Intelligent seal system real-time stamping risk interception method and intelligent seal system

Abstract

This invention proposes a real-time risk interception method and an intelligent seal system for intelligent seal usage. The method includes: a data acquisition step S101, collecting user perception data; a risk factor calculation step S102, calculating a real-time risk index value R based on the perception data; and an interception step S103, where if R ≤ T1, the seal usage operation is allowed; if R ≥ T2, the seal usage operation is intercepted and an alarm is issued; and if T1 ≤ T2 ≤ T2, the user is prompted to perform enhanced verification, where T1 is a first threshold and T2 is a second threshold. The inventive concept of this application lies in combining real-time multimodal perception with a dynamic risk fusion model to achieve comprehensive, proactive security assessment and immediate intervention for a single seal usage behavior, considering factors from "identity," "intent," "document," to "environment," thereby ensuring seal usage security.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and seal security monitoring technology, specifically to a method for real-time seal usage risk interception in an intelligent seal system, an intelligent seal system, and a storage medium. Background Technology

[0002] In the existing technology, the security management of smart seals has mainly developed along two directions: physical control and digital control, and has achieved certain results, but there are still significant and deep-seated security blind spots.

[0003] (1) Near-end security technology of physical stamping control instruments: The solution represented by CN120024135A prevents documents from being switched by taking a picture of the document before stamping and comparing it with the initial image. This technology is essentially a passive, single-point, post-verification mechanism. Its core drawback is:

[0004] The verification dimension is singular: it relies solely on static image comparison of the document's appearance and cannot assess the operator's physiological state, operational intentions, or the immediate environment.

[0005] Unable to prevent "authorized personnel from operating without authorization": under special circumstances, the stamp cannot be recognized by the equipment.

[0006] Lack of verification of the "seal context": The operator claims to be affixing "Contract A," but the actual document is "Contract B." If the two documents have similar layouts, simple image comparison may fail. The system cannot understand the semantic content of the seal.

[0007] (2) Risk warning technology for electronic seals based on historical data: such as CN119359045A, which establishes a statistical model for risk warning by analyzing historical data such as the frequency of use, the text of the event, and the authorized personnel. This type of technology is macroscopic, lagging, and pattern-based.

[0008] It cannot achieve real-time risk control for "single stamp use": its early warning is based on deviations from historical patterns, and the system cannot react immediately to single stamp use behaviors that are high-risk but appear for the first time.

[0009] Reliance on structured text analysis: The analysis of the reasons for using the seal relies on the text on the approval form, which cannot verify whether the verbal statement of the on-site operator is consistent with the approval content and the actual document content, which poses a risk.

[0010] (3) Integrated biometric authentication technology: Some high-end devices have integrated fingerprint and facial recognition modules. This solves the problem of identity binding of "who is using it", but the security model remains at the static authentication level.

[0011] Vulnerable to presentation attacks: Low-end liveness detection can be fooled by high-quality photos, videos, or 3D masks.

[0012] Complete lack of state awareness: Even if the liveness detection passes, it's impossible to know whether the operator is currently nervous or distracted. Assuming that successful authentication equates to safety is a dangerous assumption.

[0013] In summary, the closest existing technology can be considered to be an "intelligent seal control device integrating facial recognition and document image comparison." It combines identity authentication and physical document anti-tampering, forming the mainstream solution for current intelligent seal security. However, this solution still has a narrow security boundary. Its core problem is that it simplifies complex seal security to "the right person" stamping "the right document," completely ignoring the three crucial dynamic dimensions of "the right intention," "the right state," and "the right environment." The system is like a "blind security guard," only able to touch documents and recognize identification, completely unaware of the dynamic risks at play, creating a significant security gap. Summary of the Invention

[0014] In view of one or more technical defects in the prior art, the present invention proposes the following technical solution.

[0015] A method for real-time risk interception of seal usage in an intelligent seal system, wherein the intelligent seal system includes at least a camera, a microphone, a document scanner, and an environmental sensor, the method comprising:

[0016] The data collection steps include collecting user perception data, including: receiving the user's intention to use the seal via microphone or virtual keyboard when the user places a document in the smart seal system in preparation for use; the intention to use the seal is declared by the user; and collecting data via the camera. The user's facial video during time period T is captured via the microphone. The ambient audio during time period T is used to scan the document in high definition using the document scanner to obtain an electronic document. The current light intensity value and location data are read through the environmental sensor. A request is sent to the background service to obtain the most recently approved electronic process data associated with the current user and seal.

[0017] The risk factor calculation step involves calculating the real-time risk index value R based on the perceived data.

[0018] The interception step is as follows: if R ≤ T1, the current stamping operation is allowed; if R ≥ T2, the current stamping operation is intercepted and an alarm is issued; if T1... R If T2 is selected, the user is prompted to perform enhanced verification, where T1 is the first threshold and T2 is the second threshold.

[0019] Furthermore, the enhanced verification operation is as follows: the smart seal system generates and displays a temporary, one-time random verification QR code; the user scans the QR code using their pre-bound, real-name authenticated backup device or manually enters the verification code corresponding to the QR code; the smart seal system automatically packages all the sensor data collected this time into an immutable "evidence package" and generates a hash value for on-chain storage; if the user completes the QR code verification within a limited time, it is determined that "secondary authentication is passed", and the current seal operation is allowed, and the "evidence package" is marked as an associated file.

[0020] Furthermore, the operation of the risk factor calculation step is as follows: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The facial videos within time period T are processed sequentially, including face recognition, liveness detection, and parallel calculation of micro-expression tension. and heart rate variability abnormality Based on the above and The operator state risk sub-score is obtained by fusion. ;right Speech activity detection is performed on ambient audio within a time period T, separating the user statement and ambient sound. The statement is then used for text transcription. The ambient sound component is used to identify keywords. Then, the context consistency risk sub-score is calculated. Perform OCR and layout analysis on the scanned electronic documents to extract key entities. Simultaneously, it parses the electronic process data returned from the backend and extracts key entities. Fuzzy matching is performed between the two to calculate the file consistency risk sub-score. Based on time and location data, the number of people detected in the video, light intensity values, and volume data, calculate the environmental safety risk sub-score. The intelligent seal system loads the corresponding weight vector from a local or cloud-based security policy library based on the seal ID used in this application. The intelligent seal system, through Real-time calculated risk index ;in, .

[0021] Furthermore, the calculation of the operator state risk sub-score The operation is as follows: Multimodal liveness detection is used to calculate the liveness detection results. ,in, This indicates that the liveness detection has passed. Indicates failure or detection of an attack; calculates emotional tension. Extracting the intensity of motor units associated with tension and fear from facial key point sequences.

[0022]

[0023] The number of selected tension-related motor units;

[0024] : No. Importance coefficient of each action unit;

[0025] : in the time window within, no. Average intensity of each action unit, range ;

[0026] Calculate the degree of abnormality of heart rate variability Heart rate signals are estimated from facial videos using rPPG technology, and their short-time complexity is calculated.

[0027]

[0028] Estimated heart rate variability index for the current time period.

[0029] , The mean and standard deviation of the user's HRV in a resting state are compared to the baseline. : Adjustment coefficient, used to control abnormal sensitivity.

[0030] calculate :

[0031]

[0032] in, : is the fusion coefficient, and ,in It can be set to a larger value (such as 0.5) to reflect the veto power of liveness detection.

[0033] Furthermore, calculate the context consistency risk sub-score. The operation is as follows: Obtain user statements input via microphone or touchscreen. And identify the set of keywords in the environmental audio. ,

[0034] in, ;

[0035] ;

[0036] This represents a text embedding function used to convert text into a semantic vector;

[0037] Represents the calculation declaration text With all environmental audio keywords The maximum cosine similarity between the statements is used; a high similarity is determined by the presence of highly relevant content in the ambient audio. Low similarity; if the ambient sound content is completely unrelated or contains contradictory keywords, the similarity is low. high.

[0038] Furthermore, calculate the document consistency risk sub-score. The operation involves performing OCR on the scanned electronic document to extract the key entity set. (e.g., contract number, names of both parties, amount, date), and simultaneously parse the electronic process data returned from the backend and extract key entities.

[0039] in, ;

[0040] in, : indicates the intersection of sets based on fuzzy matching. For example, after calculating the string similarity (such as Levenshtein distance) between "Client: XX Technology Company" recognized by OCR in a document and "Client: XX Technology Co., Ltd." in an approval form, if the similarity is greater than a threshold, then... If the match is successful, then it is considered a successful match.

[0041] Furthermore, calculate the environmental safety risk sub-score. The operation is as follows:

[0042]

[0043] in, Indicates time risk, If it is within the preset working time period, otherwise ;

[0044] Indicates location risk. When the GPS / Bluetooth beacon is within the authorized geofence, otherwise ;

[0045] To indicate the risk of large gatherings, the system uses cameras to detect the total number of faces in the feed. and the number of authorized faces identified , ;

[0046] This indicates the threshold for the number of people at risk, excluding the operator themselves. In addition to those already included, the risk increases with each additional unidentified person. until the upper limit of 1 is reached;

[0047] Indicates the risk of sound and light environment, and the overall ambient light intensity. and average ambient volume , ,in, It is a step function. Returns 1 if the condition is met, otherwise returns 0. , These represent the preset minimum and maximum decibel thresholds, respectively.

[0048] Furthermore, for each user's operation, the intelligent seal system generates a structured digital seal usage profile, which includes: a timestamp, user ID, seal ID, and a final risk index. The system records the number of sub-scores, decision results, response actions, and the hash values ​​of the associated "evidence packages," and encrypts the digital archives of the seals used before synchronizing them to the cloud audit center to form a complete and traceable record of the seals used.

[0049] The present invention also proposes an intelligent seal system, which includes at least a processor, a memory, a camera, a microphone, a document scanner, and an environmental sensor. The processor is connected to the memory via a bus, and the memory stores a computer program that, when executed by the processor, implements any of the methods described above.

[0050] The present invention also proposes a computer-readable storage medium storing computer program code, which, when executed by a computer, performs any of the methods described above.

[0051] The technical advantages of this invention are as follows: This invention provides a real-time risk interception method for intelligent seal systems, an intelligent seal system, and a storage medium. The method includes: a data acquisition step S101, acquiring user perception data, including: when a user places a document in the intelligent seal system to prepare for seal acquisition, receiving the seal acquisition intention via a microphone or virtual keyboard, wherein the seal acquisition intention is declared by the user; and acquiring data via the camera. The user's facial video during time period T is captured via the microphone. The environmental audio during time period T is used to scan the document in high definition using the document scanner to obtain an electronic document. The current light intensity value and location data are read by the environmental sensor. A request is sent to the background service to obtain the most recently approved electronic process data associated with the current user and seal. In the risk factor calculation step S102, the real-time risk index value R is calculated based on the perceived data. In the interception step S103, if R≤T1, the current seal application is allowed; if R≥T2, the current seal application is intercepted and an alarm is issued. If T1 R If T2 is reached, the user is prompted to perform enhanced verification, where T1 is the first threshold and T2 is the second threshold. The inventive concept of this application lies in combining real-time multimodal perception with a dynamic risk fusion model to achieve comprehensive, proactive security assessment and immediate intervention for a single seal-using behavior, considering factors from "identity," "intent," "document," to "environment," thereby ensuring seal security. Attached Figure Description

[0052] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0053] Figure 1 This is a flowchart of a real-time seal usage risk interception method for an intelligent seal system according to an embodiment of the present invention.

[0054] Figure 2 This is a structural diagram of an intelligent seal system according to an embodiment of the present invention. Detailed Implementation

[0055] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0056] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0057] Figure 1 This invention discloses a method for real-time seal usage risk interception using an intelligent seal system. The intelligent seal system includes at least a camera, a microphone, a document scanner, and an environmental sensor. The method includes:

[0058] Step S101 involves collecting user perception data, including: receiving the user's intention to use the seal via a microphone or virtual keyboard when the user places a document in the smart seal system in preparation for use; the intention to use the seal is declared by the user; and collecting data via the camera. The user's facial video during time period T is captured via the microphone. The ambient audio during time period T is used to scan the document in high definition using the document scanner to obtain an electronic document. The current light intensity value and location data are read through the environmental sensor. A request is sent to the background service to obtain the most recently approved electronic process data associated with the current user and seal.

[0059] Risk factor calculation step S102: Calculate the real-time risk index value R based on the perceived data;

[0060] In interception step S103, if R ≤ T1, the current stamping operation is allowed; if R ≥ T2, the current stamping operation is intercepted and an alarm is issued; if T1... R If T2 is selected, the user is prompted to perform enhanced verification, where T1 is the first threshold and T2 is the second threshold.

[0061] The inventive concept of this application lies in combining real-time multimodal perception with a dynamic risk fusion model to achieve comprehensive, proactive security assessment and immediate intervention for a single seal-using behavior, considering factors such as "identity," "intention," "document," and "environment." This application can estimate the operator's physiological indicators (such as heart rate variability (HRV)) and micro-expression changes in real time and imperceptibly, based on a regular camera, without relying on invasive sensors (such as wearable heart rate monitors). It quantifies the operator's "tension," "stress level," or "level of consciousness." After biometric verification, it further determines whether the operation was performed autonomously, consciously, and in a normal mental state, rather than as an involuntary act under duress, extreme fatigue, or mental confusion. Simultaneously, it can capture the user's verbal or written statement of purpose for using the seal (declaration of intent) at the moment the user initiates the seal-using action, extract the substantive content of the document to be stamped (document content) through optical character recognition, and link it to the electronic process data of the back-end approval system (approval basis). Furthermore, it utilizes natural language processing technology to quickly calculate the semantic consistency score among these three elements to detect risks such as "dual contracts," "substitution," or "operator misunderstanding." This application defines and quantifies the "environmental risk" of a single seal-using operation. This includes not only static "time-location" rules (such as non-working hours, non-office locations) but also dynamic factors: such as simultaneously detecting whether there are multiple onlookers not on the authorized list at the seal-using site (through facial detection and comparison); whether the ambient sound is abnormally noisy or there are conflicting conversations; whether the ambient lighting is abnormally dim, etc. A multi-sensor data fusion instantaneous environmental risk assessment model needs to be established. This application utilizes three dimensions of perceptual data (video stream, audio stream, OCR text (Optical Character Recognition), and sensor data), which are heterogeneous in time and format. On resource-constrained edge devices (such as smart seal terminals), it achieves precise time synchronization and efficient feature extraction of these data, and integrates them into a unified, rapidly computable real-time seal-using risk index mathematical model. This model needs to support dynamic adjustment of the weights of each dimension according to different business types (such as official seals, contract seals, and financial seals), thereby ensuring seal-using security. This is the core inventive concept of this application.

[0062] In one embodiment, the system will With two preset thresholds (e.g., 0.3) (e.g., 0.7) for comparison: Decision A ( ): Determined as low risk. The system green light illuminates, indicating "Verification passed," and immediately drives the actuator to complete the physical stamping or electronic signature; Decision B ( The system has been flagged as medium risk. A yellow warning light is flashing, indicating "Enhanced verification required."

[0063] The process transitions to the enhanced verification sub-process: The system generates a temporary, one-time random verification code, which is displayed on the device screen; the user is required to scan the QR code using their pre-bound, real-name authenticated backup device (such as a mobile APP) or manually enter the verification code; at the same time, the system automatically packages all the original perception data collected this time (encrypted video, audio clips, risk logs) into an immutable "evidence package" and generates a hash value for on-chain storage; if the user completes the QR code verification within the specified time, it is determined that "secondary authentication is successful", the system stamps the document, and marks the "evidence package" as an associated file.

[0064] Decision C ( The system is deemed high-risk. A red light illuminates continuously, and an alarm sound is emitted, indicating "Security risk, operation blocked." The system immediately: physically locks the stamping mechanism to prevent any stamping action; pushes the "evidence package" and high-risk alarm through an encrypted channel in real time to multiple pre-set security personnel; and records the security event log locally on the device.

[0065] This application addresses the problem of a tiered, flexible response mechanism based on risk levels: how to avoid abrupt, one-size-fits-all approach that disrupts normal business operations when a system detects a potential risk. It requires designing a tiered response strategy that quantifies risks into different levels and triggers corresponding response actions: from "recording enhanced evidence" to "initiating secondary confirmation," and finally to "physical hard blocking and alarms." Ensuring that the security response is both powerful and precise, and that all decision-making processes and original sensing data are encrypted and archived to form an auditable "stamped digital twin record," is another key inventive aspect of this application.

[0066] In one embodiment, the operation of the risk factor calculation step S102 is as follows: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The facial videos within time period T are processed sequentially, including face recognition, liveness detection, and parallel calculation of micro-expression tension. and heart rate variability abnormality Based on the above and The operator state risk sub-score is obtained by fusion. ;right Speech activity detection is performed on ambient audio within a time period T, separating the user statement and ambient sound. The statement is then used for text transcription. The ambient sound component is used to identify keywords. Then, the context consistency risk sub-score is calculated. Perform OCR and layout analysis on the scanned electronic documents to extract key entities. Simultaneously, it parses the electronic process data returned from the backend and extracts key entities. Fuzzy matching is performed between the two to calculate the file consistency risk sub-score. Based on time and location data, the number of people detected in the video, light intensity values, and volume data, calculate the environmental safety risk sub-score. The intelligent seal system loads the corresponding weight vector from a local or cloud-based security policy library based on the seal ID used in this application. The intelligent seal system, through Real-time calculated risk index ;in, .

[0067] The core of this application is to construct a closed-loop system of "synchronous perception - feature extraction - fusion evaluation - hierarchical response", and proposes a core mathematical model. This represents the real-time stamp usage risk index, with a value range of [value range missing]. The higher the value, the higher the risk of this stamping operation.

[0068] Operator state risk sub-score. Calculated based on the operator's biometrics and physiological signals.

[0069] Context consistency risk sub-score. Calculated based on the degree of matching between the user's stated intent and the environmental information.

[0070] Document consistency risk sub-score. Calculated based on the degree of matching between document content and approval basis.

[0071] Environmental safety risk sub-score. Based on spatiotemporal rules and dynamic environmental perception calculation.

[0072] These represent the dynamic weights for the four dimensions mentioned above. The weight values ​​are loaded from a predefined configuration library based on the "seal type" used in this application. For example, for "legal person's name seal," The weight is relatively high; regarding the "contract seal", and It has a high weight.

[0073] This application solves the problem of achieving precise time synchronization and efficient feature extraction of data such as video, audio, scanned documents, and light intensity on resource-constrained edge devices (such as smart seal terminals) by constructing the aforementioned calculation model of R. It integrates these data into a unified and rapidly calculable real-time seal risk index mathematical model. This model needs to support dynamic adjustment of the weights of each dimension according to different business types (such as official seals, contract seals, and financial seals) to ensure seal security. This is one of the important inventive points of this application.

[0074] In one embodiment, the calculation of the operator state risk sub-score The operation is as follows: Multimodal liveness detection is used to calculate the liveness detection results. ,in, This indicates that the liveness detection has passed. Indicates failure or detection of an attack; calculates emotional tension. Extracting the intensity of motor units associated with tension and fear from facial key point sequences.

[0075]

[0076] The number of selected tension-related motor units;

[0077] : No. Importance coefficient of each action unit;

[0078] : in the time window within, no. Average intensity of each action unit, range ;

[0079] Calculate the degree of abnormality of heart rate variability Heart rate signals are estimated from facial videos using rPPG technology, and their short-time complexity is calculated.

[0080]

[0081] Estimated heart rate variability index for the current time period.

[0082] , The mean and standard deviation of the user's HRV in a resting state are compared to the baseline. : Adjustment coefficient, used to control abnormal sensitivity.

[0083] calculate :

[0084]

[0085] in, : is the fusion coefficient, and ,in It can be set to a larger value (such as 0.5) to reflect the veto power of liveness detection.

[0086] The calculation proposed in this embodiment The specific method can solve the problem of real-time identification of "identity misuse" and "non-voluntary operation". This allows the present application to estimate the operator's physiological indicators (such as heart rate variability HRV) and micro-expression changes in real time and imperceptibly based on ordinary cameras without relying on invasive sensors (such as wearable heart rate belts). It quantifies the operator's "tension", "stress level" or "consciousness". After biometric identification is passed, it can further determine whether the operation is in a voluntary, clear and normal mental state, rather than an involuntary behavior under coercion, extreme fatigue or mental confusion. This improves the security of the seal. This is another important inventive point of the present application.

[0087] In one embodiment, the context consistency risk sub-score is calculated. The operation is as follows: Obtain user statements input via microphone or touchscreen. And identify the set of keywords in the environmental audio. ,

[0088] in, ;

[0089] ;

[0090] This represents a text embedding function used to convert text into a semantic vector;

[0091] Represents the calculation declaration text With all environmental audio keywords The maximum cosine similarity between the statements is used; a high similarity is determined by the presence of highly relevant content in the ambient audio. Low similarity; if the ambient sound content is completely unrelated or contains contradictory keywords, the similarity is low. high.

[0092] In one embodiment, the file consistency risk sub-score is calculated. The operation involves performing OCR on the scanned electronic document to extract the key entity set. (e.g., contract number, names of both parties, amount, date), and simultaneously parse the electronic process data returned from the backend and extract key entities.

[0093] in, ;

[0094] in, : indicates the intersection of sets based on fuzzy matching. For example, after calculating the string similarity (such as Levenshtein distance) between "Client: XX Technology Company" recognized by OCR in a document and "Client: XX Technology Co., Ltd." in an approval form, if the similarity is greater than a threshold, then... If the match is successful, then it is considered a successful match.

[0095] This application is calculated , This application addresses the issue of real-time cross-verification of the "declared intent," "document content," and "approval basis." The real-time process designed in this application simultaneously captures the user's verbal or written declaration of purpose (declared intent) the moment the user initiates the affixing of the seal. It then extracts the substantive content of the document to be stamped (document content) through optical character recognition and links it to the electronic workflow data (approval basis) of the backend approval system. Furthermore, another key inventive concept of this application is how to quickly calculate the semantic consistency score among these three elements using natural language processing technology to detect risks such as "dual contracts," "substitution," or "operator misunderstanding," thereby further enhancing seal security.

[0096] In one embodiment, environmental safety risk sub-scores are calculated. The operation is as follows:

[0097]

[0098] in, Indicates time risk, If it is within the preset working time period, otherwise ;

[0099] Indicates location risk. When the GPS / Bluetooth beacon is within the authorized geofence, otherwise ;

[0100] To indicate the risk of large gatherings, the system uses cameras to detect the total number of faces in the feed. and the number of authorized faces identified , ;

[0101] This indicates the threshold for the number of people at risk, excluding the operator themselves. In addition to those already included, the risk increases with each additional unidentified person. until the upper limit of 1 is reached;

[0102] Indicates the risk of sound and light environment, and the overall ambient light intensity. and average ambient volume , ,in, It is a step function. Returns 1 if the condition is met, otherwise returns 0. , These represent the preset minimum and maximum decibel thresholds, respectively.

[0103] This application is calculated This application addresses the problem of real-time perception and quantitative assessment of dynamic environmental security risks by defining and quantifying the "environmental risk" of a single stamping operation. This includes not only static "time-location" rules (such as non-working hours, non-office locations) but also dynamic factors: such as simultaneously detecting the presence of multiple unauthorized bystanders at the stamping location (through facial detection and comparison); whether the ambient sound is abnormally noisy or involves conflicting conversations; and whether the ambient lighting is abnormally dim. The need to establish a multi-sensor data fusion model for instantaneous environmental risk assessment is another key inventive concept of this application.

[0104] In one embodiment, each time the user performs an operation, the smart seal system generates a structured digital seal usage profile, which includes: a timestamp, user ID, seal ID, and a final risk index. The system records each sub-score, decision result, response action, and the hash value of the associated "evidence package," and encrypts the digital archive of the seal usage before synchronizing it to the cloud audit center to form a complete and traceable seal usage record, ensuring seal usage security.

[0105] Figure 2 The present invention illustrates an intelligent seal system, which includes at least a processor, a memory, a camera, a microphone, a document scanner, and an environmental sensor. The processor is connected to the memory via a bus, and the memory stores a computer program that, when executed by the processor, implements any of the methods described above.

[0106] One embodiment of the present invention provides a computer storage medium storing a computer program. When the computer program on the computer storage medium is executed by a processor, the above-described method is implemented. The computer storage medium may be a hard disk, DVD, CD, flash memory, or other storage device.

[0107] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0108] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the apparatus described in various embodiments or some parts of the embodiments of this application.

[0109] Finally, it should be noted that the above embodiments are for illustration only and not for limiting the technical solutions of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention without departing from the spirit and scope of the present invention. Any modifications or partial substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for real-time seal usage risk interception in an intelligent seal system, characterized in that, The intelligent seal system includes at least a camera, a microphone, a document scanner, and an environmental sensor, and the method includes: The data collection steps include collecting user perception data, including: receiving the user's intention to use the seal via microphone or virtual keyboard when the user places a document in the smart seal system in preparation for use; the intention to use the seal is declared by the user; and collecting data via the camera. The user's facial video during time period T is captured via the microphone. The ambient audio during time period T is used to scan the document in high definition using the document scanner to obtain an electronic document. The current light intensity value and location data are read through the environmental sensor. A request is sent to the background service to obtain the most recently approved electronic process data associated with the current user and seal. The risk factor calculation step involves calculating a real-time risk index value R based on the perceived data; the risk factor calculation step specifically includes: […]. The facial videos within time period T are processed sequentially, including face recognition, liveness detection, and parallel calculation of micro-expression tension. and heart rate variability abnormality Based on the above and The operator state risk sub-score is obtained by fusion. The calculation of the operator state risk sub-score The operation is as follows: Multimodal liveness detection is used to calculate the liveness detection results. ,in, This indicates that the liveness detection has passed. Indicates failure or detection of an attack; calculates emotional tension. Extracting the intensity of motor units associated with tension and fear from facial key point sequences. ； The number of selected tension-related motor units; : No. Importance coefficient of each action unit; : in the time window within, no. Average intensity of each action unit, range ; Calculate the degree of abnormality of heart rate variability Heart rate signals are estimated from facial videos using rPPG technology, and their short-time complexity is calculated. ； Estimated heart rate variability index for the current time period. , The mean and standard deviation of the user's HRV in a resting state are compared to the baseline. : Adjustment coefficient, used to control abnormal sensitivity. calculate : ； in, : is the fusion coefficient, and ; The interception step is as follows: if R ≤ T1, the current stamping operation is allowed; if R ≥ T2, the current stamping operation is intercepted and an alarm is issued; if T1... R If T2 is selected, the user is prompted to perform enhanced verification, where T1 is the first threshold and T2 is the second threshold.

2. The method according to claim 1, characterized in that, The enhanced verification operation is as follows: the smart seal system generates a temporary, one-time random verification QR code and displays it. The user scans the QR code using their pre-bound, real-name authenticated backup device or manually enters the verification code corresponding to the QR code. The smart seal system automatically packages all the perception data collected this time into an immutable "evidence package" and generates a hash value for on-chain storage. If the user completes the QR code verification within the specified time, it will be determined as "secondary authentication passed", and the stamping operation will be allowed, and the "evidence package" will be marked as an associated file.

3. The method according to claim 2, characterized in that, The operation of the risk factor calculation step also includes: […]. Speech activity detection is performed on ambient audio within a time period T, separating the user statement and ambient sound. The statement is then used for text transcription. The ambient sound component is used to identify keywords. Then, the context consistency risk sub-score is calculated. Perform OCR and layout analysis on the scanned electronic documents to extract key entities. Simultaneously, it parses the electronic process data returned from the backend and extracts key entities. Fuzzy matching is performed between the two to calculate the file consistency risk sub-score. Based on time and location data, the number of people detected in the video, light intensity values, and volume data, calculate the environmental safety risk sub-score. The intelligent seal system loads the corresponding weight vector from a local or cloud-based security policy library based on the seal ID used in this application. The intelligent seal system, through Real-time calculated risk index ;in, 。 4. The method according to claim 3, characterized in that, Calculate the context consistency risk sub-score The operation is as follows: Obtain user statements input via microphone or touchscreen. And identify the set of keywords in the environmental audio. , in, ; ； This represents a text embedding function used to convert text into a semantic vector; Represents the calculation declaration text With all environmental audio keywords The maximum cosine similarity between the statements is used; a high similarity is determined by the presence of highly relevant content in the ambient audio. Low similarity; if the ambient sound content is completely unrelated or contains contradictory keywords, the similarity is low. high.

5. The method according to claim 4, characterized in that, Calculate the file consistency risk sub-score The operation involves performing OCR on the scanned electronic document to extract the key entity set. Simultaneously, it parses the electronic process data returned from the backend and extracts key entities. ； in, ; in, : Indicates the intersection of sets based on fuzzy matching; if the similarity is greater than a threshold... If the match is successful, then it is considered a successful match.

6. The method according to claim 5, characterized in that, Calculate environmental safety risk sub-scores The operation is as follows: ； in, Indicates time risk, If it is within the preset working time period, otherwise ; Indicates location risk. When the GPS / Bluetooth beacon is within the authorized geofence, otherwise ; To indicate the risk of large gatherings, the system uses cameras to detect the total number of faces in the image. and the number of authorized faces identified , ； This represents the risk threshold, indicating that for each additional unauthorized person besides the operator, the cumulative risk increases. until the upper limit of 1 is reached; Indicates the risk of sound and light environment, and the overall ambient light intensity. and average ambient volume , ,in, It is a step function. Returns 1 if the condition is met, otherwise returns 0. , These represent the preset minimum and maximum decibel thresholds, respectively.

7. The method according to claim 6, characterized in that, Each time a user performs an operation, the intelligent seal system generates a structured digital seal usage file, which includes: timestamp, user ID, seal ID, and final risk index. The system records the use of seals, including sub-scores, decision results, response actions, and the hash values ​​of associated "evidence packages." These seals are then encrypted and synchronized to the cloud audit center to form a complete and traceable seal usage record.

8. An intelligent seal system, characterized in that, The intelligent seal system includes at least a processor, a memory, a camera, a microphone, a document scanner, and an environmental sensor. The processor is connected to the memory via a bus, and the memory stores a computer program. When the computer program is executed by the processor, it implements the method described in any one of claims 1-7.

Images