Insurance Fishing Exercise Evaluation System Based on Behavioral Boundaries
By building a knowledge base for insurance business scenarios and identifying employee behavior boundaries, the problem of data distortion in phishing drills was solved, enabling precise evaluation and dynamic adjustment of employees' safety capabilities.
Patent Information
- Application Number
- CN202611502836.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-24
- Publication Date
- 2026-11-10
AI Technical Summary
In existing phishing email security drills, the automated access behavior of email security gateways and automatic scanning devices is mixed with the actual behavior of employees in the same record, resulting in distorted evaluation data. Furthermore, traditional methods based on single results such as clicks and reports are insufficient to reflect changes in employee behavior as risks escalate.
The scenario building unit acquires the insurance business scenario knowledge base and employee job tags, generates scenario parameter groups, uses the behavior collection and trusted chain generation unit to identify non-employee behaviors and form a trusted behavior chain, combines the behavior boundary judgment unit to identify safety rollback points, the adaptive orchestration unit adjusts scenario parameters, and the evaluation and update unit updates employee safety files.
It improves the authenticity of fishing drill behavior data and the precision of employee safety capability assessment, and can dynamically adjust drill content to revolve around the boundaries of actual employee behavior, identifying the safe withdrawal position of employees during the progression of risks.
Smart Images

Figure CN123074835A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of network security and information security behavior assessment technology, and more specifically, to an insurance phishing exercise assessment system based on behavioral boundaries. Background Technology
[0002] Existing phishing email security drills typically pre-configure drill email templates and corresponding difficulty parameters. After the drill emails are sent to target employees, the system records behaviors such as opening emails, accessing links, page operations, and security reporting to calculate employee click rates, reporting rates, or whether preset risky operations have occurred, and forms an employee security awareness evaluation result based on this. Some systems also adjust subsequent drill scenarios or drill difficulty based on historical evaluation results.
[0003] However, in the actual email environment of insurance institutions, drill emails typically pass through email security gateways, link rewriting services, and automated security scanning programs during delivery and access. These security devices may automatically access drill links before employees actually view the emails, causing automated scanning behavior and actual employee access behavior to be mixed in the same drill record, distorting the basic data for subsequent behavior evaluation. Meanwhile, existing solutions often rely on single results such as whether a click or report was made as the primary evaluation criterion, making it difficult to reflect when employees identify anomalies and stop subsequent operations during the progression of risk. Furthermore, when multiple factors such as business consistency, time urgency, and the sensitivity of the requested operation change simultaneously, it is difficult to determine the specific scenario factors that actually cause changes in employee security behavior. Summary of the Invention
[0004] To address the problems in existing phishing drill evaluation processes, such as non-employee access behavior generated by email security gateways and automatic scanning easily interfering with the evaluation of real employee behavior, and the difficulty of traditional evaluation methods based on single results like clicks and reports in depicting the actual safe retreat positions of employees during the escalation of risk, as well as the difficulty in determining the specific scenario parameters that lead to changes in employee behavior, this invention provides an insurance phishing drill evaluation system based on behavioral boundaries. The system includes: Scenario building unit: Used to acquire the insurance business scenario knowledge base, the job tags and historical security records of employees participating in the exercise, extract multiple candidate insurance business scenarios corresponding to the current employee from the insurance business scenario knowledge base based on the job tags, configure a scenario parameter group for each candidate insurance business scenario, and obtain a scenario candidate set; the scenario parameter group includes at least business event type, job adaptation tag, business consistency level, time urgency level, and interaction sensitivity level. Behavior Acquisition and Trusted Chain Generation Unit: Used to generate scenario instance identifiers and employee session identifiers for each exercise; based on the scenario instance identifiers and employee session identifiers, collect corresponding multi-source behavior events from the data terminal; identify non-employee behavior events according to the occurrence sequence, access concurrency relationship and terminal interaction confirmation status of each multi-source behavior event; delete the non-employee behavior events from the multi-source behavior events; and form a trusted behavior chain according to the occurrence sequence of the remaining behavior events. Behavior boundary determination unit: used to map each multi-source behavior event in the trusted behavior chain to multiple preset risk progression states, obtain the risk state sequence of the employee in the current exercise, identify the state transition position in the risk state sequence as the safety retreat point; and determine the employee's personal challenge boundary range based on the safety retreat points and scenario parameter groups of the same employee in multiple historical exercise scenarios. Adaptive orchestration unit: used to select the next round of target insurance business scenarios from the scenario candidate set based on the individual challenge boundary interval, the job label, and the job adaptation label of each candidate insurance business scenario, and adjust the scenario parameter group of the target insurance business scenario; when the dominant scenario parameter causing the change of the safety rollback point cannot be determined based on the historical trusted behavior chain, the scenario parameter to be detected is determined from at least two candidate scenario parameters, and a pair of detection scenarios is generated where only the scenario parameter to be detected is different and the other scenario parameters remain consistent. This is used to obtain the trusted behavior chain formed by the employee in response to the paired detection scenario, and to determine the sensitive scenario parameters of the employee based on the difference between the highest risk state, the safety backoff point and the timeliness of the safety event reporting in the two trusted behavior chains. Evaluation and Update Unit: Based on the sensitive scenario parameters, the security rollback points, and the historical trusted behavior chain, update the employee's security profile and the scenario effectiveness correlation between insurance business scenarios and positions, and feed back the updated security profile to the scenario construction unit and the adaptive orchestration unit.
[0005] This invention first correlates and filters the authenticity of multiple events, including email delivery, gateway access, terminal interaction, and security reporting. Then, based on a trusted behavior chain, it identifies the actual security retreat points where employees transition from escalating risk to ceasing access or proactively reporting. Building upon this, it establishes individual challenge boundaries using historical retreat points. When the dominant influencing factor cannot be determined, it proactively constructs paired detection scenarios, changing only one parameter to obtain behavioral results with a correlation, thereby further identifying sensitive factors in employee scenarios. This improves the authenticity of phishing drill behavioral data and the precision of employee security capability evaluation, enabling subsequent drill content to be dynamically adjusted around the boundaries of actual employee behavior.
[0006] Furthermore, the insurance business scenario knowledge base includes multiple insurance business scenario skeletons and variable parameter fields corresponding to each insurance business scenario skeleton; The scene construction unit is specifically used for: Obtain the normal business process stage corresponding to the current employee's job title tag; Candidate insurance business scenarios are obtained by filtering the target job type from the insurance business scenario knowledge base, which matches the job tag and whose business process stage corresponds to the normal business process stage. Obtain the historical behavior states corresponding to different scenario parameters in the historical security file, and initialize the variable parameter field with the historical behavior states to form the scenario parameter group; The scenario candidate set is obtained based on multiple candidate insurance business scenarios that have completed initialization.
[0007] By breaking down the actual business operations of insurance institutions, including supervision, underwriting, claims settlement, reinsurance, and customer service, into relatively stable business scenario frameworks and independently variable scenario parameter fields, the exercise scenarios maintain a semantic correspondence with employee positions and normal business processes. Simultaneously, they can independently change inducing factors such as familiarity with the source, consistency with business operations, and time urgency without altering the basic business structure. Compared to using a uniform template or selecting exercise content, this approach reduces low response times caused by scenarios irrelevant to employees' actual roles and provides a structured foundation for subsequent pairwise probing to individually modify specific scenario parameters.
[0008] Furthermore, the behavior acquisition and trusted chain generation unit is specifically used for: Before sending the target exercise email, an original access identifier corresponding to the scenario instance identifier is generated for the target exercise email, and a first mapping relationship is established between the original access identifier, the employee session identifier and the target employee. When the target exercise email passes through the email security gateway, obtain the link rewrite record and automatic scan log generated by the email security gateway, and establish a second mapping relationship between the original access identifier and the access identifier rewritten by the email security gateway. When an access event targeting the exercise email is detected, the email sending log, email security gateway log, and employee terminal access log are associated with the same scenario instance based on the first mapping relationship and the second mapping relationship. Get the temporal position of each access event relative to the email delivery completion time, the employee email viewing time, and other access events; get the number of accesses generated by the same access source for different links within a preset scanning time window. If an access event occurs before an employee views an email, or if the same access source accesses multiple different objects in the target exercise email consecutively within the preset scanning time window, and no employee terminal interaction confirmation event is detected during the corresponding access process, then the access event will be marked as a machine scan candidate event. The machine scan candidate event is sent to the interactive review process, and before the interactive review process is completed, the machine scan candidate event is set to a pending confirmation state so that it is temporarily excluded from the update of employee safety profiles and personal challenge boundary ranges.
[0009] By leveraging the mapping between the original access identifier and the gateway-rewritten access identifier, accesses modified by the email security gateway are traced back to the corresponding employees and specific exercise scenarios. Furthermore, by combining the access time, the number of accesses in a short period, and the employee's terminal interaction status, automatic scanning of candidate events is used to prevent machine actions such as email security gateway pre-scanning and link security checks from being directly recorded in employee click logs. Simultaneously, setting events with uncertain origins to a pending confirmation status instead of immediately writing them into the employee's security profile reduces the cumulative contamination of erroneous behavior records on individual challenge boundaries.
[0010] Furthermore, the methods for obtaining the employee terminal interaction confirmation event include: Scan candidate events for machines in the pending confirmation state to obtain employee terminal interaction confirmation events, email security gateway scan logs, and access source information; When the machine scan candidate event does not have an employee interaction confirmation identifier, and any one of the access timing, access source, or batch access characteristics matches the email security gateway scan log, the machine scan candidate event is determined to be a non-employee behavior event. When a machine scan candidate event contains an employee interaction confirmation flag, it is converted into a manual access event; The non-employee behavior events are excluded from the multi-source behavior events of the same scenario instance, and the trusted behavior chain is constructed according to the occurrence sequence of the remaining manual access events, email opening events, security event reporting events, and controlled drill page operation events.
[0011] Based on the initial temporal screening of candidate events by machine scanning, further verification is performed using terminal interaction confirmation events. This ensures that access events are only retained as human actions when genuine evidence of employee terminal interaction can be obtained. Events without human interaction but that can be correlated with gateway scan logs are removed from the behavior chain. This two-stage identification approach—candidate screening, interaction verification, and event characterization—reduces misjudgments caused by relying solely on IP addresses or access times, allowing subsequent risk status sequences to be built on a more realistic foundation of employee operation events.
[0012] Furthermore, the risk progression status is set according to the interaction depth between the employee and the target exercise email, including at least the exposure status, email viewing status, access trigger status, controlled page interaction status, and risk operation confirmation status; the system also sets security response statuses corresponding to the risk progression statuses, including at least the access termination status and the proactive reporting status. The behavior boundary determination unit is specifically used for: Traverse the trusted behavior chain according to the order of occurrence of behavioral events to determine the highest risk state of the employee in the current exercise. Before or after reaching the highest risk state, detect whether there is a behavioral event that changes the current risk state to a terminated access state or an active reporting state. When the aforementioned behavioral event occurs, the state transition position where the risk convergence trend first forms is determined as the safe pullback point; When an employee has not entered the access trigger state and there is an active reporting event, the security rollback point is marked as a pre-identification rollback point; When an employee has entered the access trigger state, but terminates the access or actively reports before entering the risk operation confirmation state, the security rollback point is marked as a process identification rollback point. When a safety response is generated only after an employee has reached the risk operation confirmation state, the safety pullback point is marked as a delayed identification pullback point. The highest risk state and the safe withdrawal point are used to jointly describe the employee's actual safety response capability in the current insurance business scenario.
[0013] The system expands employee behavior from a traditional binary of click / no click outcome into a sequence of risk states with a clear progressive relationship. The point where an employee first stops the risk progression is designated as the safety pullback point. This allows the system to not only know whether an employee ultimately reports the risk, but also to distinguish whether the employee identified the anomaly while viewing the email, after clicking, or after completing the risky action. This improves the behavioral granularity of employee safety awareness evaluation, enabling employees with different risk processes but the same final reporting result to receive different capability descriptions.
[0014] Furthermore, the specific steps for determining the personal challenge boundary range include: For the same employee, historical exercise records belonging to the same insurance business scenario type or having the same dominant scenario parameters are sorted according to the level of scenario challenge; The highest scenario challenge level for continuously maintaining either a pre-identification type pullback point or a process identification type pullback point is obtained from the historical exercise records and used as the safety identification boundary. The minimum level of challenge in a scenario where a delayed identification pullback point occurs or the risk operation confirmation state is reached is used as the risk breach boundary. The personal challenge boundary range is obtained based on the range of scenario parameters between the security identification boundary and the risk breach boundary.
[0015] By utilizing the changes in the safe withdrawal positions of the same employee across multiple historical drills, the upper limit for maintaining an effective safety response and the lower limit for when a significant risk breach begins to occur can be determined. These two limits constitute the employee's current personal challenge boundary range. This allows subsequent drills to focus on the boundary areas that truly differentiate employees' safety capabilities, avoiding drills that are too simple and lack differentiation, or drills that are too difficult and cause employees to generally enter a risky state, thus losing their evaluative significance.
[0016] Furthermore, the generation of the paired detection scenes specifically includes: Based on the employee's historical trusted behavior chain, identify at least two candidate scenario parameters that are related to the change in the safety rollback point; Select one scenario parameter to be detected from the candidate scenario parameters, and generate a first detection scenario and a second detection scenario with the same insurance business scenario skeleton; The business event types, job matching tags, and other scenario parameters (excluding the parameters of the scenario to be detected) of the first and second detection scenarios are kept in a consistent state, and the parameters of the scenario to be detected are placed at different parameter levels to generate the paired detection scenarios.
[0017] To address situations where multiple scenario parameters change simultaneously in historical drills, making it impossible to pinpoint the specific source of influence, paired detection scenarios with univariate differences are proactively generated. This ensures that both drills maintain the same business scenario framework, job suitability conditions, and other inducing parameters, changing only the level of the parameters in the scenario to be detected. This reduces the interference of business content, job differences, and other scenario parameters on the comparison results, providing behavioral data with a comparative relationship for subsequent assessments of whether a specific parameter actually drives the relocation of employee safety withdrawal points.
[0018] Furthermore, the specific steps for determining the sensitive scenario parameters of the employee include: The system acquires the first and second trusted behavior chains generated by employees for the first and second detection scenarios, respectively; it also acquires comparison data of the first and second trusted behavior chains, including the highest risk state, safety retreat point, first safety response time, and active reporting status. When at least one comparison data point crosses the security identification boundary or risk breach boundary, the scene parameter to be detected is determined as a sensitive scene parameter. Pairwise probing is performed on other candidate scenario parameters that could not be ruled out in turn to form a set of sensitive parameters for employees.
[0019] The system compares the highest risk status, safe retreat location, safe response time, and proactive reporting status of the same employee in paired detection scenarios, using the employee's own safety identification boundary and risk breach boundary as judgment criteria. Only when changing a scenario parameter actually causes the behavioral outcome to cross the boundary is that parameter identified as a sensitive scenario parameter, thereby improving the correlation between the sensitive factor identification results and the employee's actual risk behavior.
[0020] Furthermore, the specific steps for updating employee safety records include: The identification capability evaluation results are determined based on the employee's highest risk status and safe withdrawal point; The alertness assessment result is determined based on whether a safe termination operation is performed after reviewing the email and whether an event is proactively reported. The evaluation result of anti-induction ability is determined based on the state changes corresponding to the personal challenge boundary interval and sensitive scenario parameters. The movement direction is determined based on the safety identification boundary and risk breach boundary of multiple consecutive exercise cycles to determine the behavioral persistence evaluation results; The security profile is updated by integrating the evaluation results of recognition ability, vigilance, resistance to inducement, and behavior persistence.
[0021] The employee safety record will be expanded from a single click rate or drill score to include differentiated behavioral evaluation dimensions such as identification ability, vigilance, resistance to inducement, and behavioral persistence. This will reduce the impact of a single, incidental behavior on the employee's long-term safety record.
[0022] Furthermore, updating the scenario effectiveness correlation between the insurance business scenarios and job positions specifically includes: Based on the job tags and insurance business scenario types in the current trusted behavior chain, determine the target job-scenario combination; Based on the highest risk state and the safe withdrawal point, the behavior trigger type for the target position in the current insurance business scenario is determined. The behavior trigger type includes at least low response trigger, pre-identification trigger, process identification trigger, and risk breakthrough trigger. Write the behavior trigger type, security response status, and sensitive scenario parameters into the historical scenario response set of the target job-scenario combination; Based on the historical scenario response set, update the business trigger status, behavior differentiation status, and security response status of the target position-scenario combination; When the number of times the same sensitive scenario parameter causes the safety retreat point to move towards a high-risk state in the target job-scenario combination reaches a preset condition, the correlation between the sensitive scenario parameter and the target job-scenario combination is increased. Based on the updated business trigger status, behavior differentiation status, security response status, and the correlation degree of sensitive scenario parameters, the scenario effectiveness correlation between the target position and the insurance business scenario is updated.
[0023] The credible behavioral results generated from individual employee drills are further fed back to the job-to-insurance business scenario level. This allows the system to continuously assess whether a particular scenario has sufficient business authenticity for a specific job, whether it can differentiate between employees with different safety levels, and which scenario parameters are more likely to cause employees in that job to shift their risk boundaries. Scenario effectiveness is therefore no longer a pre-defined, manually assigned label, but can be continuously updated based on the real behavioral results of multiple employees. When new employees do not yet have sufficient personal historical data, the system can also leverage existing scenario effectiveness relationships within the same job to select initial drill scenarios with high triggerability and discriminative power, improving scenario adaptability during the cold start phase.
[0024] One or more technical solutions provided by this invention have at least the following technical effects or advantages: 1. This invention addresses the common email operating environment in insurance institutions' email security gateways, which often involves link rewriting, automatic pre-access, and security scanning. By mapping between scenario instance identifiers, employee session identifiers, and the original access identifier and rewritten access identifier, it associates email sending logs, email security gateway logs, and employee terminal behavior logs with the same scenario instance. Furthermore, it combines access timing, short-term batch access characteristics, and terminal interaction confirmation status to perform a secondary verification of machine scanning behavior, reducing the possibility of mistaking automatic access by security devices for genuine employee clicks, and providing a reliable behavioral data foundation for subsequent employee security capability evaluation.
[0025] 2. This invention transforms the behavioral process between employees and training emails into a continuous, progressively increasing risk state based on the depth of interaction. The point where the risk first progresses to termination of access or proactive reporting is used as the safety retreat point. This transforms the evaluation object from a single result indicator of whether or not the email was clicked or reported to a process indicator of where the employee discovers and stops an anomaly in the risk chain. Furthermore, by combining the safety identification boundary and the risk breach boundary to form a personal challenge boundary range, it can describe the differences in real safety behavior between different employees and between different stages of the same employee in a more granular way.
[0026] 3. This invention addresses the problem that multiple inducing parameters often change simultaneously in historical drills, making it difficult to determine which factor causes the deterioration of employee safety behavior. When the dominant scenario parameters cannot be determined, it constructs paired detection scenarios with only one different scenario parameter based on the same insurance business scenario framework. It identifies sensitive scenario parameters by the differences between the highest risk state, safety pullback point, safety response timeliness, and proactive reporting results. This further refines the types of emails on which employees are prone to problems into the specific inducing factors that drive them to cross the safety boundary.
[0027] 4. This invention further feeds back individual challenge boundaries and sensitive scenario parameters to the subsequent scenario construction and adaptive orchestration process, enabling the next round of insurance business scenarios to be selected based on the employee's current actual capability boundaries, rather than using a uniform template or a fixed difficulty increment method; at the same time, it updates the scenario effectiveness correlation between the job and the insurance business scenario with the credible behavioral results of multiple employees, so that individual assessment results and job group scenario experience can be continuously accumulated, forming a closed-loop evaluation mechanism of scenario construction—credible behavior collection—behavioral boundary judgment—sensitive parameter detection—file and scenario update—next round scenario reselection. Attached Figure Description
[0028] The accompanying drawings, which are provided to further illustrate embodiments of the invention and constitute a part of this invention, are not intended to limit the scope of the invention. Figure 1 This is the overall flowchart of the insurance fishing exercise evaluation system based on behavioral boundaries; Figure 2 This is a flowchart of behavior collection and trusted behavior chain generation; Figure 3 This is a flowchart for determining safe retreat points and personal challenge boundaries. Detailed Implementation
[0029] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, where there is no conflict, the embodiments of the present invention and the features thereof can be combined with each other.
[0030] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.
[0031] Example 1 refer to Figure 1This embodiment provides an insurance phishing exercise evaluation system based on behavioral boundaries. The system can be deployed in the internal network of an insurance institution or in a securely isolated exercise network, and establishes data connections with at least the email sender, email security gateway, employee terminal, security event reporting interface, and controlled exercise server.
[0032] The email sending client is used to send authorized security awareness training emails; the email security gateway is the insurance institution's existing email security equipment or email security service, used to perform link rewriting, malicious content detection, and automatic scanning; the employee terminal is the desktop email client, web email client, or mobile office terminal used by the employees participating in the training; the security incident reporting interface is the insurance institution's existing security reporting plugin, security operations platform interface, or dedicated reporting portal; the controlled training server is used to host the training pages and record preset behavioral events. The controlled training server does not deploy actual malicious payloads, and for simulated information input, it only records whether input occurred and whether submission was triggered, without saving the employee's real account, password, or customer business data.
[0033] The system includes: Scenario building unit: Used to acquire the insurance business scenario knowledge base, the job tags and historical security records of employees participating in the exercise, extract multiple candidate insurance business scenarios corresponding to the current employee from the insurance business scenario knowledge base based on the job tags, configure a scenario parameter group for each candidate insurance business scenario, and obtain a scenario candidate set; the scenario parameter group includes at least business event type, job adaptation tag, business consistency level, time urgency level, and interaction sensitivity level. Specifically, the insurance business scenario knowledge base includes multiple insurance business scenario skeletons and variable parameter fields corresponding to each insurance business scenario skeleton; The insurance business scenario framework defines the basic semantic structure of a type of email within the insurance business process. It includes at least four scenarios: regulatory compliance, internal management, underwriting, claims, reinsurance collaboration, and customer service. For example, regulatory compliance scenarios might include supplementing regulatory materials, verifying anti-money laundering documents, and conducting internal compliance checks; underwriting scenarios might include supplementing application materials, confirming underwriting opinions, and changing policy information. The specific business names mentioned above are only used to illustrate how the scenario framework is formed and do not require the scenario knowledge base to contain specific business names.
[0034] Each insurance business scenario framework includes the business initiator role type (e.g., internal supervisor, business collaboration department), target position type (target position), business object type (e.g., policy, claims, customer information), business process stage (the position where the email appears in the normal business process), normal business response time limit (the reasonable processing time limit for the business), and allowed business interaction types (whether the business normally requires viewing attachments, entering the internal system, or performing a confirmation operation). The variable parameter field is used to change the difficulty of the exercise and the behavioral induction characteristics without changing the basic semantic structure, so that multiple exercise instances with different parameter levels can be generated for the same business scenario. These parameters include at least the source familiarity (the degree of familiarity between the email initiator and the target employee's daily work relationship), the consistency of business content (the degree of matching between the email content and the target employee's current job responsibilities and recent business activities), the time urgency (the time pressure required by the email to respond), the sensitivity of the requested operation (the degree of proximity of the operation required by the email to the employee to a high-risk operation), and the type of interaction carrier (which carrier the employee's subsequent interaction occurs in, such as the email body, controlled link page, simulated attachment, or internal exercise entry).
[0035] In this embodiment, each variable parameter field can be represented by a discrete level, such as three levels: low, medium, and high, or more discrete levels can be used.
[0036] The scene construction unit is specifically used for: Based on the current employee's job label (which includes at least the job category, and may also include the business line and job authority scope), the system can obtain the normal business process stages corresponding to the job label. For example, if the target employee's job label is "claims review position", the system can obtain the normal business process stages corresponding to the claims review position, including the stages of material supplementation after case acceptance, responsibility review, amount review, compensation approval, and customer communication.
[0037] Candidate insurance business scenarios are obtained by filtering the target job type from the insurance business scenario knowledge base, where the target job type matches the job tag and the business process stage corresponds to the normal business process stage. Specifically, a first-level filtering can be performed according to the target job type, and a second-level filtering can be performed according to the business process stage. When a scenario skeleton allows multiple jobs to participate, the business process branch that is directly related to the current job is retained.
[0038] The system retrieves historical behavioral states corresponding to different scenario parameters from the historical security archive, and initializes the variable parameter fields with these historical behavioral states to form the scenario parameter group. The historical security archive at least stores the scenario type, scenario parameter group, highest risk state, safety pullback point, first safety response time, proactive reporting status, and identified sensitive scenario parameters for each employee in historical drills. The system reads historical records similar to or comparable to the current candidate insurance business scenario to determine the typical risk state of the employee at different parameter levels, thereby selecting an initial level for the variable parameter fields of the current candidate scenario.
[0039] For example, if historical records show that a claims review employee is usually able to proactively report before clicking a link under normal business consistency and medium time urgency, but stops only after entering the controlled page multiple times under high business consistency and high time urgency, then when the system forms the next round of scenario candidate sets, it will no longer uniformly adopt the highest parameter level, but will configure several adjacent parameter levels around the employee's historical security identification boundary and risk breach boundary, so that the candidate scenarios simultaneously include scenario instances below the boundary, close to the boundary, and located within the boundary range.
[0040] The scenario candidate set is obtained based on multiple candidate insurance business scenarios that have completed initialization; each candidate insurance business scenario retains its scenario skeleton identifier, scenario parameter group, job matching tag and corresponding initial scenario challenge level for selection by the subsequent adaptive orchestration unit.
[0041] Behavior Acquisition and Trusted Chain Generation Unit: This unit generates scenario instance identifiers and employee session identifiers for each exercise. Based on these identifiers, it collects corresponding multi-source behavior events from the data source (including email sender, email security gateway, employee terminal, and security event reporting interface). It identifies non-employee behavior events generated by email preloading, email security gateway scanning, or automated security detection based on the occurrence sequence, access concurrency relationship, and terminal interaction confirmation status of each multi-source behavior event. It then deletes these non-employee behavior events from the multi-source behavior events and forms a trusted behavior chain according to the occurrence sequence of the remaining behavior events. In this embodiment, the email sending end outputs at least the email generation time, delivery time, and scenario instance identifier; the email security gateway outputs at least the link rewrite record, automatic scan time, and scan source identifier; the employee terminal outputs at least the authorized email viewing events and the interactive confirmation events generated by the controlled drill page; and the security event reporting interface outputs at least the reporting trigger time and the corresponding scenario instance identifier.
[0042] refer to Figure 2 , specifically: Before sending the target exercise email, an original access identifier corresponding to the scenario instance identifier is generated for the target exercise email, and a first mapping relationship is established between the original access identifier, the employee session identifier, and the target employee.
[0043] The original access identifier can be embedded in the internal parameters of the controlled exercise link, or it can be stored in a mapping table by the email sender to distinguish specific exercise instances generated for different employees using the same scenario template. The employee session identifier is used to associate multiple behavioral events of the same employee in this exercise, and it expires or is archived after the exercise ends.
[0044] When the target exercise email passes through the email security gateway, some email security gateways will replace the original access address in the email with the gateway rewritten address, and the gateway will automatically access the rewritten address to perform security checks. To avoid the inability to trace back to the original exercise instance after rewriting, the system obtains the link rewriting records and automatic scan logs generated by the email security gateway, and establishes a second mapping relationship between the original access identifier and the access identifier rewritten by the email security gateway; the second mapping relationship records at least the original access identifier, the access identifier after rewriting, the gateway scan time, and the gateway source identifier.
[0045] When an access event targeting a training email is detected, the system associates the email sending log, email security gateway log, and employee terminal access log with the same scenario instance based on the first and second mapping relationships. Thus, even if the email link has been rewritten by the enterprise gateway, the system can still determine which employee, which training session, and which business scenario a particular access corresponds to, without mixing access records from different employees or different training periods.
[0046] The system obtains the temporal position of each access event relative to the email delivery completion time, the employee's email viewing time, and other access events. It also obtains the number of accesses generated by the same access source for different links within a preset scanning time window. In other words, it not only looks at the access itself, but also at the time point before and after the email delivery and the employee opening the email, and whether the same source accessed many different links in a very short period of time. These time and quantity characteristics are used to determine whether the access is more like a human operation or an automatic scan.
[0047] The preset scan time window is used to identify automated behavior of continuously accessing multiple different links within a short period of time. Its length can be configured according to the actual scanning characteristics of the enterprise email security gateway.
[0048] If an access event occurs before an employee views an email, or if the same access source accesses multiple different objects in the target exercise email consecutively within the preset scanning time window, and no employee terminal interaction confirmation event is detected during the corresponding access process, then the access event will be marked as a machine scan candidate event. The machine scan candidate events are sent to the interactive review process. Before the interactive review process is completed, the machine scan candidate events are set to a pending confirmation status, so that they are temporarily excluded from the update of employee security profiles and personal challenge boundary ranges. The purpose of the pending confirmation status is to prevent unconfirmed access events from prematurely affecting the employee's security awareness evaluation. After the review is completed, the event is converted into either a non-employee behavior event or a manual access event.
[0049] Specifically, the methods for obtaining the employee terminal interaction confirmation event include: Scan candidate events for machines in the pending confirmation state to obtain employee terminal interaction confirmation events, email security gateway scan logs, and access source information; Employee terminal interaction confirmation events can be generated by a controlled practice page within an employee-authorized practice environment. When the controlled practice page gains focus, the system only records states related to the behavior confirmation, such as whether the page is visible, whether a pointer click has occurred, whether keyboard input has been performed, whether a controlled page button has been triggered, and whether there is a sequential relationship between adjacent interactive actions. It does not collect the actual text content entered by the employee. If at least two types of human interaction actions with a sequential relationship are detected within a preset interaction time range, an employee interaction confirmation identifier can be generated for the current access session, forming an employee terminal interaction confirmation event.
[0050] Access source information may include at least the gateway source identifier, the endpoint session source identifier, and the identifier of the automatically detected service registered within the enterprise, which are used to distinguish between access from the security infrastructure and access from the target employee session.
[0051] When the machine scan candidate event does not have an employee interaction confirmation identifier, and any one of the access timing, access source, or batch access characteristics matches the email security gateway scan log, the machine scan candidate event is determined to be a non-employee behavior event. Matching means that the corresponding characteristics can find a clear scan record of the same scenario instance or the same rewritten access identifier in the email security gateway scan log. For example, if the scan log has recorded that the email security gateway automatically accessed the link after the email was delivered, then the access event can be confirmed as a non-employee behavior event.
[0052] When a candidate event for machine scanning contains an employee interaction confirmation flag, it is converted into a manual access event. Even if the access event coincides with the gateway scanning period, as long as the continuous interaction status of the employee terminal can prove that the subsequent behavior comes from the target employee's session, the system will still retain the corresponding manual access part and will not exclude the employee behavior because the gateway has scanned the same link before.
[0053] From the multi-source behavioral events of the same scenario instance, the non-employee behavioral events are excluded. The trusted behavioral chain is then constructed according to the occurrence sequence of the remaining manual access events, email opening events, security incident reporting events, and controlled drill page operation events. The trusted behavioral chain is ordered primarily by event time, while also retaining the event source type and corresponding risk status mapping identifier. For example, a trusted behavioral chain can be represented as: email opening event—manual access event—controlled page gains focus—simulated input action occurs—access terminates event—active reporting event.
[0054] Behavior boundary determination unit: used to map each multi-source behavior event in the trusted behavior chain to multiple preset risk progression states, obtain the risk state sequence of the employee in the current exercise, identify the state transition position in the risk state sequence from risk continuing to progress to risk stopping or safety response as the safety retreat point; based on the safety retreat points and scenario parameter groups of the same employee in multiple historical exercise scenarios, determine the employee's personal challenge boundary range; refer to Figure 3 , specifically, The risk progression status is set according to the depth of interaction between the employee and the target exercise email, and includes at least the exposure status, email viewing status, access trigger status, controlled page interaction status, and risk operation confirmation status. The "Exposure" status indicates that the exercise email has successfully reached the target employee's email environment, but there is no evidence that the employee has viewed it. The "Email Viewing" status indicates that there is a credible event that the target employee has viewed or previewed the email. The "Access Triggered" status indicates that the employee has entered the controlled exercise link or triggered the corresponding interaction entry through the exercise email. The "Controlled Page Interaction" status indicates that the employee has performed further operations on the controlled exercise page, such as clicking on the page, simulating input actions, or viewing simulated attachments. The "Risk Operation Confirmation" status indicates that the employee has triggered a preset high-risk simulated operation, such as pressing the simulated submit button on the controlled page. The "Risk Operation Confirmation" status only records whether the operation occurred, and does not save the actual submitted content.
[0055] The system also sets security response states corresponding to the progressive risk states. These security response states include at least two: an access termination state and a proactive reporting state. The access termination state indicates that the employee proactively exits the controlled exercise page, cancels preset operations, or stops subsequent interactions before the risk progresses to a higher level. The proactive reporting state indicates that the employee marks or reports the email as suspicious through the security incident reporting interface. The security response state is not the next level in the progressive risk state, but rather a bypass state indicating that the risky behavior has begun to converge.
[0056] The behavior boundary determination unit is specifically used for: Traverse the trusted behavior chain according to the order of occurrence of the behavior events, match each behavior event to the corresponding risk progression state or security response state, and obtain a risk state sequence arranged in time. Determine the highest risk state an employee is in during the current exercise. The highest risk state refers to the deepest risk progression state the employee has actually reached in the current trusted behavior chain, not the highest risk level preset in the email itself. For example, if the scenario allows the employee to enter the simulated submission page, but the employee only opens the email and then actively reports it, their highest risk state is still the email viewing state; if the employee has entered the controlled page and performed a simulated input action but has not triggered submission, their highest risk state is the controlled page interaction state.
[0057] Before or after reaching the highest risk state, detect whether there is a behavioral event that changes the current risk state to a terminated access state or an active reporting state; when an employee continues to enter a higher risk state from a lower risk state, it indicates that the risk is still progressing; when an employee stops entering a higher risk state and triggers a terminated access state or an active reporting, it indicates that the risk begins to converge.
[0058] When the aforementioned behavioral event occurs, the state transition point where the risk convergence trend first forms is determined as the safety rollback point. The state transition point can be represented by the pre-transition risk state plus the safety response state, such as email viewing state – proactive reporting state, access triggering state – terminated access state, or controlled page interaction state – proactive reporting state. By retaining the pre-transition risk state, it is possible to distinguish at which point in the risk process the employee became aware of the anomaly.
[0059] When an employee does not enter the access trigger state and there is an active reporting event, the security rollback point is marked as a pre-identification rollback point; this type indicates that the employee has stopped the risk progression before actually entering the controlled link, for example, the employee finds the source or business content to be abnormal after checking the email and reports it directly.
[0060] When an employee has entered the access trigger state, but terminates access or actively reports it before entering the risk operation confirmation state, the security rollback point is marked as a process identification type rollback point. This type indicates that the employee initially failed to complete the identification at the email level, but discovered the anomaly after opening the controlled page or engaging in further interaction, and stopped before the high-risk operation was confirmed.
[0061] When a safety response is generated only after an employee has reached the risk operation confirmation state, the safety pullback point is marked as a delayed identification pullback point. This type is used to indicate that the employee has already passed the main risk operation position preset in this exercise, and then the access is terminated or reported.
[0062] The system uses the highest risk state and the safety rollback point to jointly describe an employee's actual safety response capability in the current insurance business scenario. For example, if two employees both proactively report events, and the first employee's highest risk state is email viewing and forms a pre-identification rollback point, while the second employee's highest risk state is controlled page interaction and forms a process identification rollback point, then the system will record the two as different behavioral results.
[0063] Specifically, the steps for determining the personal challenge boundary range include: For the same employee, historical exercise records belonging to the same insurance business scenario type or having the same dominant scenario parameters are sorted according to the level of scenario challenge; The dominant scenario parameter refers to the parameter that plays the main role in changing the level of scenario challenge in the current set of historical records. For example, if the same set of records mainly changes the level of time urgency while other parameters are at similar levels, then the level of time urgency can be used as the dominant scenario parameter for that set of records.
[0064] In this embodiment, the level of challenge in a scenario can be ranked using discrete levels. The system first determines the initial challenge level for each historical exercise based on the scenario parameter level table, and then corrects it based on the highest risk state actually reached by employees in the same position under that parameter combination. If a certain parameter combination consistently results in employees being identified as being in a low-risk state, its challenge level remains unchanged or decreases; if the same parameter combination, based on multiple reliable behavior records, shows that employees are more likely to enter a deeper risk state, its challenge level can be increased.
[0065] The highest scenario challenge level for continuously maintaining a pre-identification pullback point or a process identification pullback point is obtained from the historical exercise records and used as the safety identification boundary. For example, it can be required that no delayed identification pullback point or risk operation confirmation state occurs in several adjacent similar exercises, and then the highest challenge level is identified as the current safety identification boundary.
[0066] The minimum level of scenario challenge at which a delayed identification pullback point occurs or the risk operation confirmation state is reached is obtained as the risk breach boundary; the risk breach boundary is used to indicate at what level of challenge an employee has a real possibility of entering a major risk operation.
[0067] The personal challenge boundary range is obtained based on the range of scenario parameters between the security identification boundary and the risk breach boundary; Adaptive orchestration unit: used to select the next round of target insurance business scenarios from the scenario candidate set based on the individual challenge boundary interval, the job label, and the job adaptation label of each candidate insurance business scenario, and adjust the scenario parameter group of the target insurance business scenario; when the dominant scenario parameter causing the change of the safety rollback point cannot be determined based on the historical trusted behavior chain, the scenario parameter to be detected is determined from at least two candidate scenario parameters, and a pair of detection scenarios is generated where only the scenario parameter to be detected is different and the other scenario parameters remain consistent. Specifically, under normal orchestration conditions, the adaptive orchestration unit first eliminates candidate scenarios where the job suitability label does not match the current employee, and then prioritizes scenarios from the remaining candidate scenarios whose challenge level is within the individual challenge boundary range. When multiple scenarios meet the conditions, the selection can be made by combining the scenario cooling status and historical coverage, so as to avoid the same business theme appearing repeatedly in the short term, causing employees to obtain the correct results due to memorized templates rather than their safe identification ability.
[0068] When historical records indicate a significant change in the safety rollback point, but multiple scenario parameters also change simultaneously, the system includes these parameters that occurred concurrently with the rollback point change in candidate scenario parameters. For example, if an employee's historical records show a change from a pre-emptive rollback point to a process-oriented rollback point, which increases both the time urgency and the consistency of business content, the existing historical trusted behavior chain can only indicate that at least one of these two factors may have been involved in the change, and the dominant scenario parameter cannot yet be determined.
[0069] Based on the employee's historical trusted behavior chain, at least two candidate scenario parameters that are related to the change of the safety rollback point are determined; specifically, adjacent or comparable similar scenario records within a preset historical window can be traversed to record the scenario parameters that change synchronously when the safety rollback point moves backward; parameters that appear repeatedly in multiple rollback point shift records are retained as candidate scenario parameters.
[0070] Select one scenario parameter to be detected from the candidate scenario parameters, and generate a first detection scenario and a second detection scenario with the same insurance business scenario skeleton; The business event types, job matching tags, and other scenario parameters (excluding the parameters of the scenario to be detected) of the first and second detection scenarios are kept in a consistent state (i.e., other parameters are kept at the same level or within a preset consistency range), and the parameters of the scenario to be detected are kept at different parameter levels to generate the paired detection scenarios.
[0071] For example, when the parameter of the scenario to be detected is the time urgency level, both the first and second detection scenarios use the same claim materials to supplement the scenario skeleton, the same target position, the same business object type, and the same interaction carrier. The first detection scenario uses the normal business response time limit, while the second detection scenario uses a higher time urgency level. Non-critical fields such as simulated case number and date in the email can be replaced with equivalent ones, but without changing the business semantics and other parameter levels.
[0072] This is used to obtain the trusted behavior chain formed by the employee for the paired detection scenario, and to determine the sensitive scenario parameters of the employee based on the difference between the highest risk state, the safety backoff point and the timeliness of the safety event reporting in the two trusted behavior chains.
[0073] Specifically: The system acquires the first and second trusted behavior chains generated by employees for the first and second detection scenarios, respectively; it also acquires comparison data of the first and second trusted behavior chains, including the highest risk state, safety retreat point, first safety response time, and active reporting status. The initial security response time can be represented as the time from when an employee first trusts and views the email to when access is first terminated or the employee proactively reports the issue. This is used to distinguish between rapid identification and identification that occurs only after a longer period of time.
[0074] The comparative data is then assessed for directional consistency. If, after the parameter of the scenario to be detected is raised from a lower level to a higher level, the employee's highest risk state shifts to a deeper risk state, the safety retreat point shifts from an upfront identification type to a process identification type or a delayed identification type, the initial safety response time is significantly prolonged (e.g., exceeding a preset time), or the proactive reporting changes from existing to non-existent, then the change in this parameter is considered to provide behavioral evidence of a risk progression. Conversely, if the behavioral outcome does not change substantially, the parameter is not temporarily identified as a sensitive scenario parameter.
[0075] When at least one comparison data point crosses the security identification boundary or risk breach boundary, the scene parameter to be detected is determined as a sensitive scene parameter. For example, in the first detection scenario, the employee forms a pre-identification withdrawal point, while in the second detection scenario, the employee reaches the risk operation confirmation state. In this case, the second detection scenario has already shown that the risk has been breached.
[0076] Pairwise probing is performed on other candidate scenario parameters that could not be ruled out in turn to form a set of sensitive parameters for employees.
[0077] For parameters that do not cause boundary crossing after detection, they can be temporarily removed from the current candidate set; for multiple parameters that are identified as sensitive scene parameters, the system saves the parameter name, the parameter level when the boundary crossing occurs, and the corresponding scene type, instead of forcibly combining multiple parameters into a single score.
[0078] Evaluation and Update Unit: Based on the sensitive scenario parameters, the security rollback points, and the historical trusted behavior chain, update the employee's security profile and the scenario effectiveness correlation between insurance business scenarios and positions, and feed back the updated security profile to the scenario construction unit and the adaptive orchestration unit.
[0079] Specifically, the steps for updating an employee's safety record include: The evaluation result of an employee's identification capability is determined based on their highest risk status and safe rollback points. If an employee can establish proactive identification rollback points while viewing emails in multiple different business scenarios, the evaluation result of their identification capability improves. If an employee frequently only generates a safety response after entering a controlled page interaction state or even after confirming a risky operation, the evaluation result of their identification capability decreases. Updates to identification capability primarily reflect how early in the risk chain anomalies an employee can detect.
[0080] The alertness rating is determined based on whether a security termination operation is performed after reviewing the email and whether a proactively reported event is generated. For records where not only is the operation stopped after an anomaly is detected, but the operation is also proactively reported through the security event reporting interface in a timely manner, the alertness rating is increased. For records where the operation is stopped for a long period of time without any reporting, the identification capability can be maintained, but the proactive alertness rating is not increased accordingly. For records where no security response is given after multiple reviews and the system continues to enter a deeper risk state, the alertness rating is decreased.
[0081] The evaluation result of resistance to induction is determined based on the state changes corresponding to the individual challenge boundary range and sensitive scenario parameters. Resistance to induction is not simply determined by whether or not a click is made, but rather by observing whether the employee can maintain their original safe retreat position when their known sensitive parameters increase. For example, if an employee has been identified as sensitive to time urgency, but is still able to maintain the pre-identified retreat point under high time urgency levels, it indicates that their resistance to induction factors has improved.
[0082] The behavioral persistence evaluation results are determined based on the movement direction of the safety identification boundary and risk breach boundary over multiple consecutive training cycles. If the safety identification boundary gradually moves towards a higher level of challenge over multiple training cycles, and the risk breach boundary moves towards a higher level of challenge or remains stable, it indicates that the employee's improved safety behavior is persistent. If the boundary increases for a period of time after training is completed, and then continuously falls back to a lower level of challenge, the behavioral persistence evaluation decreases and may trigger subsequent retraining strategies.
[0083] The security profile is updated by integrating the evaluation results of recognition ability, vigilance, resistance to inducement, and behavior persistence.
[0084] Specifically, updating the scenario effectiveness association between the insurance business scenarios and job positions includes: Based on the job tags and insurance business scenario types in the current trusted behavior chain, target job-scenario combinations are determined. A target job-scenario combination can correspond to a data unit in the scenario validity matrix, such as claims review position - customer complaint escalation scenario or underwriting position - supplementary insurance application information scenario.
[0085] Based on the highest risk state and the safe withdrawal point, the behavior trigger type for the target position in the current insurance business scenario is determined. The behavior trigger type includes at least low response trigger, pre-identification trigger, process identification trigger, and risk breakthrough trigger. Pre-identification trigger corresponds to pre-identification type withdrawal point; process identification trigger corresponds to process identification type withdrawal point; risk breakthrough trigger is used to indicate that the employee has formed a delayed identification type withdrawal point or reached a risk operation confirmation state.
[0086] The behavior trigger type, security response status, and sensitive scenario parameters are written into the historical scenario response set of the target job-scenario combination. Each historical scenario response record includes at least a scenario instance identifier, job tag, scenario type, scenario parameter group, behavior trigger type, security response status, highest risk status, security rollback point, and sensitive scenario parameters. To reduce the long-term impact of earlier data on the current business status, response records from the most recent preset period or the most recent preset number can be retained as the current update set according to the exercise time.
[0087] Based on the historical scenario response set, update the business trigger status, behavior differentiation status, and security response status of the target position-scenario combination; The business trigger status indicates whether a scenario can induce employees in that role to perform actual business processing actions. For example, if a large number of trusted records remain in the exposed state without being viewed or proactively reported, it indicates that the scenario may not be sufficiently relevant to the actual work of that role, and the business trigger status is reduced. If most employees at least view the email or make a valid access, the business trigger status is increased.
[0088] The behavioral differentiation status is used to indicate whether the same scenario can produce differentiated behavioral outcomes for different employees in the same position or at different security levels. If the historical scenario response set contains pre-identification triggers, process identification triggers, and a small number of risk breach triggers, it indicates that the scenario can map employees with different security capabilities to different risk positions, and the behavioral differentiation status is high. If almost all employees exhibit the same low response trigger, or almost all employees directly reach the risk breach trigger, then the scenario has a low differentiation effect on evaluating differences in capabilities.
[0089] The safety response status indicates the overall performance of employees in a given role in proactively responding to safety issues in a given scenario. It can be updated by considering whether proactive reporting occurred and the time of the first safety response. For example, if most employees in the same role proactively report issues in an early risk state, the safety response status of that role-scenario combination is good; conversely, if most reports are only generated after a deep risk state has occurred, or if there are no proactive reports for an extended period, the safety response status will be correspondingly lowered.
[0090] When the number of times the same sensitive scenario parameter causes the safety retreat point to move towards a high-risk state in the target job-scenario combination reaches a preset condition, the correlation between the sensitive scenario parameter and the target job-scenario combination is increased. The degree of correlation can be represented by discrete levels such as low, medium, and high. For example, in customer complaint scenarios involving multiple different claims staff, if increasing the time urgency level consistently results in the safety rollback point shifting from pre-identification to process identification, then the system will increase the correlation between time urgency level, claims review staff, and customer complaint scenario from medium to high.
[0091] If subsequent drills show that the same sensitive scenario parameter no longer causes the safety pullback point to shift backward, or if employees are generally able to maintain the forward-identifying pullback point after increasing the parameter, then the system can reduce the correlation between the sensitive scenario parameter and the target job-scenario combination.
[0092] Based on the updated business trigger status, behavior differentiation status, security response status, and the correlation degree of sensitive scenario parameters, the scenario effectiveness correlation between the target position and the insurance business scenario is updated.
[0093] Scenario validity correlation can be used for scenario cold start when personal historical data is insufficient. For example, for newly hired claims review employees, when their personal historical security profile is insufficient to determine the personal challenge boundary range, the scenario construction unit can select scenarios with high business trigger status and good behavior differentiation status from the scenario validity correlation corresponding to the claims review position as initial exercise scenarios, and use neutral parameter levels for the first test; after accumulating personal trustworthy behavior chains, it can then gradually transition to adaptive selection based on personal challenge boundary ranges.
[0094] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.
[0095] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
Claims
1. A safety fishing exercise evaluation system based on behavioral boundaries, characterized in that, The system includes: Scenario building unit: Used to acquire the insurance business scenario knowledge base, the job tags and historical security records of employees participating in the exercise, extract multiple candidate insurance business scenarios corresponding to the current employee from the insurance business scenario knowledge base based on the job tags, configure a scenario parameter group for each candidate insurance business scenario, and obtain a scenario candidate set; the scenario parameter group includes at least business event type, job adaptation tag, business consistency level, time urgency level, and interaction sensitivity level. Behavior Acquisition and Trusted Chain Generation Unit: Used to generate scenario instance identifiers and employee session identifiers for each exercise; based on the scenario instance identifiers and employee session identifiers, collect corresponding multi-source behavior events from the data terminal; identify non-employee behavior events according to the occurrence sequence, access concurrency relationship and terminal interaction confirmation status of each multi-source behavior event; delete the non-employee behavior events from the multi-source behavior events; and form a trusted behavior chain according to the occurrence sequence of the remaining behavior events. Behavior boundary determination unit: used to map each multi-source behavior event in the trusted behavior chain to multiple preset risk progression states, obtain the risk state sequence of the employee in the current exercise, identify the state transition position in the risk state sequence as the safety retreat point; and determine the employee's personal challenge boundary range based on the safety retreat points and scenario parameter groups of the same employee in multiple historical exercise scenarios. Adaptive orchestration unit: used to select the next round of target insurance business scenarios from the scenario candidate set based on the individual challenge boundary interval, the job label, and the job adaptation label of each candidate insurance business scenario, and adjust the scenario parameter group of the target insurance business scenario; when the dominant scenario parameter causing the change of the safety rollback point cannot be determined based on the historical trusted behavior chain, the scenario parameter to be detected is determined from at least two candidate scenario parameters, and a pair of detection scenarios is generated where only the scenario parameter to be detected is different and the other scenario parameters remain consistent. This is used to obtain the trusted behavior chain formed by the employee in response to the paired detection scenario, and to determine the sensitive scenario parameters of the employee based on the difference between the highest risk state, the safety backoff point and the timeliness of the safety event reporting in the two trusted behavior chains. Evaluation and Update Unit: Based on the sensitive scenario parameters, the security rollback points, and the historical trusted behavior chain, update the employee's security profile and the scenario effectiveness correlation between insurance business scenarios and positions, and feed back the updated security profile to the scenario construction unit and the adaptive orchestration unit.
2. The insurance fishing exercise evaluation system based on behavioral boundaries according to claim 1, characterized in that, The insurance business scenario knowledge base includes multiple insurance business scenario skeletons and variable parameter fields corresponding to each insurance business scenario skeleton. The scene construction unit is specifically used for: Obtain the normal business process stage corresponding to the current employee's job title tag; Candidate insurance business scenarios are obtained by filtering the target job type from the insurance business scenario knowledge base, which matches the job tag and whose business process stage corresponds to the normal business process stage. Obtain the historical behavior states corresponding to different scenario parameters in the historical security file, and initialize the variable parameter field with the historical behavior states to form the scenario parameter group; The scenario candidate set is obtained based on multiple candidate insurance business scenarios that have completed initialization.
3. The insurance fishing exercise evaluation system based on behavioral boundaries according to claim 2, characterized in that, The behavior acquisition and trusted chain generation unit is specifically used for: Before sending the target exercise email, an original access identifier corresponding to the scenario instance identifier is generated for the target exercise email, and a first mapping relationship is established between the original access identifier, the employee session identifier and the target employee. When the target exercise email passes through the email security gateway, obtain the link rewrite record and automatic scan log generated by the email security gateway, and establish a second mapping relationship between the original access identifier and the access identifier rewritten by the email security gateway. When an access event targeting the exercise email is detected, the email sending log, email security gateway log, and employee terminal access log are associated with the same scenario instance based on the first mapping relationship and the second mapping relationship. Get the temporal position of each access event relative to the email delivery completion time, the employee email viewing time, and other access events; get the number of accesses generated by the same access source for different links within a preset scanning time window. If an access event occurs before an employee views an email, or if the same access source accesses multiple different objects in the target exercise email consecutively within the preset scanning time window, and no employee terminal interaction confirmation event is detected during the corresponding access process, then the access event will be marked as a machine scan candidate event. The machine scan candidate event is sent to the interactive review process, and before the interactive review process is completed, the machine scan candidate event is set to a pending confirmation state so that it is temporarily excluded from the update of employee safety profiles and personal challenge boundary ranges.
4. The insurance fishing exercise evaluation system based on behavioral boundaries according to claim 3, characterized in that, The methods for obtaining the employee terminal interaction confirmation event include: Scan candidate events for machines in the pending confirmation state to obtain employee terminal interaction confirmation events, email security gateway scan logs, and access source information; When the machine scan candidate event does not have an employee interaction confirmation identifier, and any one of the access timing, access source, or batch access characteristics matches the email security gateway scan log, the machine scan candidate event is determined to be a non-employee behavior event. When a machine scan candidate event contains an employee interaction confirmation flag, it is converted into a manual access event; The non-employee behavior events are excluded from the multi-source behavior events of the same scenario instance, and the trusted behavior chain is constructed according to the occurrence sequence of the remaining manual access events, email opening events, security event reporting events, and controlled drill page operation events.
5. The insurance fishing exercise evaluation system based on behavioral boundaries according to claim 4, characterized in that, The risk progression status is set according to the interaction depth between the employee and the target exercise email, and includes at least the exposure status, email viewing status, access trigger status, controlled page interaction status, and risk operation confirmation status; the system also sets security response statuses corresponding to the risk progression statuses, and the security response statuses include at least the access termination status and the proactive reporting status. The behavior boundary determination unit is specifically used for: Traverse the trusted behavior chain according to the order of occurrence of behavioral events to determine the highest risk state of the employee in the current exercise. Before or after reaching the highest risk state, detect whether there is a behavioral event that changes the current risk state to a terminated access state or an active reporting state. When the aforementioned behavioral event occurs, the state transition position where the risk convergence trend first forms is determined as the safe pullback point; When an employee has not entered the access trigger state and there is an active reporting event, the security rollback point is marked as a pre-identification rollback point; When an employee has entered the access trigger state, but terminates the access or actively reports before entering the risk operation confirmation state, the security rollback point is marked as a process identification rollback point. When a safety response is generated only after an employee has reached the risk operation confirmation state, the safety pullback point is marked as a delayed identification pullback point. The highest risk state and the safe withdrawal point are used to jointly describe the employee's actual safety response capability in the current insurance business scenario.
6. The insurance fishing exercise evaluation system based on behavioral boundaries according to claim 5, characterized in that, The specific steps for determining the personal challenge boundary range include: For the same employee, historical exercise records belonging to the same insurance business scenario type or having the same dominant scenario parameters are sorted according to the level of scenario challenge; The highest scenario challenge level for continuously maintaining either a pre-identification type pullback point or a process identification type pullback point is obtained from the historical exercise records and used as the safety identification boundary. The minimum level of challenge in a scenario where a delayed identification pullback point occurs or the risk operation confirmation state is reached is used as the risk breach boundary. The personal challenge boundary range is obtained based on the range of scenario parameters between the security identification boundary and the risk breach boundary.
7. The insurance fishing exercise evaluation system based on behavioral boundaries according to claim 6, characterized in that, The generation of the paired detection scenarios specifically includes: Based on the employee's historical trusted behavior chain, identify at least two candidate scenario parameters that are related to the change in the safety rollback point; Select one scenario parameter to be detected from the candidate scenario parameters, and generate a first detection scenario and a second detection scenario with the same insurance business scenario skeleton; The business event types, job matching tags, and other scenario parameters (excluding the parameters of the scenario to be detected) of the first and second detection scenarios are kept in a consistent state, and the parameters of the scenario to be detected are placed at different parameter levels to generate the paired detection scenarios.
8. The insurance fishing exercise evaluation system based on behavioral boundaries according to claim 7, characterized in that, The specific steps for determining the sensitive scenario parameters of the employee include: The system acquires the first and second trusted behavior chains generated by employees for the first and second detection scenarios, respectively; it also acquires comparison data of the first and second trusted behavior chains, including the highest risk state, safety retreat point, first safety response time, and active reporting status. When at least one comparison data point crosses the security identification boundary or risk breach boundary, the scene parameter to be detected is determined as a sensitive scene parameter. Pairwise probing is performed on other candidate scenario parameters that could not be ruled out in turn to form a set of sensitive parameters for employees.
9. The insurance fishing exercise evaluation system based on behavioral boundaries according to claim 8, characterized in that, The specific steps for updating an employee's safety record include: The identification capability evaluation results are determined based on the employee's highest risk status and safe withdrawal point; The alertness assessment result is determined based on whether a safe termination operation is performed after reviewing the email and whether an event is proactively reported. The evaluation result of anti-induction ability is determined based on the state changes corresponding to the personal challenge boundary interval and sensitive scenario parameters. The movement direction is determined based on the safety identification boundary and risk breach boundary of multiple consecutive exercise cycles to determine the behavioral persistence evaluation results; The security profile is updated by integrating the evaluation results of recognition ability, vigilance, resistance to inducement, and behavior persistence.
10. The insurance fishing exercise evaluation system based on behavioral boundaries according to claim 8, characterized in that, Updating the scenario effectiveness correlation between the insurance business scenarios and job positions specifically includes: Based on the job tags and insurance business scenario types in the current trusted behavior chain, determine the target job-scenario combination; Based on the highest risk state and the safe withdrawal point, the behavior trigger type for the target position in the current insurance business scenario is determined. The behavior trigger type includes at least low response trigger, pre-identification trigger, process identification trigger, and risk breakthrough trigger. Write the behavior trigger type, security response status, and sensitive scenario parameters into the historical scenario response set of the target job-scenario combination; Based on the historical scenario response set, update the business trigger status, behavior differentiation status, and security response status of the target position-scenario combination; When the number of times the same sensitive scenario parameter causes the safety retreat point to move towards a high-risk state in the target job-scenario combination reaches a preset condition, the correlation between the sensitive scenario parameter and the target job-scenario combination is increased. Based on the updated business trigger status, behavior differentiation status, security response status, and the correlation degree of sensitive scenario parameters, the scenario effectiveness correlation between the target position and the insurance business scenario is updated.
