Intelligent security check decision-making system based on risk management

Abstract

The invention discloses an intelligent security check decision-making system based on risk management, and relates to the technical field of security check risk management, and the system comprises a basic building module, a model training module, a fusion updating module, and an identification decision-making module. A decentralized infrastructure is constructed through a customized interface network, security check data are verified and stored among nodes through a consensus mechanism, authority and access management is realized in cooperation with an intelligent contract, edge equipment is connected with an X-ray machine to collect security check data, multiple rounds of neural network training and an anomaly detection model are supported, the threat recognition capability is improved, and the security check efficiency is improved. Multi-node model parameter fusion, dynamic adjustment of fusion weight, generation of a global model and distribution to edge equipment for local reasoning, real-time loading of a latest model, automatic analysis of risk data, generation of a risk score and a grade label, calling of an intelligent contract storage result, and automatic triggering of an exception handling measure according to a set decision rule are realized. And the automation, the safety and the real-time reaction capability of the security check process are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of security inspection risk management, and in particular to an intelligent security inspection decision-making system based on risk management. Background Art

[0002] With the advancement of technology, modern security inspection systems are becoming increasingly intelligent and information-based. Traditional security inspection processes rely primarily on manual operations and single-device inspections, resulting in cumbersome processes, low efficiency, and susceptibility to human intervention and operational errors. In recent years, some research has introduced intelligent models based on image recognition and behavioral analysis to improve the accuracy of threat detection. However, these often rely on centralized systems, posing security risks in data transmission and processing, and making model updates and integration difficult. Furthermore, traditional security inspection systems lack a sound trust mechanism, making it difficult to achieve data traceability and full-process oversight, and therefore unable to adequately address increasingly complex security threats.

[0003] Existing technologies have the following shortcomings: Existing security inspection systems rely on centralized data storage and management, resulting in insufficient data security and vulnerability to tampering or leakage. Existing models are slow to update, making it difficult to quickly integrate data from multiple sources for threat identification, resulting in insufficient model adaptability and reduced detection effectiveness. Information silos are prevalent in existing systems, making it difficult to integrate and share data between different security checkpoints, impacting overall security assessments. Traditional security inspection processes rely heavily on manual operations, have low automation levels, and are limited in efficiency, making it difficult to meet the needs of efficient access in large-scale traffic flows. The lack of a robust permission control mechanism increases the risk of permission and data abuse, resulting in low model recognition rates and a high risk of false positives and missed detections, impacting security effectiveness.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent security inspection decision-making system based on risk management to solve the problems in the above-mentioned background technology.

[0006] In order to achieve the above-mentioned purpose, the present invention provides the following technical solutions: a smart security inspection decision-making system based on risk management, comprising a basic construction module, a model training module, a fusion update module and an identification decision module;

[0007] Basic building module: Deploy a customized interface network and define alliance chain nodes. Use the consensus mechanism to participate in security data verification and decentralized storage. Implement permission control and access management through smart contracts. Install edge devices at each node to connect to X-ray machines to collect security data.

[0008] Model training module: This module loads security inspection data through edge devices and conducts multiple rounds of training. It uses neural networks to analyze images or behavioral sequences and combines them with anomaly detection models to identify potential threats.

[0009] Fusion update module: Receives multiple model parameters from each node, fuses the models based on a dynamic fusion weight algorithm, triggers the calculation of the fused global model parameters, generates a new model version, and sends it to the edge node. After receiving it, the edge device automatically loads it into the local inference engine for application;

[0010] Identification and decision-making module: loads the latest global security inspection model, receives security inspection data and performs risk identification, generates and packages risk scores and level labels, calls smart contracts, writes the packaged results to the consortium chain, sets decision-making rules, and automatically triggers corresponding security inspection measures based on risk scores and levels;

[0011] Preferably, a customized interface network is deployed, and security checkpoints, management agencies, and regulatory units are defined as alliance chain nodes and their respective responsibilities are assigned. A consensus mechanism is deployed, and each alliance chain node participates in security check data verification according to the consensus mechanism, and saves a full copy of the customized interface network. A smart contract framework is established, and permission control of the security check process is realized through the smart contract of the customized interface network. The permission roles are defined, and access control based on the identity of the security check permission role is established. Edge devices are installed on each alliance chain node, and the X-ray machine is connected to collect security check data. The AI model and training environment are installed, TLS / SSL is set up, and a unified key management system is configured.

[0012] Preferably, the respective responsibilities of security checkpoints, management agencies, and regulatory units are allocated as follows: the responsibilities of security checkpoint nodes include collecting security data, executing risk assessment models, uploading security data, and calling on-chain smart contracts; the responsibilities of management agency nodes include monitoring operating status, approving authority, reviewing on-chain security events, and archiving audit records; the responsibilities of regulatory unit nodes include participating in model training, parameter fusion, security data analysis, and saving a full copy of the customized interface network for each alliance chain node.

[0013] Preferably, the authority roles specifically include: security personnel, managers, management agencies and AI model managers, among which security personnel can only upload security data and provide execution suggestions based on AI decisions; managers have the authority to view and verify all security data; management agencies are responsible for monitoring the flow of security data; AI model managers are responsible for adjusting and optimizing the algorithms of AI models and defining new security decision logic.

[0014] Preferably, the smart contract framework includes: defining a security inspection data contract to automatically record and verify the upload, modification, and query operations of security inspection data; defining an AI decision-making contract to automatically call the AI model to analyze the security inspection data after a security inspection point completes security inspection data collection, and generate decisions based on preset rules; and defining a compliance contract to verify whether the security inspection process meets compliance requirements.

[0015] Preferably, the initial parameters of the global security inspection model are generated, initialized with random numbers, security inspection data are loaded through edge devices, and multiple rounds of training of the global security inspection model are performed. The model parameters are encrypted by a homomorphic encryption algorithm and uploaded to the aggregation node on the alliance chain. A neural network is selected to perform sequence analysis of images or behaviors of security inspection data, an anomaly detection model is used to identify potential threats, security inspection data is incrementally trained, transfer learning is used to reduce training time, an early stopping mechanism is set, all updateable parameters in the model are extracted, asymmetric encryption technology is used to encrypt the model parameters during transmission, and the data are uploaded to the aggregation node on the alliance chain through a secure communication line.

[0016] Preferably, the model parameters from each consortium chain node are received through the smart contract on the consortium chain, and the model parameter fusion algorithm is defined in the consortium chain smart contract. The specific formula is:

[0017]

[0018] in, represents the dynamic fusion weight of the i-th alliance chain node in the t-th round, α represents the influence of the model performance on the weight, It represents the model performance index of the current alliance chain node on the current round of verification set. β represents the influence of the stability and credibility of the alliance chain node on the weight. It represents the evaluation score of the stability and credibility of the alliance chain node, and γ represents the impact of the hardware environment of the alliance chain node on the weight. The indicator representing the hardware environment where the alliance chain node is located triggers the smart contract call, calculates the integrated global model parameters, and generates a new global model version. The alliance chain actively sends the latest global model parameters to the edge node through a secure channel. After the edge device receives the model parameters, it automatically loads them into the local inference engine.

[0019] Preferably, the latest global security inspection model is loaded through the edge device, security inspection data is received and risk identification is performed, a risk score and risk level label of 0-100 is generated, the risk score, risk level label, passenger identity ID and timestamp are packaged, the smart contract is called through a secure channel, the packaged result is written to the alliance chain to form a traceable certificate, and decision-making rules are set in the smart contract, including: if the risk score is >80, "secondary security inspection" or "manual verification" is triggered; if the risk level is "high risk", manual intervention is automatically prompted or passage is denied; otherwise, normal passage is allowed.

[0020] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0021] 1. By introducing advanced AI intelligent analysis technology, the automated analysis and decision-making process of security inspection data has been achieved, significantly improving the objectivity and scientific nature of security detection. A deep learning-based intelligent model is used to quickly and comprehensively analyze and score collected security inspection data, ensuring the scientificity and objectivity of each judgment. This system enables structured management of multi-source and multi-channel data, effectively integrating multi-dimensional information from different security inspection points, equipment, and scenarios to provide a global perspective on the overall situation. Automated analysis supports real-time early warning and risk assessment, promptly identifying potential threats and improving emergency response capabilities. The automated decision-making process significantly reduces human bias, alleviates staff workload, and mitigates safety hazards caused by human error. Data structuring and integrated management promote the digitalization and intelligence of security inspection processes, continuously improve the system's security capabilities, free up human resources, and enable security personnel to focus on handling highly complex or emergency incidents. This not only ensures the objectivity and accuracy of security assessments, but also realizes the digital transformation of security management, laying a solid foundation for the development of smart security inspection and promoting the industry's upgrade from experience-based to data-driven.

[0022] 2. Through intelligent analysis and comprehensive evaluation, we no longer rely on personal experience, greatly reducing the deviation of human judgment and ensuring that each safety judgment is based on scientific data and algorithm models, thereby improving the accuracy of overall detection. The automated operation of the system can reduce dependence on staff, reduce labor intensity, and relieve staff pressure, allowing operators to focus more on processing high-value or complex scenarios, effectively alleviating the problem of tight staffing. By adopting a digital and structured management concept, various types of heterogeneous business data can be seamlessly connected and integrated, supporting multi-channel and multi-process data integration to form a complete security information chain. Combined with customized interfaces and smart contract technology, it ensures the security, immutability and traceability of data throughout the entire process, thereby improving the transparency and credibility of management. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0024] Figure 1 This is a method flow chart of the intelligent security inspection decision-making system based on risk management of the present invention.

[0025] Figure 2 This is a module diagram of the risk management-based intelligent security inspection decision-making system of the present invention. DETAILED DESCRIPTION

[0026] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0027] Example 1

[0028] The present invention provides Figure 2 The risk management-based intelligent security inspection decision-making system shown in the figure includes a basic construction module, a model training module, a fusion update module, and an identification and decision-making module;

[0029] Basic building module: Deploy a customized interface network and define alliance chain nodes. Use the consensus mechanism to participate in security data verification and decentralized storage. Implement permission control and access management through smart contracts. Install edge devices at each node to connect to X-ray machines to collect security data.

[0030] Deploy a customized interface network, define security checkpoints, management agencies, and regulatory units as alliance chain nodes and assign their respective responsibilities, deploy a consensus mechanism, and each alliance chain node participates in security check data verification according to the consensus mechanism and saves a full copy of the customized interface network to ensure decentralized storage of security check data. Establish a smart contract framework, implement permission control of the security check process through the smart contract of the customized interface network, define permission roles, establish access control based on security check permission role identity, install edge devices on each alliance chain node, connect to the X-ray machine to collect security check data, install AI models and training environment, set up TLS / SSL and configure a unified key management system.

[0031] The respective responsibilities of security checkpoints, management agencies, and regulatory units are allocated as follows: the responsibilities of security checkpoint nodes include collecting security data, executing risk assessment models, uploading security data, and calling on-chain smart contracts; the responsibilities of management agency nodes include monitoring operating status, approving authority, reviewing on-chain security events, and archiving audit records; the responsibilities of regulatory unit nodes include participating in model training, parameter fusion, and security data analysis, and saving a full copy of the customized interface network for each alliance chain node to ensure decentralized storage of data.

[0032] The specific permission roles include: security personnel, managers, management agencies and AI model managers. Among them, security personnel can only upload security data and provide execution suggestions based on AI decisions; managers have the authority to view and verify all security data; management agencies are responsible for monitoring the flow of security data; AI model managers are responsible for adjusting and optimizing the algorithms of AI models and defining new security decision-making logic.

[0033] The smart contract framework includes: defining a security data contract to automatically record and verify the upload, modification, and query operations of security data; defining an AI decision-making contract so that when a security check point completes security data collection, the smart contract automatically calls the AI model to analyze the security data and generate decisions based on preset rules; and defining a compliance contract to verify whether the security check process meets compliance requirements.

[0034] Model training module: This module loads security inspection data through edge devices and conducts multiple rounds of training. It uses neural networks to analyze images or behavioral sequences and combines them with anomaly detection models to identify potential threats.

[0035] Generate the initial parameters of the global security inspection model, initialize it with random numbers, load security inspection data through edge devices, and conduct multiple rounds of training for the global security inspection model. Encrypt the model parameters with a homomorphic encryption algorithm and upload them to the aggregation node on the alliance chain. Select a neural network (such as CNN, LSTM) to perform sequence analysis of images or behaviors of security inspection data. Use anomaly detection models (such as autoencoders, isolation forests) to identify potential threats. Incrementally train security inspection data, use transfer learning to reduce training time, set an early stopping mechanism to prevent overfitting, extract all updateable parameters in the model, use asymmetric encryption technology to encrypt model parameters during transmission, and upload them to the aggregation node on the alliance chain through a secure communication line.

[0036] Fusion update module: Receives multiple model parameters from each node, fuses the models based on a dynamic fusion weight algorithm, triggers the calculation of the fused global model parameters, generates a new model version, and sends it to the edge node. After receiving it, the edge device automatically loads it into the local inference engine for application;

[0037] The model parameters from each consortium chain node are received through the smart contract on the consortium chain, and the model parameter fusion algorithm is defined in the consortium chain smart contract. The specific formula is:

[0038]

[0039] in, represents the dynamic fusion weight of the i-th alliance chain node in the t-th round, α represents the influence of the model performance on the weight, It represents the model performance index of the current alliance chain node on the current round of verification set. β represents the influence of the stability and credibility of the alliance chain node on the weight. It represents the evaluation score of the stability and credibility of the alliance chain node, and γ represents the impact of the hardware environment of the alliance chain node on the weight. The indicator representing the hardware environment where the alliance chain node is located triggers the smart contract call, calculates the integrated global model parameters, and generates a new global model version. The alliance chain actively sends the latest global model parameters to the edge node through a secure channel. After the edge device receives the model parameters, it automatically loads them into the local inference engine.

[0040] Identification and decision-making module: loads the latest global security inspection model, receives security inspection data and performs risk identification, generates and packages risk scores and level labels, calls smart contracts, writes the packaged results to the consortium chain, sets decision-making rules, and automatically triggers corresponding security inspection measures based on risk scores and levels;

[0041] The latest global security inspection model is loaded through the edge device, security inspection data is received and risk identification is performed, a risk score and risk level label of 0-100 is generated, the risk score, risk level label, passenger identity ID and timestamp are packaged, the smart contract is called through a secure channel, the packaged result is written into the alliance chain to form a traceable certificate, and decision-making rules are set in the smart contract, including: Risk score >80: trigger "secondary security inspection" or "manual verification"; risk level is "high risk", automatic prompt for manual intervention or denial of passage; otherwise, normal passage is allowed.

[0042] Example 2

[0043] The following is another embodiment of the present invention, which provides an intelligent security inspection decision-making system based on risk management. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0044] After running the baggage collection module, similarity measurement module, shared screening module, baggage identification module, human body analysis module, and decision guidance module of an automated baggage handling system based on X-ray inspection, the foundational module simulated a consortium blockchain network, defining security checkpoints, management agencies, and regulatory bodies as nodes. A consensus mechanism (such as PBFT or Raft) was deployed to achieve security check data verification and decentralized storage. Edge devices were installed at each node, connecting to X-ray machines to collect radiation images and behavioral data. The system implemented TLS / SSL security protocols and deployed efficient smart contracts for permission control and access management. During the experiment, security checkpoints collected large-scale, multi-dimensional data (an average of 2,000 security check images and behavioral sequences per day) and uploaded it to a customized interface network. Monitoring nodes generated 100 blocks per hour, with a total on-chain storage capacity of 200GB. The system achieved efficient data verification and evidence storage. The management agency monitored system status, reviewed login logs, and archived over 2,000 audit records to ensure transparent and trustworthy operations. Regulatory agencies participate in model training, using edge devices to load training data and conduct multiple rounds of training (each round of data involves approximately 500MB). CNN and LSTM are used to optimize the detection model and improve recognition rates. After multiple rounds of training, model parameters are uploaded to the consortium chain using homomorphic encryption for parameter fusion, with the fusion coefficient dynamically adjusted based on node performance and stability. After fusion, the model's accuracy on the validation set increased from 78% to 92%. Edge devices utilize the latest models for real-time risk identification, automatically assessing each passenger's risk score (0-100) and risk level (low / medium / high), with an average single analysis time of less than 150 milliseconds. For risk scores greater than 80, the system automatically triggers a "secondary security check," with an average measured response time of 200 milliseconds and a false alarm rate of less than 3%. All data upload and processing processes are recorded in real time on the consortium chain to ensure data immutability and traceability. The overall simulation results show that the system performs excellently in improving recognition rate, ensuring data security, reducing personnel burden and enhancing response capabilities, effectively verifying the security, accuracy and automation advantages of the invention in practical applications.

[0045] By introducing advanced AI intelligent analysis technology, the automated analysis and decision-making process of security inspection data has been realized, greatly improving the objectivity and scientific nature of security detection. A deep learning-based intelligent model is used to quickly and comprehensively analyze and score the collected security inspection data, ensuring the scientificity and objectivity of each judgment. It can achieve structured management of multi-source and multi-channel data, effectively integrating multi-dimensional information from different security inspection points, equipment and scenarios, providing a global perspective on the overall situation. Automated analysis supports real-time early warning and risk assessment, timely identifying potential threats, and improving emergency response capabilities. The use of automated decision-making processes greatly reduces human bias, alleviates staff workload, and reduces safety hazards caused by human errors. Data structuring and integrated management promote the digitalization and intelligence of security inspection processes, continuously improve the system's security assurance capabilities, free up human resources, and enable security personnel to focus on handling highly complex or emergency incidents. This not only ensures the objectivity and accuracy of security assessments, but also realizes the digital transformation of security management, laying a solid foundation for the development of smart security inspections and promoting the industry's upgrade from experience-based to data-driven.

[0046] Through intelligent analysis and comprehensive evaluation, we no longer rely on personal experience, greatly reduce the deviation of human judgment, and ensure that each safety judgment is based on scientific data and algorithm models, thereby improving the accuracy of overall detection. The automated operation of the system can reduce dependence on staff, reduce labor intensity, and relieve staff pressure, allowing operators to focus more on processing high-value or complex scenarios, effectively alleviating the problem of tight staffing. It adopts a digital and structured management concept to achieve seamless docking and integration of various types of heterogeneous business data, support multi-channel and multi-process data integration, form a complete security information chain, and combine customized interfaces and smart contract technology to ensure data security, immutability and full-process traceability, thereby improving the transparency and credibility of management.

[0047] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. The intelligent security inspection decision-making system based on risk management is characterized by: It includes basic construction module, model training module, fusion update module and recognition decision module; Basic building module: Deploy a customized interface network and define alliance chain nodes. Use the consensus mechanism to participate in security data verification and decentralized storage. Implement permission control and access management through smart contracts. Install edge devices at each node to connect to X-ray machines to collect security data. Model training module: This module loads security inspection data through edge devices and conducts multiple rounds of training. It uses neural networks to analyze images or behavioral sequences and combines them with anomaly detection models to identify potential threats. Fusion update module: Receives multiple model parameters from each node, fuses the models based on a dynamic fusion weight algorithm, triggers the calculation of the fused global model parameters, generates a new model version, and sends it to the edge node. After receiving it, the edge device automatically loads it into the local inference engine for application; Identification and decision-making module: loads the latest global security inspection model, receives security inspection data and performs risk identification, generates and packages risk scores and level labels, calls smart contracts, writes the packaged results into the alliance chain, sets decision-making rules, and automatically triggers corresponding security inspection measures based on risk scores and levels.

2. The risk management-based intelligent security inspection decision-making system according to claim 1 is characterized by: In the basic building module, the alliance chain nodes are security checkpoint nodes, management agency nodes, and regulatory unit nodes, and their respective responsibilities include: the responsibilities of the security checkpoint nodes include collecting security check data, executing risk assessment models, uploading security check data, and calling on-chain smart contracts; the responsibilities of the management agency nodes include monitoring operating status, approving authority, reviewing on-chain security check events, and archiving audit records; the responsibilities of the regulatory unit nodes include participating in model training, parameter fusion, and security check data analysis, and saving a full copy of the customized interface network for each alliance chain node.

3. The risk management-based intelligent security inspection decision-making system according to claim 1 is characterized by: In the basic building module, each alliance chain node participates in security inspection data verification according to the consensus mechanism, saves a full copy of the customized interface network, establishes a smart contract framework, implements permission control of the security inspection process through the smart contract of the customized interface network, defines permission roles, establishes access control based on the identity of security inspection permission roles, installs edge devices on each alliance chain node, and connects to X-ray machines to collect security inspection data.

4. The risk management-based intelligent security inspection decision-making system according to claim 1 is characterized by: The model training module generates initial parameters of the global security inspection model, initializes it with random numbers, loads security inspection data through edge devices, and performs multiple rounds of training on the global security inspection model. The model parameters are encrypted using a homomorphic encryption algorithm and uploaded to the aggregation node on the alliance chain. A neural network is selected to perform sequence analysis of images or behaviors of the security inspection data, an anomaly detection model is used to identify potential threats, security inspection data is incrementally trained, transfer learning is used to reduce training time, an early stopping mechanism is set to prevent overfitting, all updateable parameters in the model are extracted, and asymmetric encryption technology is used to encrypt the model parameters during transmission. The parameters are uploaded to the aggregation node on the alliance chain through a secure communication line.

5. The risk management-based intelligent security inspection decision-making system according to claim 1 is characterized by: The fusion update module receives model parameters from each alliance chain node through the smart contract on the alliance chain, defines the model parameter fusion algorithm in the alliance chain smart contract, triggers the smart contract call, calculates the fused global model parameters, and generates a new global model version. The alliance chain actively sends the latest global model parameters to the edge node through a secure channel. After receiving the model parameters, the edge device automatically loads them into the local inference engine.

6. The risk management-based intelligent security inspection decision-making system according to claim 5 is characterized by: The specific formula of the model parameter fusion algorithm is: in, represents the dynamic fusion weight of the i-th alliance chain node in the t-th round, α represents the influence of the model performance on the weight, It represents the model performance index of the current alliance chain node on the current round of verification set. β represents the influence of the stability and credibility of the alliance chain node on the weight. It represents the evaluation score of the stability and credibility of the alliance chain node, and γ represents the impact of the hardware environment of the alliance chain node on the weight. An indicator that represents the hardware environment where the consortium chain node is located.

7. The risk management-based intelligent security inspection decision-making system according to claim 1 is characterized by: The identification and decision-making module loads the latest global security inspection model through the edge device, receives security inspection data and performs risk identification, generates a risk score and risk level label from 0 to 100, packages the risk score, risk level label, passenger identity ID and timestamp, calls the smart contract through a secure channel, writes the packaged results into the alliance chain to form a traceable certificate, and sets the decision-making rules in the smart contract.

8. A computer device comprising: memory and processor; The memory stores a program of an intelligent security inspection decision-making system based on risk management, which is characterized in that when the processor executes the system, the module function of any system described in claims 1-7 is realized.

Images