent event information. The association event information that has an individual association strength less than a preset association truncation threshold and belongs to the same homogeneous relationship cluster is removed. The maximum value of the individual association strength of each association event information in the removed homogeneous relationship cluster is extracted as the calculation benchmark value. The calculation benchmark value and the number of association event information in the removed homogeneous relationship cluster are input into a preset logarithmic saturation function to obtain the cluster association strength of the removed homogeneous relationship cluster. The values ​​of the cluster association strength that are greater than the preset low-risk threshold are adjusted to the values ​​of the preset low-risk threshold. Based on the cluster association strength of each homogeneous relation cluster after removal and the individual association strength of each independent event information, the comprehensive association strength between the object group and the corresponding event unit is obtained.

7. The information security protection method for e-commerce platforms according to claim 5, characterized in that, Before determining the conflict of interest level between candidate expert objects and candidate suppliers in the object group corresponding to the association strength based on the association strength and the preset judgment threshold, the e-commerce platform information security protection method further includes: If the difference between the overall correlation strength of an event unit and the preset high-risk threshold or the difference between the overall correlation strength of an event unit and the preset low-risk threshold is within the preset critical range, the event information missing analysis and timestamp logic conflict analysis will be performed on the related event information in the corresponding event unit to obtain the analysis results. If the analysis results indicate that there are missing event information or timestamp logic conflicts in the related event information of each event unit, the corresponding event unit will be marked as a feedback event unit and the acquisition of the conflict of interest level of the feedback event unit will be stopped. After obtaining the feedback event information of the corresponding event unit to be fed back, the feedback object group corresponding to the feedback event information is determined, the feedback association strength between the feedback event information and the corresponding feedback object group is determined, and the conflict of interest level between the candidate expert objects and candidate suppliers in the feedback object group corresponding to the feedback association strength is determined according to the feedback association strength and the preset judgment threshold.

8. The information security protection method for e-commerce platforms according to claim 5, characterized in that, After determining the conflict of interest level between candidate expert objects and candidate suppliers in the object group corresponding to the association strength based on the association strength and the preset judgment threshold, the e-commerce platform information security protection method further includes: If a conflict of interest is detected between a candidate expert and a candidate supplier exceeding the preset limit, and the conflict level is either high-risk or medium-risk, the corresponding candidate expert will be marked as a high-risk candidate expert and removed.

9. The information security protection method for e-commerce platforms according to claim 5, characterized in that, The step of generating control instructions based on the conflict of interest level to determine the access permissions of candidate expert objects in the corresponding object group to the bidding project data of candidate suppliers in the same object group includes: For candidate experts in an object group with a low-risk conflict of interest level, generate control instructions to grant them full access to bidding project data of candidate suppliers in the same object group. For candidate experts in an object group with a medium-risk conflict of interest level, control instructions are generated to restrict access to bidding project data of candidate suppliers in the same object group, based on the corresponding association strength and the event type of the associated event information in the object group. For candidate experts in a high-risk conflict of interest group, a pre-set data security protection mechanism is used to generate control instructions that do not grant access to the bidding project data of candidate suppliers in the same group. The access restriction settings are used to configure field-level or operation-level access restrictions in the access control list.

10. An information security protection system for an e-commerce platform, characterized in that, include: The information acquisition module is used to acquire multiple candidate expert objects and multiple candidate suppliers, and perform multi-database retrieval based on the multiple candidate expert objects to obtain multi-source event information of each candidate expert object. The associated event module is used to combine each candidate expert object with each candidate supplier to obtain multiple object groups, determine multi-source event information that has a direct or indirect relationship with both the candidate expert objects and candidate suppliers in the object group, and obtain the associated event information of each object group. The association strength module is used to identify the association depth between the associated event information and the corresponding object group, and to determine the event type of the associated event information. Based on the association depth and event type of the associated event information, the association strength between the associated event information and the corresponding object group is determined. The conflict level module is used to determine the conflict level between candidate expert objects and candidate suppliers in the object group corresponding to the association strength based on the association strength and the preset judgment threshold. The instruction control module is used to generate control instructions based on the conflict of interest level to determine the access permissions of candidate expert objects in the corresponding object group to the bidding project data of candidate suppliers in the same object group.

Images