Data processing method and device of integrated virtual employee care saas platform

Abstract

The application provides a data processing method and device of an integrated virtual employee care Saas platform, and the method comprises the following steps: obtaining the associated data of an employee through the integrated virtual employee care Saas platform; the associated data comprises historical behavior data, working state data and emotional fluctuation data; the preset psychological state recognition algorithm is adjusted based on the associated data, and an adjusted psychological state recognition algorithm is obtained; the associated data is evaluated based on the adjusted psychological state recognition algorithm, and an employee psychological state evaluation result is generated; when the employee psychological state evaluation result triggers an early warning threshold, a care scheme information is automatically pushed; and the associated data is stored by using a dynamic encryption algorithm. In the application, the psychological state evaluation accuracy is improved, and the data security is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a data processing method and device of an integrated virtual employee care Saas platform. BACKGROUND

[0002] With the popularization of digital office and the upgrading of enterprise management concept, employee care has become the core link for enterprises to improve cohesion and productivity, especially in the fields of psychological health support and career development guidance. At present, some enterprises begin to use AI assisted tools or single function platforms to provide care services for employees, such as answering basic psychological problems through chat robots and conducting career planning evaluation by means of independent systems.

[0003] However, the existing technology has the following significant limitations:

[0004] Employee behavior data (such as platform usage records), work state data (such as task completion), and emotional fluctuation data (such as emotional expression in text / voice) are usually stored in different systems, lacking cross-dimensional correlation analysis, resulting in psychological state assessment relying on a single data source and insufficient accuracy.

[0005] Most platforms use general psychological state recognition algorithms, without considering individual differences (such as personality traits and job stress thresholds) and dynamic changes (such as career stages and life event influences), resulting in a large deviation between evaluation results and actual needs.

[0006] Employee emotional fluctuation data and psychological test results are highly sensitive information, and the existing storage method has the risk of data leakage, especially in data cross-platform transmission and long-term storage, with insufficient privacy protection. SUMMARY

[0007] The main purpose of the present application is to provide a data processing method and device of an integrated virtual employee care Saas platform, aiming to overcome the defects of inaccurate evaluation results caused by current psychological state assessment relying on a single data and easy data leakage.

[0008] To achieve the above purpose, the present application provides a data processing method of an integrated virtual employee care Saas platform, comprising the following steps:

[0009] Obtaining the correlation data of employees through an integrated virtual employee care Saas platform; the correlation data includes historical behavior data, work state data and emotional fluctuation data;

[0010] Based on the correlation data, the preset psychological state recognition algorithm is adjusted to obtain an adjusted psychological state recognition algorithm;

[0011] Evaluate the associated data based on the optimized psychological state recognition algorithm to generate an employee psychological state evaluation result, and automatically push care scheme information when the employee psychological state evaluation result triggers an early warning threshold;

[0012] The associated data is encrypted and stored by using a dynamic encryption algorithm.

[0013] Further, the integrated virtual employee care Saas platform integrates mental health support, career planning suggestions, social interaction services, and life guidance function modules into the same SaaS platform. The SaaS platform receives employee service requests through a unified interface and automatically matches corresponding function modules according to the request type.

[0014] Further, the emotional fluctuation data includes sentiment tendency values of text chat records and tone feature parameters of voice interaction.

[0015] Further, the psychological state recognition algorithm is optimized based on the associated data to obtain an optimized psychological state recognition algorithm, including:

[0016] Extract personalized strong correlation features from the associated data. In the same scenario, the Pearson correlation coefficient between the negative indicators in the emotional fluctuation data and the task delay rate in the work state data of the same employee is ≥0.7, and it appears stably ≥4 times in the past half year, which is marked as a personalized strong correlation feature.

[0017] Construct a correction factor based on the personalized strong correlation feature.

[0018] Embed the correction factor into the feature extraction layer of the preset psychological state recognition algorithm. For data matching to the personalized strong correlation feature, automatically increase its feature weight in the algorithm, and freeze the group general feature weight unrelated to the employee to obtain the optimized psychological state recognition algorithm.

[0019] Further, the calculation formula of the correction factor is:

[0020] Correction factor = (individual feature value - group mean) / group standard deviation × 0.6 + feature occurrence frequency × 0.4.

[0021] Further, the method further comprises:

[0022] According to the service usage frequency and interaction participation of employees on the platform, an integral incentive mechanism is established. Points can be exchanged for professional psychological assessment services or career planning courses, and through the social interaction module, employees can help each other and share experiences.

[0023] Further, the associated data is encrypted and stored by using a dynamic encryption algorithm, including:

[0024] sensitivity classification of the associated data;

[0025] Based on the classification corresponding to each associated data, determine the corresponding split rule, and split the corresponding associated data into 512-byte encrypted data blocks based on the split rule, and attach a unique check code to each encrypted data block;

[0026] The encrypted data blocks are stored in three physically isolated storage nodes, and any storage node stores no more than 40% of the total amount of encrypted data blocks;

[0027] When calling the associated data, the calling conditions need to be met:

[0028] The IP address of the calling node is within the preset whitelist;

[0029] Get the dynamic authorization token of the three storage nodes, and the token generation interval is within the preset time, and the token contains the recombination algorithm version number of the data block;

[0030] Through real-time face feature verification of employees;

[0031] After the three conditions are passed, the recombination algorithm is automatically triggered, and the complete data is spliced according to the check code time sequence.

[0032] Further, the associated data is encrypted and stored by using a dynamic encryption algorithm, comprising:

[0033] Calculate the data activity index of the associated data, and the calculation formula is: (access frequency in a preset period × 0.5) + (associated early warning trigger times × 0.3) + (data update frequency × 0.2);

[0034] When the data activity index is greater than a first preset value, the AES-256-GCM algorithm is used for encryption storage, and the anti-quantum computing hash lock is added, and the encryption is automatically performed once every preset time;

[0035] When the data activity index is less than a second preset value, the core data in the associated data is retained and the original detail data is deleted, and a lightweight encryption algorithm is used for encryption storage; The first preset value is greater than the second preset value;

[0036] When the data activity index is not greater than the first preset value and not less than the second preset value, a preset encryption algorithm is used for encryption storage.

[0037] Further, the associated data is encrypted and stored by using a dynamic encryption algorithm, comprising:

[0038] Obtain the multi-dimensional features of the associated data, and classify them to obtain first, second and third dimensional features;

[0039] obtaining an encryption algorithm array, wherein the encryption algorithm array is an array including a plurality of encryption algorithms;

[0040] deforming the encryption algorithm array based on the first dimension feature to obtain a deformed encryption algorithm array;

[0041] screening the deformed encryption algorithm array based on the second dimension feature to obtain a target encryption algorithm;

[0042] optimizing the target encryption algorithm based on the third dimension feature to obtain an optimized encryption algorithm, and encrypting and storing the associated data based on the optimized encryption algorithm.

[0043] The application further provides a data processing device of an integrated virtual employee care Saas platform, comprising:

[0044] an obtaining unit configured to obtain associated data of an employee through an integrated virtual employee care Saas platform, wherein the associated data comprises historical behavior data, working state data and emotional fluctuation data;

[0045] an optimization unit configured to optimize a preset psychological state recognition algorithm based on the associated data to obtain an optimized psychological state recognition algorithm;

[0046] an evaluation unit configured to evaluate the associated data based on the optimized psychological state recognition algorithm to generate an employee psychological state evaluation result, and automatically push care scheme information when the employee psychological state evaluation result triggers an early warning threshold;

[0047] a storage unit configured to encrypt and store the associated data by using a dynamic encryption algorithm.

[0048] The application provides a data processing method and device of an integrated virtual employee care Saas platform, comprising the following steps: obtaining associated data of an employee through an integrated virtual employee care Saas platform, wherein the associated data comprises historical behavior data, working state data and emotional fluctuation data; optimizing a preset psychological state recognition algorithm based on the associated data to obtain an optimized psychological state recognition algorithm; evaluating the associated data based on the optimized psychological state recognition algorithm to generate an employee psychological state evaluation result, and automatically pushing care scheme information when the employee psychological state evaluation result triggers an early warning threshold; and encrypting and storing the associated data by using a dynamic encryption algorithm. In the application, the integrated virtual employee care Saas platform is used to obtain the associated data of an employee in multiple dimensions, and the psychological state recognition algorithm is optimized, so that the accuracy of psychological state evaluation is improved; and the dynamic encryption algorithm is used to encrypt and store the associated data, so that the data security is ensured. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 is a data processing method step schematic diagram of an integrated virtual employee care Saas platform in an embodiment of the present application;

[0050] Figure 2 is a data processing device structure block diagram of an integrated virtual employee care Saas platform in an embodiment of the present application;

[0051] Figure 3 is a structure schematic block diagram of a computer device in an embodiment of the present application.

[0052] The implementation, functional characteristics and advantages of the present application will be further described with reference to the accompanying drawings in conjunction with embodiments. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0054] Referring to Figure 1 , an embodiment of the present application provides a data processing method of an integrated virtual employee care Saas platform, comprising the following steps:

[0055] Step S1, obtaining the associated data of an employee through an integrated virtual employee care Saas platform; the associated data comprises historical behavior data, working state data and emotional fluctuation data;

[0056] Step S2, based on the associated data, adjusting a preset psychological state recognition algorithm to obtain an adjusted psychological state recognition algorithm;

[0057] Step S3, based on the adjusted psychological state recognition algorithm, evaluating the associated data to generate an employee psychological state evaluation result, and when the employee psychological state evaluation result triggers an early warning threshold, automatically pushing care scheme information;

[0058] Step S4, using a dynamic encryption algorithm to encrypt and store the associated data.

[0059] In the present embodiment, as described in the above step S1, the purpose is to realize the integrated collection of multi-dimensional employee data, providing a basis for subsequent psychological state evaluation and algorithm adjustment. Specifically, the Saas (Software as a Service) platform synchronously obtains three types of core associated data through a unified data interface system:

[0060] Historical behavior data: includes the operation trajectory of employees within the platform, such as mental health counseling records (including counseling duration, problem type), access frequency and duration of career planning module, speech content of social interaction module and feedback (likes, comments and emotional tendency) to others' sharing, etc., which is realized through log recording and behavior embedding technology to realize full-process tracking;

[0061] Work status data: connected to enterprise internal OA, ERP and other systems, to obtain the task completion progress of employees (such as delay rate, quality score), overtime length and frequency, communication response speed in cross-department collaboration, and can be associated with wearable devices (such as smart bracelet) to synchronize physiological indicators (such as sleep duration, daily step count), forming the linkage data of work and physiological state;

[0062] Emotional fluctuation data: through natural language processing (NLP) technology to extract emotional keywords (such as "stress" and "anxiety") and semantic tendency of employees in text chat, combined with tone characteristics (such as pitch fluctuation rate and speech speed change) in voice interaction, and micro-expression characteristics (such as frown frequency and eye contact duration) in video interaction, to build a multi-modal emotional feature library.

[0063] After the above data is de-duplicated and cleaned, it is stored according to the time stamp and the unique identifier (such as the hash value of the employee number) of the employee to ensure the integrity and correlation of the data.

[0064] As described in step S2 above, the preset psychological state recognition algorithm is adjusted based on the associated data to obtain an adjusted psychological state recognition algorithm, and the recognition accuracy of the algorithm for the individual psychological state of the employee is improved through a dynamic learning mechanism. The core is to continuously optimize the algorithm parameters using the associated data. In one embodiment, it specifically includes:

[0065] Sample pool construction: effective training samples are selected from the associated data, which requires the sample to contain complete behavior-work-emotion three-dimensional features, and the time span covers at least 90 days, and is labeled according to emotion type-impact factor (such as "work stress type anxiety" and "life event type depression"), and the sample with a labeling consistency rate of ≥85% is included in the core training set;

[0066] Feature weight iteration: using gradient descent algorithm, the weight of the feature extraction layer of the algorithm is adjusted based on the individual data in the sample set. For example, if a certain employee's "continuous 3-day overtime of more than 4 hours" and "sudden increase of 30% of text negative word proportion" have strong correlation (Pearson correlation coefficient ≥0.7), then the weight of this combination feature in the algorithm is automatically increased;

[0067] Personalized calibration: For individual differences among employees, a two-dimensional model of personal baseline-group reference system is constructed. The average psychological state of employees after 3 months of employment is taken as the personal baseline, and the average data of the same post group is compared to calculate the deviation value. When the deviation continues to exceed 2 times the standard deviation of the group, a lightweight sub-model is customized for it, which realizes the individualization of the algorithm by freezing the general feature weight and strengthening the individual unique features (such as the emotional fluctuation rule before the project deadline).

[0068] It can be understood that through the parameter iteration every 30 days, the recognition accuracy of the algorithm for individual psychological state is improved by ≥10% compared with the initial version.

[0069] As described in step S3 above, the associated data is evaluated based on the tuned psychological state recognition algorithm to generate employee psychological state evaluation results, and when the warning threshold is triggered, the care scheme information is automatically pushed to realize the closed loop from data evaluation to intervention execution. The core is to output accurate evaluation results through the tuned algorithm, and to link the intelligent intervention mechanism.

[0070] The specific process is:

[0071] The tuned algorithm performs multi-dimensional analysis on the real-time collected associated data, and outputs evaluation results including emotion type (such as anxiety, burnout), influence degree (1-10 points), duration, and associated factors (work / life ratio). During the evaluation process, the algorithm preferentially calls the individual calibrated feature weight to ensure that the results fit the individual characteristics of the employees;

[0072] The preset three-level warning threshold (based on industry psychological crisis standards and enterprise historical data) is, for example: the first warning (mild fluctuation) is 10%-20% deviation of emotional score from the personal baseline, the second warning (significant fluctuation) is 20%-40% deviation and lasts ≥72 hours, and the third warning (high-risk fluctuation) is ≥40% deviation or detection of extreme emotional signals (such as self-harm tendency keywords);

[0073] According to the warning level, the differentiated scheme is automatically matched:

[0074] First-level warning: push preventive content, such as meditation audio based on historical preferences, time management tool recommendations;

[0075] Second-level warning: trigger AI psychological counseling robot real-time dialogue, synchronize push anonymous confession portal and professional psychological assessment questionnaire;

[0076] Third-level warning: immediately link to the pool of signed psychological consultants, push the contact information of professional personnel within a 30-kilometer range that can respond immediately, and generate an encrypted emotional snapshot (key features after desensitization) to assist rapid intervention.

[0077] At the same time, the platform records the employee's response data (such as click rate, completion rate) to the care program, providing feedback for subsequent algorithm optimization.

[0078] As described in step S4 above, the associated data is stored using a dynamic encryption algorithm, ensuring the security of the associated data (especially sensitive information) through a multi-level security mechanism, balancing privacy protection and data availability.

[0079] In one embodiment, it can include:

[0080] Data is divided into three levels according to sensitivity:

[0081] First-level sensitive data (such as psychological test raw scores, voice emotion feature sequences) uses a chaotic encryption algorithm, with a key generated from employee biometric features (such as fingerprint hash value), platform timestamp (accurate to milliseconds), and random prime number. The key is automatically rotated every 10 encrypted data, and the key fragments are distributed and stored in three physically isolated nodes.

[0082] Second-level sensitive data (such as work performance scores, overtime records) uses a hybrid mode combining AES-256 algorithm and homomorphic encryption, supporting statistical operations (such as monthly task completion rate calculation) without decryption. The key is bound to the employee's dynamic password (updated every 60 seconds).

[0083] Third-level sensitive data (such as service access frequency) uses regular hash encryption, only retaining statistical features.

[0084] A role-based access control matrix is established, for example: a psychologist accessing first-level data requires three levels of verification: "biometric features + dynamic password + approval authorization", single access duration ≤ 10 minutes, and operation logs are stored in real time on the chain for evidence. For first-level data that exceeds 180 days and has not triggered an early warning, automatic desensitization processing is performed (such as retaining emotion trend curves and deleting specific interaction content), reducing storage risks while retaining data analysis value. In this embodiment, through the combination of dynamic encryption and permission control, the security of data in the entire process of storage, transmission, and access is ensured.

[0085] In one embodiment, the integrated virtual employee care Saas platform integrates psychological health support, career planning suggestions, social interaction services, and life guidance function modules into the same SaaS platform. The SaaS platform receives employee service requests through a unified interface and automatically matches corresponding function modules according to the request type.

[0086] In this embodiment, the psychological health support module provides psychological testing, emotional counseling, emergency psychological intervention, and other services, including a professional psychologist interface channel and anonymous confession function.

[0087] Career planning suggestion module: Integrates career assessment tools, industry development information, skill improvement course recommendations, and other content to support employees in querying job promotion paths and analyzing the matching degree of their abilities.

[0088] Social interaction service module: Builds an interactive space for employees, including interest communities, experience sharing zones, team collaboration topic areas, and other functions such as real-time chat, dynamic publishing, and comment liking.

[0089] Life guidance function module: Provides life service recommendations (such as nearby convenience facilities and parenting activity information) and health management suggestions (such as diet planning and exercise plans).

[0090] The Saas platform receives service requests from employees through a unified access interface (such as a web portal, a mobile app, and a WeChat plug-in). When an employee initiates a request (for example, inputs "I am under a lot of pressure and need counseling" or "I want to know the promotion conditions"), the platform analyzes the request intent through natural language understanding technology, automatically matches it to the corresponding function module, and calls relevant service resources. For example, if an employee sends a text request saying "I am under a lot of pressure recently," the platform will directly jump to the mental health support module and prioritize displaying the emotion self-assessment tool and the instant counseling robot portal, eliminating the need for the employee to manually switch between different modules. This one-stop experience significantly reduces the operation process and improves the convenience of employee use.

[0091] In an embodiment, the emotional fluctuation data includes sentiment tendency values of text chat records and tone feature parameters of voice interactions.

[0092] In this embodiment, the sentiment tendency value of the text chat record refers to the result of sentiment quantification analysis of the text interaction content (such as one-on-one chat, community speech, and psychological counseling messages) of the employee in the platform through natural language processing (NLP) technology. Specifically, the text is processed by word segmentation, identifying explicit emotional words (such as "anxiety," "tiredness," and "happy") and implicit emotional expressions (such as "recently always insomnia" corresponding to "excessive stress"), and assigning a weight value to each word based on a pre-set sentiment dictionary (including positive, negative, and neutral word libraries) (for example, "collapse" is assigned a value of -0.8, and "relaxed" is assigned a value of +0.6). Finally, the sentiment tendency value of the entire text is calculated by weighting (the value range is usually -1 to +1, negative values indicate negative tendencies, positive values indicate positive tendencies, and the larger the absolute value, the higher the emotional intensity).

[0093] The tone feature parameter of voice interaction refers to the quantitative index of acoustic feature extraction of the voice signal of the employee in the voice call, voice message, and other scenarios. Key parameters related to tone are extracted through voice recognition technology, including:

[0094] Pitch fluctuation: the variation of the fundamental frequency (base frequency) in speech signals, for example, the pitch fluctuation amplitude is usually more than 30% when the emotion is excited;

[0095] Speech rate variation: the number of phonemes per unit of time, for example, the speech rate may be more than 20% higher than the normal level in an anxious state;

[0096] Pause interval duration: the time interval of natural pauses in speech, for example, the pause interval may be more than 1.5 times the normal level when the emotion is low;

[0097] The above parameters are converted into calculable feature vectors by algorithms such as Mel-frequency cepstral coefficients (MFCC) to reflect the emotional fluctuation state (such as tension, calmness, excitement, etc.) in speech.

[0098] In an embodiment, the preset psychological state recognition algorithm is tuned based on the association data to obtain a tuned psychological state recognition algorithm, comprising:

[0099] Extracting individualized strong association features from the association data; wherein the Pearson correlation coefficient of the negative indicators in the emotional fluctuation data and the task delay rate in the work state data of the same employee in the same scenario is ≥0.7, and appears stably ≥4 times in the past half year, marked as individualized strong association features;

[0100] Constructing a correction factor based on the individualized strong association features;

[0101] Embedding the correction factor into the feature extraction layer of the preset psychological state recognition algorithm, automatically increasing the feature weight of the data matched to the individualized strong association features in the algorithm, while freezing the group general feature weight irrelevant to the employee, to obtain the tuned psychological state recognition algorithm.

[0102] In this embodiment, first, the emotional and work state association mode with individual specificity and stable existence is selected from the association data to provide accurate individualized basis for algorithm tuning. Specifically, first, the association data of the employee is divided according to the scene dimension, which includes but is not limited to the project critical period, the quarterly performance evaluation period, the heavy meeting preparation period, and other situations with clear time nodes or work characteristics. In each scene, the negative indicators in the emotional fluctuation data (such as the proportion of negative words in text chat records, the pitch fluctuation rate in voice interaction, and the duration of negative emotions in micro-expression features) and the task delay rate in the work state data (the proportion of actual task completion time exceeding the planned time in a certain period) are extracted respectively. The correlation strength of the two is calculated by the Pearson correlation coefficient in statistics. When the calculation result meets the following conditions, it is marked as an individualized strong association feature:

[0103] Pearson correlation coefficient ≥ 0.7 indicates that there is a strong positive correlation between the two types of data (i.e., when negative indicators increase, task delay rates also increase significantly);

[0104] This correlation pattern appears ≥ 4 times in the past half year (180 days) and the scene type is consistent each time (e.g., all occurring in "3 days before project launch"), excluding accidental correlations and ensuring the stability and repeatability of the features.

[0105] For example, a certain employee has "text negative word proportion increased from 15% to more than 50%" and "task delay rate increased from an average of 5% to 30%" in the past half year, 4 times before the project launch, and the Pearson correlation coefficient is 0.78. This combination feature is marked as the employee's individual strong correlation feature.

[0106] Furthermore, based on the individual strong correlation feature, a correction factor is constructed to convert the individual strong correlation feature into a correction parameter that can directly affect the algorithm, achieving accurate measurement of feature importance. The construction of the correction factor takes the core attributes of the individual strong correlation feature as input and uses a weighted calculation model.

[0107] Then, the correction factor is embedded into the feature extraction layer of the preset psychological state recognition algorithm to obtain the optimized psychological state recognition algorithm. By integrating the correction factor and the algorithm structure, the individual features are strengthened and irrelevant features are suppressed, improving the recognition accuracy of the algorithm for individual psychological states. Specifically, the constructed correction factor is embedded into the feature extraction layer of the preset psychological state recognition algorithm (a module responsible for extracting key features from raw data and assigning initial weights), forming a dynamic adjustment mechanism:

[0108] 1. When the algorithm processes the correlation data that matches the marked individual strong correlation feature (e.g., detects that a certain employee is in the "3 days before project launch" scenario and simultaneously appears the corresponding negative indicators and task delay rate changes), the correction factor is automatically called to multiply the initial weight of the feature in the algorithm by the correction factor (e.g., the initial weight is 0.3 and the correction factor is 1.4, then the adjusted weight is 0.42), significantly improving its influence in feature fusion and model reasoning;

[0109] 2. At the same time, the algorithm automatically freezes the group general feature weights unrelated to the employee (e.g., typical stress features of other post employees, emotional features unrelated to the employee's work scenario), i.e., keeping the weight values of these features fixed and not participating in this optimization, avoiding the interference of group data on individual specific features.

[0110] Through this process, the preset psychological state recognition algorithm is converted from a universal model to a personalized optimization model, which can more sensitively capture the characteristic signals strongly related to the individual employee, reduce the interference of irrelevant information, and ultimately improve the accuracy of the optimized psychological state recognition algorithm by ≥10% compared to the pre-optimization.

[0111] In an embodiment, the calculation formula of the correction factor is:

[0112] Correction factor = (individual characteristic value - group mean) / group standard deviation x 0.6 + characteristic occurrence frequency x 0.4.

[0113] In this embodiment, the calculation formula of the correction factor is used to quantify the core attributes of the individual strong correlation characteristics as adjustment parameters that can be applied to the algorithm. The meanings and calculation logic of each part are as follows:

[0114] (individual characteristic value - group mean) / group standard deviation x 0.6: This part measures the deviation of individual characteristics from the group benchmark through statistical methods. The core is to standardize the individual characteristic value to a comparable deviation coefficient. The individual characteristic value refers to the quantified value of the individual strong correlation characteristic of a certain employee (e.g., the Pearson correlation coefficient between "emotional negative indicators" and "task delay rate" of the employee in a specific scenario);

[0115] The group mean is the average value of the characteristic of employees in the same position in the same scenario; the group standard deviation reflects the dispersion of the characteristic values of employees in the same position, and is used to eliminate the influence of internal differences in the group on the deviation calculation; the three are calculated by (individual - mean) / standard deviation to obtain the standardized deviation value, and then multiplied by the weight coefficient 0.6 to highlight the influence of the difference between the individual and the group on the correction factor. For example, if the characteristic value of a certain employee is 0.8, the group mean is 0.3, and the group standard deviation is 0.2, then the result of this part is (0.8-0.3) / 0.2x0.6=1.5.

[0116] Characteristic occurrence frequency x 0.4: This part strengthens the stability weight of the characteristic by the frequency of its occurrence. The characteristic occurrence frequency refers to the number of times the individual strong correlation characteristic has appeared stably in the past half year (subject to the prerequisite of ≥4 times). After being multiplied by the weight coefficient 0.4, it reflects the contribution of the stability of repeated occurrence of the characteristic to the correction factor. For example, if the characteristic occurs 5 times, then the result of this part is 5x0.4=2.0.

[0117] The sum of the results of the above two parts is the value of the correction factor. Through this formula, the deviation degree of the individual and the group and the stability of the characteristic occurrence are integrated into a single quantitative indicator, making the correction factor not only reflect the individual specificity of the characteristic, but also embody the reliability of its repeated occurrence. Finally, it is used to adjust the weight of the corresponding characteristic in the algorithm, improving the adaptability of the psychological state recognition algorithm to individual differences.

[0118] In an embodiment, the method further comprises:

[0119] According to the service use frequency and interactive participation of employees on the platform, an integral incentive mechanism is established, and the integral can be exchanged for professional psychological evaluation services or career planning courses, and the integral mutual assistance and experience sharing among employees are realized through a social interaction module.

[0120] In an embodiment, the associated data is stored by using a dynamic encryption algorithm, comprising:

[0121] The associated data is classified according to the sensitivity degree;

[0122] Based on the classification corresponding to each associated data, a corresponding splitting rule is determined, and the corresponding associated data is split into 512-byte encrypted data blocks based on the splitting rule, and a unique check code is attached to each encrypted data block;

[0123] The encrypted data blocks are stored in three physically isolated storage nodes, and the encrypted data blocks stored in any storage node do not exceed 40% of the total amount;

[0124] When the associated data is called, the calling conditions need to be met:

[0125] The IP address of the calling node is within the preset whitelist;

[0126] Get the dynamic authorization token of the three storage nodes, and the token generation interval is within the preset time, and the token contains the recombination algorithm version number of the data block;

[0127] Through real-time face feature verification of the employee;

[0128] After the three conditions are passed, the recombination algorithm is automatically triggered, and the complete data is spliced according to the check code time sequence.

[0129] In this embodiment, the associated data is first classified according to the sensitivity degree, aiming to classify the associated data according to the risk level caused by data leakage, and provide classification basis for subsequent encrypted storage. Specifically, the associated data is divided into three levels according to the privacy attribute of the data content and the enterprise data security specification:

[0130] First sensitive data: including voice emotion feature sequence in emotional fluctuation data, psychological test original score, emergency psychological intervention dialogue record, etc. If this kind of data is leaked, it may cause serious damage to the privacy of employees;

[0131] Second sensitive data: including project performance score, overtime length and task delay rate in work state data, and high-frequency access record of psychological health module in historical behavior data, etc. After the leakage, it may affect the professional image of employees;

[0132] Third-level sensitive data: including service access frequency in historical behavior data, like like records of public social interactions, etc. The privacy attribute is low, and the leakage risk is relatively controllable.

[0133] The grading process is automatically completed by a preset sensitive word library and a rule engine. For example, text data containing keywords such as "depression" and "anxiety" are automatically classified as first-level sensitive data. Manual review and adjustment are also supported to ensure the accuracy of grading.

[0134] Furthermore, based on the grading, the splitting rule is determined, and the data is split into encrypted data blocks and attached with a check code. This step improves the security and integrity of data storage through structured splitting and checking mechanisms. The specific process is as follows:

[0135] Determine the splitting rule: According to the data sensitivity level, develop a differentiated splitting strategy. First-level sensitive data is split by time slicing and feature dimension (e.g., every 30 seconds of voice data is a time slice, and then split by frequency, amplitude, and other feature dimensions). Second-level sensitive data is split by record entry (e.g., each task performance record is independently split). Third-level sensitive data is split by batch (e.g., daily access records are split after being aggregated).

[0136] Split and encrypt: Based on the above rules, the sensitive data of all levels is uniformly split into fixed-size (512 bytes) encrypted data blocks. During the splitting process, the corresponding encryption algorithm of the same level (e.g., SM9 algorithm for first-level data and AES-256 algorithm for second-level data) is used for encryption processing.

[0137] Add a unique check code: Generate a unique check code for each encrypted data block. This check code contains the original sequence number of the data block (identifying its position in the complete data), the timestamp (recording the splitting time), the sensitive level label, and the data block hash value, ensuring that the data block cannot be tampered with or replaced during storage and transmission.

[0138] Then, the encrypted data blocks are stored in three physically isolated storage nodes to reduce the risk of single-point leakage and ensure the security of physical data storage through distributed storage strategies. Specifically:

[0139] Three storage nodes with physical isolation properties are selected, including server rooms in different geographic locations (such as one each in North China, East China, and South China), servers of different network operators (such as one each of China Telecom, China Unicom, and China Mobile), and no direct data interaction link between nodes to avoid data complete disclosure due to a single node being compromised; a dynamic allocation algorithm is used to store the encrypted data blocks in the three nodes, and the storage proportion of a single node is strictly limited during the allocation process: the storage amount of each node for first-level sensitive data is not more than 35% of the total amount, and the storage amount of each node for second-level and third-level sensitive data is not more than 40%, ensuring that when any node is accessed illegally, the data blocks obtained cannot form complete information.

[0140] Finally, condition verification and data recombination when calling associated data are performed to ensure the legality of data calling and the safe reconstruction of complete data through a multi-dimensional verification mechanism, as follows:

[0141] Verify that the calling node IP is in the preset whitelist: a list of legal node IPs allowed to call data (such as enterprise intranet servers and authorized psychological counselor terminal IPs) is preset, and when a calling request is initiated, the IP address of the initiating node is first verified to be in the whitelist, and if not, the request is directly rejected, blocking illegal access at the network layer;

[0142] Verify the dynamic authorization token of the three storage nodes: after a legal calling request is triggered, the three storage nodes generate a dynamic authorization token, which includes the version number of the data block recombination algorithm (to ensure that the recombination logic matches the splitting logic), the token generation timestamp, and the node signature; verify that the token generation interval is within the preset time (such as ≤10 seconds to avoid long-term validity of the token after being intercepted), and the recombination algorithm version numbers of the three tokens are consistent, otherwise the token is considered invalid;

[0143] Verify the employee's real-time facial features: when calling personal sensitive data, the employee's real-time facial image is captured through a camera and compared with the pre-stored facial feature template (captured at the time of employment), and a matching degree ≥95% is considered to pass the verification, ensuring that the data calling subject is the employee himself or an authorized personnel.

[0144] When the above three conditions are met, the recombination algorithm corresponding to the version number in the token is automatically triggered, the dispersedly stored data blocks are spliced into complete original data in chronological order according to the original sequence number and timestamp in the encrypted data block check code, the integrity of each data block is verified through the hash value in the check code during the splicing process, and finally the complete and unaltered associated data is returned to the calling node.

[0145] In an embodiment, a dynamic encryption algorithm is used to encrypt and store the associated data, including:

[0146] calculating a data activity index of the associated data, and the calculation formula is: (access frequency in a preset period of time*0.5)+(number of times of triggering an associated early warning*0.3)+(data update frequency*0.2);

[0147] When the data activity index is greater than a first preset value, the AES-256-GCM algorithm is used for encryption storage, and a quantum-resistant hash lock is added, and automatic re-encryption is performed once every preset time;

[0148] When the data activity index is less than a second preset value, core data in the associated data is retained and original detail data is deleted, and a lightweight encryption algorithm is used for encryption storage;

[0149] When the data activity index is not greater than the first preset value and not less than the second preset value, a preset encryption algorithm is used for encryption storage.

[0150] In the embodiment, first, a data activity index of the associated data is calculated, the dynamic use frequency and importance of the associated data are evaluated through a quantitative index, and a basis is provided for subsequent adaptation of an encryption strategy. Specifically, the calculation of the data activity index takes three types of core indexes of the associated data in a preset period of time (such as 30 days) as input, and uses a weighted summation model, and the formula is:

[0151] Data activity index=(access frequency in a preset period of time*0.5)+(number of times of triggering an associated early warning*0.3)+(data update frequency*0.2);

[0152] Wherein, the access frequency in the preset period of time refers to the total number of times that the associated data is called and viewed by authorized users (such as employees themselves and psychological consultants), and the weight 0.5 reflects the core influence of the frequency of data access on the activity;

[0153] The number of times of triggering an associated early warning refers to the number of times that the data participates in psychological state assessment and triggers an early warning threshold, and the weight 0.3 reflects the close association between the data and the abnormal psychological state of the employee;

[0154] The data update frequency refers to the number of times (such as newly added text emotion records in emotional fluctuation data) of data addition or modification per unit time (such as per day), and the weight 0.2 reflects the degree of dynamic change of the data.

[0155] The higher the index value calculated by the formula, the higher the real-time, importance and use frequency of the data.

[0156] Further, when the data activity index is greater than a first preset value, encryption storage is performed, when the data activity index is greater than a first preset value (such as the preset value is 8, representing high activity data), a high-intensity encryption strategy is used to cope with the security risks brought by high-frequency access, specifically including:

[0157] Encryption algorithm selection: AES-256-GCM algorithm is used for encrypted storage, which combines high encryption strength (256-bit key) and integrity check function (generate authentication tag through GCM mode to prevent data tampering);

[0158] Anti-quantum computing enhancement: A hash lock based on lattice-based cryptography is attached to the encrypted data header. The complexity of breaking this hash lock increases exponentially with the increase of quantum computing power, resisting the risk of breaking traditional encryption algorithms by future quantum computers.

[0159] Automatic re-encryption mechanism: Set to trigger the re-encryption process automatically every pre-set time (such as 12 hours), generate a new random key (unrelated to the original key) during re-encryption, and update the hash lock synchronously. Even if the key is leaked at a certain time, it will only affect the encrypted data of that period, greatly reducing the risk of long-term storage.

[0160] Such high-activity data usually includes recent high-frequency access psychological counseling records, frequent trigger warning emotional fluctuation data, etc. High-intensity encryption strategy matches its high security demand.

[0161] When the data activity index is less than the second preset value, the encrypted storage is used. When the data activity index is less than the second preset value (such as the preset value is 3, representing low-activity data), a lightweight encryption and data reduction strategy is used to optimize storage efficiency while ensuring basic security, which includes:

[0162] Data reduction processing: Keep the core data in the associated data (such as emotional trend curve in emotional fluctuation data, monthly task completion rate average in work status data), delete the original detail data (such as specific chat text, hourly task progress record), only keep the key features supporting subsequent statistical analysis, reduce storage redundancy;

[0163] Lightweight encryption algorithm: Use a low-complexity encryption algorithm (such as SM3 hash algorithm combined with Base64 encoding), which meets the basic encryption needs (prevents unauthorized access) while reducing the system resource occupation of the encryption and decryption process.

[0164] Such low-activity data usually includes historical behavior records that have not been accessed for more than 6 months, early emotional data that has not triggered any warning, etc. The reduction and lightweight encryption strategy balances its security needs and storage costs.

[0165] The encryption storage when the data activity index is between the first preset value and the second preset value, when the data activity index is not greater than the first preset value and not less than the second preset value (i.e. between 3-8, representing medium activity data), the preset regular encryption algorithm is used for storage, which needs to meet:

[0166] The encryption strength is moderate (such as AES-128 algorithm), which balances security and computing efficiency;

[0167] Supports periodic key rotation (such as updating the key once a week), without high-frequency re-encryption;

[0168] Preserves the complete structure of the data (without simplification), ensuring that the data can be used for regular psychological state assessment and historical tracing.

[0169] Such medium activity data usually includes the recent career planning consultation records of employees, occasional access to work state data, etc. The regular encryption strategy is suitable for its security requirements and usage frequency, achieving a balance between security and efficiency.

[0170] In an embodiment, the associated data is stored by using a dynamic encryption algorithm, comprising:

[0171] Obtaining multi-dimensional features of the associated data and classifying them to obtain first, second and third dimensional features;

[0172] Obtaining an encryption algorithm array; wherein the encryption algorithm array is an array including a plurality of encryption algorithms;

[0173] Based on the first dimensional feature, the encryption algorithm array is deformed to obtain a deformed encryption algorithm array;

[0174] Based on the second dimensional feature, the deformed encryption algorithm array is screened to obtain a target encryption algorithm;

[0175] Based on the third dimensional feature, the target encryption algorithm is optimized to obtain an optimized encryption algorithm; and the associated data is stored by using the optimized encryption algorithm.

[0176] In this embodiment, first, multi-dimensional features of the associated data are obtained and classified, and multi-dimensional feature extraction and classification are used to provide fine basis for the adaptation of the encryption algorithm. Specifically, three types of core dimensional features are extracted from the associated data:

[0177] The first dimensional feature: reflects the sensitive properties of the data, including the privacy level involved in the data content (such as whether it contains psychological diagnosis results) and the impact degree that data leakage may cause (such as the potential risk to the employee's professional reputation), using a quantitative score of 1-10 (10 being the highest sensitivity);

[0178] Second dimension feature: embody the structural characteristics of data, including data format (text, voice, image), data volume (such as the number of text characters, voice duration), data update frequency (static data / dynamic real-time update data), for example, the structural characteristics of "5-minute voice emotion record" can be described as "audio format + 300KB + daily update";

[0179] Third dimension feature: represent the use scenario of data, including access subject (employee himself / administrator / psychologist), access frequency (such as daily access / monthly access), access purpose (daily view / early warning analysis / emergency intervention), for example, "historical emotional data called by psychologist for psychological crisis intervention" belongs to high emergency use scenario.

[0180] The classification process is automatically matched by feature labels, for example, text data containing negative tendency keywords, its first dimension feature score is 10 points, second dimension feature is "text + 500 words + real-time update", and third dimension feature is psychologist + emergency access.

[0181] Further, an encryption algorithm array is obtained, which builds a candidate pool of diversified encryption algorithms and provides a basis for subsequent dynamic adaptation. The encryption algorithm array is a collection of various encryption algorithms, covering different encryption strength, applicable scenarios and calculation complexity algorithm types, for example:

[0182] High security algorithm: SM4, AES-256, RSA-2048, etc., suitable for high sensitive data;

[0183] Lightweight algorithm: DES, RC4, SM3 hash algorithm, etc., suitable for low sensitive or large volume data;

[0184] Special scene algorithm: voiceprint encryption algorithm for voice data, homomorphic encryption algorithm for text data, etc.

[0185] The algorithm array is sorted in a two-dimensional matrix, which facilitates subsequent quick positioning of candidate algorithms according to data characteristics.

[0186] Further, the encryption algorithm array is deformed based on the first dimension feature, which adjusts the priority of the algorithm array according to the sensitive attributes of the data, and realizes the preliminary matching of encryption strength and sensitive level. Specifically:

[0187] For high-sensitive data with a first-dimension feature score of 8 or more, the priority of high-security algorithms (such as AES-256 and SM4) in the array is increased by 30% through weight adjustment, while the candidate weight of lightweight algorithms is reduced. For low-sensitive data with a first-dimension feature score of 3 or less, the weight is adjusted in the opposite direction, so that the priority of lightweight algorithms (such as RC4) is increased, and the occupation of system resources by high-intensity algorithms is reduced. The transformation process is achieved through dynamic sorting of the algorithm array. In the transformed array, the algorithm that best matches the data sensitivity level will be placed at the front. For example, in the transformed array for high-sensitive data, the AES-256 algorithm will be placed at the front.

[0188] Based on the second-dimension feature, the target encryption algorithm is selected. This step combines the structural characteristics of the data to select the single algorithm with the best adaptability from the transformed algorithm array. The specific selection logic includes:

[0189] For text data, algorithms that support variable-length block encryption (such as AES-256-CBC mode) are preferred to avoid padding redundancy caused by fixed block length.

[0190] For multimedia data such as voice and images, algorithms that support stream encryption (such as RC4) are selected to improve encryption efficiency.

[0191] For large-volume data (such as historical behavior records exceeding 1GB), algorithms with extremely high computational complexity (such as RSA-4096) are excluded, and algorithms that balance efficiency and security (such as AES-128) are selected.

[0192] For example, for data with a first-dimension feature score of 9 and a second-dimension feature of "text + 1000 words + daily update", the AES-256-GCM algorithm that supports variable-length text, medium efficiency, and high intensity is selected from the transformed array as the target encryption algorithm.

[0193] Finally, based on the third-dimension feature, the target encryption algorithm is optimized and encrypted for storage. According to the use scenario of the data, the algorithm parameters are optimized to achieve precise encryption storage.

[0194] In one embodiment, the optimization content includes:

[0195] Key management optimization: For emergency access scenarios (such as the third-dimension feature being "emergency intervention"), the key validity period is shortened from the default 7 days to 24 hours, and key distribution requires multiple approvals.

[0196] Encryption mode adjustment: For high-frequency access data (such as daily view of work status data), a "encryption + cache" hybrid mode is adopted, and a temporary decryption cache (validity period of 1 hour) is retained after the first encryption to reduce the computational overhead of repeated encryption.

[0197] Additional security mechanism: For the administrator access scenario, the access log encryption function is added based on the target algorithm, and the time, IP address and other information of the access behavior are recorded and stored in encrypted form, ensuring that the operation is traceable.

[0198] The optimized encryption algorithm is used to encrypt the associated data, and the encryption result is stored in association with the unique identifier of the data (such as employee ID + timestamp). At the same time, an encryption parameter log (including algorithm type, key version, and tuning parameters) is generated, providing a basis for subsequent decryption and security auditing.

[0199] In an embodiment, when the encrypted data blocks are stored in three physically isolated storage nodes, the three physically isolated storage nodes form a dynamic topology network, and the connection relationship between each storage node is randomly adjusted every preset time, and the node IP address is updated in a dynamic camouflage technology.

[0200] Each encrypted data block is stored with a topology location tag, which records its logical position in the current topology structure, and the topology location tag is automatically updated with the topology network adjustment;

[0201] When calling data, the current node connection relationship needs to be parsed through the preset topology key, and then the data block is located according to the topology location tag.

[0202] In this embodiment, the three storage nodes are first deployed in different data centers with a geographical distance ≥ 500 kilometers, and each node is configured with independent power supply, network link and security protection system, to ensure that a single physical location failure or attack will not affect other nodes.

[0203] A quantum random number generator (QRNG) is used to generate a node connection matrix every 5 minutes, which defines the logical connection relationship between nodes. The connection matrix contains the following parameters:

[0204] Master node (responsible for receiving external data access requests);

[0205] Data flow path (such as node A→node C→node B);

[0206] Backup link priority (automatically switched when the main link delay > 100ms);

[0207] Each time a new topology is generated, the node connection probability is adjusted through the small-world network model (Watts-Strogatz model) in graph theory, so that the network maintains local clustering characteristics and global connectivity, and the average path length is controlled between 2-3 hops.

[0208] The dynamic IP camouflage mechanism includes:

[0209] Quantum Key Distribution (QKD): Each storage node periodically exchanges session keys with other nodes through quantum entanglement channels, with a 90-second key update period, ensuring unconditional security of communication content.

[0210] IP Dynamic Transformation Protocol: Each node is configured with a pool of 1024 pre-generated virtual IP addresses, which are transformed every 30 seconds according to the following rules: a pseudo-random number generated using the Mersenne Twister algorithm is used as an IP index; the new IP address must satisfy a Hamming distance ≥ 16 bits from the previous IP; during the transformation process, zero-knowledge proof (ZKP) is used to prove the identity continuity to other nodes.

[0211] Traffic Obfuscation Strategy: During IP transformation, fake traffic packets (30% of total traffic) are continuously sent to unrelated IP addresses, with the size, frequency, and protocol type of the fake packets mimicking real data access patterns to interfere with external traffic analysis.

[0212] Topology Position Tag System:

[0213] Multi-dimensional Coordinate Encoding: The topology position tag of each encrypted data block contains four-dimensional coordinates:

[0214] Timestamp (accurate to nanoseconds, recording the data block generation time);

[0215] Node ID (quantum fingerprint hash value of the current storage node);

[0216] Link Vector (records the shortest path from the master node to this node);

[0217] Version Number (incremented each time the topology is adjusted to prevent misuse of old version tags).

[0218] Dynamic Tag Update: When the topology structure is adjusted, each node updates the data block tag through a distributed consensus algorithm (PBFT):

[0219] Master node broadcasts new topology structure and timestamp;

[0220] Each node calculates the new link vector of the locally stored data block;

[0221] Verify the legality of the update operation through Elliptic Curve Digital Signature Algorithm (ECDSA).

[0222] Label Encryption Storage: The topology position tag is encrypted using the same dynamic encryption algorithm as the data block, with the hash value of the current topology structure included in the key, ensuring the integrity of the binding between the tag and the topology.

[0223] Data Call and Positioning Process:

[0224] Topology key negotiation: When the client initiates a data access request, it needs to obtain the current topology key through a quantum key distribution protocol first. This key contains:

[0225] The Merkle tree root hash of the latest topology structure;

[0226] The public key certificate chain of each node;

[0227] The effective time window (accurate to the microsecond level).

[0228] Location tag analysis: The client uses the topology key to decrypt the topology location tag of the data block and verifies its validity through the following steps:

[0229] Check if the timestamp is within the current topology validity period;

[0230] Verify if the link vector conforms to the current topology connection relationship;

[0231] Compare the version number with the latest topology version;

[0232] Multi-path data retrieval: According to the parsed link vector, the client simultaneously sends parallel retrieval requests to multiple possible nodes and uses the following strategies to optimize response speed:

[0233] Prioritize the path with the lowest network delay;

[0234] Automatically switch to an alternative path when a certain path response times out (> 200ms);

[0235] Perform integrity verification (such as SHA-3 hash comparison) on the returned data block;

[0236] Through the above mechanisms, secure storage and efficient retrieval of encrypted data blocks in a dynamic topology network are achieved, making the data storage location have time dimension uncertainty (topology changes every 5 minutes) and spatial dimension confusion (IP address dynamic camouflage), significantly improving the data attack resistance and privacy protection level.

[0237] In an embodiment, when the encrypted data blocks are stored in three physically isolated storage nodes, a unique virtual storage address is assigned to each encrypted data block. The virtual storage address has no fixed mapping relationship with the physical storage node, and the virtual storage address is updated based on a random number generator every preset time, and the update record is only kept in the encrypted block of the blockchain;

[0238] The storage nodes transmit data through false data blocks, with a false data block ratio of 30%. The false data blocks have the same format as the real data blocks but are false content. When an abnormal access attempt is detected, the false data blocks are automatically returned to the attacker, and the real data blocks are silently migrated to the backup storage node.

[0239] In this embodiment, to achieve dynamic decoupling of encrypted data blocks and physical storage nodes, a unique virtual storage address is assigned to each encrypted data block, which is not fixedly associated with the physical node. The generation of the virtual address integrates multi-dimensional random sources, including quantum physical noise, precise timestamp of data generation, and device hardware features, which are expanded into an unpredictable address sequence through a chaotic algorithm, ensuring that the address itself does not carry any physical location information.

[0240] The virtual address is automatically updated at preset time intervals (such as 90 seconds), and the update process is driven by a random number generator. After each update, the new address is not logically associated with the historical address. All address change records are stored in the encrypted block of the blockchain after encryption, and the block is maintained by a distributed consensus mechanism. Only authorized nodes can verify the validity of the update through a specific key, and the historical records cannot be tampered with or forged.

[0241] The above design makes it impossible for external attackers to infer the physical storage location of the data through the virtual address. Even if the address information at a certain time is obtained, it will be invalid due to the timed update of the address, fundamentally breaking the static association between the virtual identifier and the physical location.

[0242] When transmitting data between storage nodes, a fixed proportion (30%) of false data blocks are mixed in. The false data blocks and real data blocks use the same format specification, including header identification, encryption method, and check structure, and only the content is invalid information simulating the distribution characteristics of real data. The generation of false data blocks is based on a statistical feature training model of real data, ensuring that the format, length, and encryption strength cannot be distinguished from real data blocks, thereby confusing the attacker's identification of valid information.

[0243] When an abnormal access attempt is detected (such as continuous illegal signature requests, access frequency or source that does not conform to normal behavior patterns), the defense mechanism is automatically triggered: returning carefully constructed false data blocks to the request source. These false data blocks not only have complete formats, but also contain seemingly reasonable but actually invalid content to mislead the attacker's analysis direction.

[0244] At the same time, the silent migration process of real data blocks is started. The migration process is carried out through an encrypted channel, targeting pre-deployed backup storage nodes. After migration is complete, the real data on the original node is completely removed, leaving only false data blocks. The entire migration process is completed without user awareness, ensuring that the attacker cannot obtain valid data and ensuring the continuity of subsequent access by authorized users.

[0245] When calling data, the connection relationship of the current storage node is first parsed through the preset topology key. The topology key contains the node association rules of the current network, is generated regularly and distributed encrypted, ensuring that only authorized subjects can obtain the effective connection information between nodes.

[0246] After the node connection relationship is determined, the specific data block is located according to the topology location label attached when the encrypted data block is stored. The topology location label records the logical position of the data block in the current dynamic topology network. This label is automatically updated with the adjustment of the connection relationship of the storage node, and always keeps consistent with the actual storage state of the data block. Through the cooperative action of the topology key and the topology location label, authorized users can accurately trace the current storage location of the data block, complete the legal calling and reorganization of the data, and attackers cannot locate the complete data even if they intercept part of the data block because they cannot obtain real-time topology information.

[0247] In an embodiment, a dynamic encryption algorithm is used to encrypt and store the associated data, including:

[0248] Updating the encryption algorithm based on the associated data;

[0249] Generating a first ternary tree structure based on the algorithm parameters of the updated encryption algorithm;

[0250] Generating a second ternary tree structure based on the data characteristics of the associated data;

[0251] Superimposing the first ternary tree structure and the second ternary tree structure, and sequentially comparing each node of the superimposed nodes; according to a pre-set rule, target nodes are selected from the compared nodes based on the comparison results of each node;

[0252] Generating an encryption key based on the character data on each target node; and encrypting and storing the associated data based on the encryption key.

[0253] In this embodiment, the core parameters of the basic encryption algorithm are dynamically adjusted by analyzing the feature changes of the associated data in real time, realizing real-time adaptation of the encryption logic and data characteristics. Specifically, the system continuously monitors the data volume fluctuation (such as the number of new records in a single day), the feature dimension change (such as the newly added voice feature parameter in the sentiment fluctuation data), and the sensitive level migration (such as the upgrade from level two sensitive to level one sensitive), and quantifies these changes as algorithm adjustment factors.

[0254] For example, when detecting that the proportion of high-sensitive fields (such as psychological test raw scores) in the associated data increases by more than 20%, the key length of the basic encryption algorithm is extended from 128 bits to 256 bits, and a more complex block cipher mode (such as switching from CBC mode to GCM mode) is enabled; if the data update frequency is reduced to less than 10 times per day, the algorithm's round function iteration number is appropriately simplified to optimize encryption efficiency while ensuring security. The algorithm update process is synchronized to all storage nodes through an encrypted channel, and the update record is written to an immutable audit log to ensure traceability of algorithm iterations.

[0255] The first ternary tree structure is a structured mapping of the core parameters of the updated encryption algorithm, used to convert abstract algorithm characteristics into calculable tree nodes. Extract the key parameters of the updated encryption algorithm, including key length, iteration rounds, block size, hash function type, etc., and convert each parameter into a character sequence according to a pre-set rule (e.g., "256-bit key" is converted to "K256", and "16-round iteration" is converted to "I16").

[0256] Subsequently, a ternary tree is constructed based on these character sequences: the root node is the unique identifier of the algorithm (e.g., the hash value of the algorithm name + update timestamp), each parent node derives three child nodes according to the logic "parameter type - parameter value - check code", and the depth of the child nodes is positively related to the importance of the parameters (e.g., the key length parameter corresponds to a node with a depth of 3, and the hash function type corresponds to a node with a depth of 2). Each node of the tree structure is attached with a check value based on SHA-256 to ensure that the parameter conversion and tree structure generation process have not been tampered with.

[0257] The second ternary tree structure is used to capture the characteristic attributes of the associated data, realizing the structured association between data characteristics and encryption algorithms. Multi-dimensional features are extracted from the associated data, including data format distribution (such as the proportion of text, voice, and numerical data), timestamp sequence (such as the time distribution of data generated in the past 7 days), and sensitive field occurrence frequency (such as the number of occurrences of keywords like "anxiety" and "stress"), and these features are quantified into standardized character labels (such as "text proportion 60%" is converted to "T60", and "average daily generation 50 records" is converted to "D50").

[0258] Based on these labels, a ternary tree is constructed: the root node is the unique identifier of the batch of associated data, each parent node derives three child nodes according to the logic of "feature category - feature value - feature weight" (for example, the child nodes of the "sensitive feature" category include "negative sentiment word frequency", "number of psychological interventions", etc.), and the level of the node corresponds to the degree of influence of the feature on data security (for example, features directly related to privacy correspond to deeper nodes). As in the first ternary tree, each node contains a check code to ensure the accuracy of the data feature mapping.

[0259] The first ternary tree and the second ternary tree are superimposed by aligning the root nodes, so that the nodes of the two trees form overlapping or independent node sets in the same spatial coordinate system. During the superimposition process, overlapping nodes with the same position in the two trees are identified by comparing the character labels and check codes of the nodes, as well as non-overlapping nodes unique to each tree.

[0260] Subsequently, the overlapping nodes are evaluated according to pre-set screening rules: if the semantic association degree of the character labels of the overlapping nodes of the two trees is ≥80% (for example, "K256" and "T60" are both associated with highly sensitive data and are judged to be strongly associated), and the matching degree of the node check codes is ≥95%, then the node is marked as a candidate node; further, the depth of the candidate node is screened, and nodes with a depth ≥3 (to ensure core features) are retained, and the target node is finally determined. The screening rules are solidified through a smart contract, and the association threshold can be dynamically adjusted according to the data security level (for example, the threshold for first-level sensitive data is increased to 90%).

[0261] The character data on the target node contains the fusion information of the encryption algorithm parameters and the data features, and these character data are serialized and spliced from deep to shallow according to the node level (for example, the characters of nodes with a depth of 5 are spliced first), forming the original key material. The material is subjected to multiple rounds of hash operations to generate an encryption key of a fixed length (such as 256 bits), and quantum random numbers are introduced as salt values in the key generation process to ensure that different keys are generated for the same character sequence.

[0262] The encryption key is distributed through a key management system, decrypted only temporarily during encryption operations, and immediately destroyed after use, avoiding long-term residence of the key in memory. Before storage, the associated data is encrypted using the encryption key, and the encrypted ciphertext is stored in distributed nodes according to the data sharding strategy. The head of each shard is attached with a key index and an encryption timestamp to ensure accurate positioning of the corresponding key version during decryption.

[0263] Referring to Figure 2 In another embodiment of the present application, a data processing device of an integrated virtual employee care Saas platform is also provided, comprising:

[0264] The acquisition unit is configured to acquire associated data of an employee through an integrated virtual employee care Saas platform, wherein the associated data comprises historical behavior data, working state data and emotional fluctuation data.

[0265] The tuning unit is configured to tune a preset psychological state recognition algorithm based on the associated data to obtain a tuned psychological state recognition algorithm.

[0266] The evaluation unit is configured to evaluate the associated data based on the tuned psychological state recognition algorithm to generate an employee psychological state evaluation result, and automatically push care scheme information when the employee psychological state evaluation result triggers an early warning threshold.

[0267] The storage unit is configured to store the associated data by using a dynamic encryption algorithm.

[0268] In the embodiment, the specific implementation of each unit in the above device embodiment can refer to the description in the above method embodiment, and will not be repeated here.

[0269] Referring to Figure 3 In the embodiment, a computer device can be a server, and the internal structure thereof can be as shown in Figure 3 The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store corresponding data in the embodiment. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the above method.

[0270] Those skilled in the art can understand Figure 3 that the structure shown in the embodiment is only a block diagram of part of the structure related to the scheme, and does not constitute a limitation on the computer device to which the scheme is applied.

[0271] The embodiment also provides a computer readable storage medium having a computer program stored thereon, and the computer program is executed by the processor to implement the above method. It can be understood that the computer readable storage medium in the embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0272] In summary, the data processing method and device of the integrated virtual employee care Saas platform provided in the embodiments of the present application, comprising: obtaining the associated data of the employee through the integrated virtual employee care Saas platform; the associated data comprises historical behavior data, working state data and emotional fluctuation data; based on the associated data, the preset psychological state recognition algorithm is adjusted to obtain the adjusted psychological state recognition algorithm; based on the adjusted psychological state recognition algorithm, the associated data is evaluated to generate the employee psychological state evaluation result, when the employee psychological state evaluation result triggers the early warning threshold, the care scheme information is automatically pushed; the dynamic encryption algorithm is used to encrypt and store the associated data. In the present application, the multiple dimensions of the associated data of the employee are obtained through the integrated virtual employee care Saas platform, and the psychological state recognition algorithm is adjusted, so that the psychological state evaluation accuracy is improved; the dynamic encryption algorithm is used to encrypt and store the associated data, so that the data security is ensured.

[0273] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium provided by the present application and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM, etc.

[0274] It is to be understood that the terminology "including", "comprising", or any other variation thereof, is intended to cover a non-exclusive inclusion such that process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises a... " does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0275] The above description is merely the preferred embodiments of the present application, and is not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made according to the content of the present application specification and drawings, or directly or indirectly applied to other related technical fields, is also included in the patent protection scope of the present application.

Claims

1. A data processing method for an integrated virtual employee care SaaS platform, characterized in that, Includes the following steps: Employee data is acquired through an integrated virtual employee care SaaS platform; this data includes historical behavioral data, work status data, and emotional fluctuation data. The preset psychological state recognition algorithm is optimized based on the associated data to obtain the optimized psychological state recognition algorithm. The associated data is evaluated based on the optimized psychological state recognition algorithm to generate employee psychological state evaluation results. When the employee psychological state evaluation results trigger the warning threshold, care plan information is automatically pushed. The associated data is encrypted and stored using a dynamic encryption algorithm, including: dynamically updating the basic encryption algorithm based on the associated data; generating a first ternary tree structure based on the algorithm parameters of the updated basic encryption algorithm; the algorithm parameters include key length, number of iteration rounds, group size, and hash function type; generating a second ternary tree structure based on the data characteristics of the associated data; the data characteristics include the proportion of text, voice, and numerical data, timestamp sequences, and frequency of occurrence of sensitive fields; superimposing the first ternary tree structure and the second ternary tree structure, and comparing each overlapping node in turn; selecting target nodes based on the comparison results of each node according to pre-defined rules; generating encryption keys based on the character data on each target node; and encrypting and storing the associated data based on the encryption keys.

2. The data processing method of the integrated virtual employee care SaaS platform according to claim 1, characterized in that, The integrated virtual employee care SaaS platform integrates psychological health support, career planning advice, social interaction services, and life guidance into a single SaaS platform. The SaaS platform receives employee service requests through a unified interface and automatically matches the corresponding functional modules according to the request type.

3. The data processing method of the integrated virtual employee care SaaS platform according to claim 1, characterized in that, The emotional fluctuation data includes the emotional tendency value of text chat records and the intonation feature parameters of voice interaction.

4. The data processing method of the integrated virtual employee care SaaS platform according to claim 1, characterized in that, Based on the associated data, the preset psychological state recognition algorithm is optimized to obtain the optimized psychological state recognition algorithm, including: Personalized strong correlation features are extracted from the associated data; among them, for the same employee in the same scenario, the Pearson correlation coefficient between the negative indicators in the emotional fluctuation data and the task delay rate in the work status data is ≥0.7, and the occurrence is stable at ≥4 times in the past six months, which is marked as a personalized strong correlation feature. A correction factor is constructed based on the aforementioned personalized strong correlation features; The correction factor is embedded into the feature extraction layer of the preset psychological state recognition algorithm. For data that matches personalized strong correlation features, the feature weight in the algorithm is automatically increased, while the weight of general group features that are unrelated to employees is frozen, thus obtaining an optimized psychological state recognition algorithm.

5. The data processing method of the integrated virtual employee care SaaS platform according to claim 4, characterized in that, The formula for calculating the correction factor is as follows: Correction factor = (individual characteristic value - population mean) / population standard deviation × 0.6 + characteristic frequency × 0.

4.

6. The data processing method of the integrated virtual employee care SaaS platform according to claim 1, characterized in that, The method further includes: A points-based incentive mechanism is established based on employees' frequency of service usage and interaction on the platform. Points can be redeemed for professional psychological assessment services or career planning courses, and employees can share points and experiences through social interaction modules.

7. The data processing method of the integrated virtual employee care SaaS platform according to claim 1, characterized in that, The associated data is encrypted and stored using a dynamic encryption algorithm, including: The sensitivity level of the associated data is classified. Based on the hierarchical structure of each associated data, the corresponding splitting rules are determined, and the associated data is split into 512-byte encrypted data blocks based on the splitting rules. A unique verification code is attached to each encrypted data block. The encrypted data blocks are distributed and stored across three physically isolated storage nodes, and the encrypted data blocks stored on any single storage node do not exceed 40% of the total. When accessing the associated data, the following conditions must be met: The IP address of the calling node is within the preset whitelist; Obtain dynamic authorization tokens from 3 storage nodes, with the token generation interval within a preset time. The tokens contain the reassembly algorithm version number of the data blocks. Employees are verified through real-time facial recognition. Once all three conditions are met, the reassembly algorithm is automatically triggered, and the complete data is assembled according to the check code sequence.

8. The data processing method of the integrated virtual employee care SaaS platform according to claim 1, characterized in that, The associated data is encrypted and stored using a dynamic encryption algorithm, including: The data activity index of the associated data is calculated using the following formula: (access frequency within a preset time period × 0.5) + (number of associated alerts triggered × 0.3) + (data update frequency × 0.2). When the data activity index is greater than the first preset value, the AES-256-GCM algorithm is used for encrypted storage, and a quantum-resistant hash lock is added. The data is automatically re-encrypted every preset time period. When the data activity index is less than the second preset value, the core data in the associated data is retained and the original detailed data is deleted, and a lightweight encryption algorithm is used for encrypted storage; the first preset value is greater than the second preset value. When the data activity index is not greater than the first preset value and not less than the second preset value, a preset encryption algorithm is used for encrypted storage.

9. The data processing method of the integrated virtual employee care SaaS platform according to claim 1, characterized in that, The associated data is encrypted and stored using a dynamic encryption algorithm, including: The multi-dimensional features of the associated data are obtained and classified to obtain the first-dimensional features, the second-dimensional features, and the third-dimensional features; Obtain an array of encryption algorithms; wherein the array of encryption algorithms is an array comprising multiple encryption algorithms; Based on the first dimension feature, the encryption algorithm array is deformed to obtain a deformed encryption algorithm array; The target encryption algorithm is obtained by filtering the array of variant encryption algorithms based on the second dimension feature; The target encryption algorithm is optimized based on the third dimension feature to obtain an optimized encryption algorithm; the associated data is then encrypted and stored based on the optimized encryption algorithm.

10. A data processing device for an integrated virtual employee care SaaS platform, characterized in that, include: The acquisition unit is used to acquire relevant data of employees through the integrated virtual employee care SaaS platform; the relevant data includes historical behavior data, work status data, and emotional fluctuation data. The optimization unit is used to optimize the preset psychological state recognition algorithm based on the associated data to obtain the optimized psychological state recognition algorithm. The evaluation unit is used to evaluate the associated data based on the optimized psychological state recognition algorithm, generate employee psychological state evaluation results, and automatically push care plan information when the employee psychological state evaluation results trigger an early warning threshold. A storage unit is used to encrypt and store the associated data using a dynamic encryption algorithm, comprising: dynamically updating a basic encryption algorithm based on the associated data; generating a first ternary tree structure based on the algorithm parameters of the updated basic encryption algorithm; the algorithm parameters including key length, number of iteration rounds, group size, and hash function type; generating a second ternary tree structure based on the data characteristics of the associated data; the data characteristics including the proportion of text, voice, and numerical data, timestamp sequence, and frequency of occurrence of sensitive fields; superimposing the first ternary tree structure and the second ternary tree structure, and comparing each overlapping node in turn; selecting target nodes based on the comparison results of each node according to pre-defined rules; generating an encryption key based on the character data on each target node; and encrypting and storing the associated data based on the encryption key.

Images