Quantum security protection-based community police information early warning model training system

Abstract

The invention relates to model training, in particular to a community police information early warning model training system based on quantum safety protection, which comprises a multivariate data identification unit, a data processing unit, a data processing unit, a data processing unit, a data processing unit and a data processing unit, model training samples corresponding to the modal data are sent to the model training management unit; the model training management unit verifies the model training samples sent by the multivariate data identification unit, performs model training on models corresponding to various modal data, and optimizes the models according to model training quality and model operation efficiency scores provided by the model deployment unit; the model deployment unit is used for deploying the trained model in the operation environment and monitoring and evaluating the operation state and the operation efficiency of the model; according to the technical scheme provided by the invention, the defects of low accuracy and efficiency of community police information early warning caused by difficulty in effective model training can be effectively overcome.

Description

Technical Field

[0001] The present invention relates to model training, and in particular to a community policing information early warning model training system based on quantum security protection. Background Art

[0002] Community policing information analysis and early warning technology is an important part of current community policing work. It uses advanced technologies such as big data and artificial intelligence to conduct in-depth mining and analysis of community policing information to achieve the purpose of early warning, crime prevention and improved police efficiency.

[0003] With the widespread application of video surveillance, traditional community policing information analysis technology mainly focuses on various scene analyses based on video data and using edge computing gateways or intelligent servers as hardware. This type of technology extracts video sample data to train the model analysis engine, and then uses various scene models to identify segments in the video stream that meet specific scenes, thereby achieving the purpose of intelligent analysis of video data. For example, scenes such as traffic violations, illegal road occupation, and personnel control in key places all use video streams as the basis for identifying corresponding scenes.

[0004] However, with the rapid development of Internet technology, community policing information has also shown a multimodal and diversified development trend. Using video data as the only source of community policing information will inevitably lead to the loss of a lot of important information, which will greatly reduce the integrity of community policing information analysis and cause low accuracy and efficiency of community policing information warning. Summary of the invention

[0005] 1. Technical issues to be solved

[0006] In view of the above-mentioned shortcomings of the prior art, the present invention provides a community police information early warning model training system based on quantum security protection, which can effectively overcome the defects of the prior art in that the accuracy and efficiency of community police information early warning are low due to the difficulty in effective model training.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A community policing information early warning model training system based on quantum security protection, comprising a multivariate data recognition unit, a model training management unit, a model deployment unit, a quantum security protection unit and an input and output unit;

[0010] A multivariate data recognition unit performs multivariate type classification after preprocessing the multivariate data, extracts samples from various types of modal data, and sends model training samples corresponding to various types of modal data to the model training management unit;

[0011] The model training management unit verifies the model training samples sent by the multivariate data recognition unit, performs model training on the models corresponding to various modal data, and optimizes the models according to the model training quality and the model operation efficiency score provided by the model deployment unit;

[0012] The model deployment unit deploys the trained model in the operating environment, monitors and evaluates the operating status and operating efficiency of the model, obtains the operating status score and operating efficiency score of the model corresponding to various modal data, optimizes the response service according to the operating status score of the model, and updates the operating efficiency score to the corresponding model;

[0013] The quantum security protection unit provides traceable quantum keys for data that needs to be protected and matches quantum encryption algorithms to achieve encryption and decryption of sensitive information;

[0014] The input and output unit checks the legitimacy of the input information and then sends it to the corresponding processing module. At the same time, it uses the quantum key provided by the quantum security protection unit and the matching quantum encryption algorithm to encrypt the data to be protected and then exchange it. It also completes data decryption and data verification before outputting it to the output terminal to output sensitive information to the output terminal.

[0015] Preferably, the multivariate data identification unit includes a data preprocessing module, a data classification identification module and a data sample generation module;

[0016] The data preprocessing module preprocesses the multivariate data according to the data format and its multimodal characteristics, removes unqualified data with abnormal data format and missing information, and repairs the data errors that can be repaired;

[0017] The data classification and recognition module pre-reads the header information of the pre-processed data, matches the data based on its multimodal features, assigns a classification identification number to the matching data, and assigns a non-classification identification number to the unmatched data;

[0018] The data sample generation module matches the corresponding sample data extraction method by reading the classification identification number of each type of modal data, extracts samples from each type of modal data using the sample data extraction method, generates model training samples corresponding to each type of modal data, and appends the hash value of each data to the end of the corresponding original data in the model training sample.

[0019] Preferably, the data classification and identification module pre-reads the header information of the pre-processed data, matches the data in combination with the multimodal features to which the data belongs, assigns a classification identification number to the matching data, and assigns an unclassified identification number to the unmatched data, including:

[0020] S1. Read the header information of the specified length of the preprocessed data header and calculate the hash value of the header information:

[0021] IHH i =HSAH(DATA i );

[0022] Among them, IHHV i DATA is the header information of the i-th preprocessed data i HSAH(●) is the hash operation;

[0023] S2. Compare the hash value of the header information with the characteristic values ​​of each type of modal data. If they are equal, write the characteristic value of the corresponding category of modal data into the preprocessed data as the classification identification number.

[0024] Preferably, the model training management unit includes a sample pre-verification module, a sample training implementation module and a sample training detection module;

[0025] The sample pre-verification module performs integrity verification on the model training samples sent by the multivariate data recognition unit through a hash algorithm, assigns a normal status flag to the model training samples that pass the verification, and discards the model training samples that fail the verification;

[0026] The sample training implementation module inputs the model training samples containing the normal state mark into the model corresponding to the corresponding category modal data for model training;

[0027] The sample training detection module tracks and monitors various indicators of the model training process in real time, checks the model training results, evaluates the model training quality, optimizes the model parameters based on the model training quality, and performs secondary optimization of the model parameters based on the model operation efficiency score provided by the model deployment unit.

[0028] Preferably, the model deployment unit includes a model deployment conversion module, a model status monitoring module and a model efficiency evaluation module;

[0029] The model deployment conversion module packages the parameters, configurations and environments of the trained model into a model deployment package, and deploys the model deployment package to the operating environment;

[0030] The model status monitoring module monitors the status of models in each operating environment, marks abnormal status of models and presets recovery operations, sorts the running status scores of each model running instance from high to low, and responds to service requests based on the score ranking;

[0031] The model efficiency evaluation module regularly calculates the prediction accuracy of the models in each operating environment, calculates the model's operating efficiency score based on the model's prediction accuracy, and updates the operating efficiency score to the corresponding model.

[0032] Preferably, the quantum security protection unit includes a quantum monitoring module, a quantum key management module and a quantum algorithm management module;

[0033] The quantum monitoring module performs necessary initialization on the quantum security hardware and then monitors its operating status in real time, and issues an alarm when the operating status is abnormal;

[0034] The quantum key management module obtains the quantum random number from the quantum monitoring module, converts the quantum random number into a quantum key through a conversion algorithm, outputs the quantum key after assigning a unique identity, and tracks the use of each quantum key;

[0035] The quantum algorithm management module matches the optimal quantum encryption algorithm according to the type of encrypted data and encryption requirement level that the data requester needs to transmit, and dynamically updates the capability set and availability of the internal quantum encryption algorithm to support dynamic addition and reduction of algorithms.

[0036] Preferably, the input-output unit includes an input information detection module, a data exchange module and a data output terminal management module;

[0037] The input information detection module performs a validity check on the input parameters and configuration information, identifies risky input information, stores it in the illegal input buffer, and notifies the information input party;

[0038] The data exchange module uses the quantum key and matching quantum encryption algorithm provided by the quantum security protection unit to encrypt the data exchanged between different modules and systems before exchanging them, and completes data decryption and data verification before outputting to the output terminal, thus building an effective data encryption and decryption coordination mechanism for the data exchange process;

[0039] The data output terminal management module enters and manages the attributes of the output terminal, attaches the output capability sets of different output terminals to the terminal information, and outputs the received sensitive information within the range of the output capability set.

[0040] (III) Beneficial effects

[0041] Compared with the prior art, the community police information early warning model training system based on quantum security protection provided by the present invention proposes to use a quantum encryption algorithm with anti-quantum attack capability in combination with a quantum key, and at the same time combine multivariate data for model training, thereby greatly expanding the community police information early warning model training data set. Various sensitive information that could not be exported outside the public security domain due to security reasons is protected by quantum encryption before model training is performed outside the domain. On the premise of ensuring data security, social computing resources are fully utilized to make up for the lack of public security proprietary computing power. In particular, for the situation where grassroots police work units have limited computing resources and insufficient technical personnel, the present invention greatly reduces the difficulty and application threshold of community police information analysis and early warning, and helps to promote the rapid development of intelligent community policing work. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 It is a system schematic diagram of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] A community police information warning model training system based on quantum security protection, such as Figure 1 As shown, it includes a multivariate data recognition unit, a model training management unit, a model deployment unit, a quantum security protection unit and an input and output unit;

[0046] A multivariate data recognition unit performs multivariate type classification after preprocessing the multivariate data, extracts samples from various types of modal data, and sends model training samples corresponding to various types of modal data to the model training management unit;

[0047] The model training management unit verifies the model training samples sent by the multivariate data recognition unit, performs model training on the models corresponding to various modal data, and optimizes the models according to the model training quality and the model operation efficiency score provided by the model deployment unit;

[0048] The model deployment unit deploys the trained model in the operating environment, monitors and evaluates the operating status and operating efficiency of the model, obtains the operating status score and operating efficiency score of the model corresponding to various modal data, optimizes the response service according to the operating status score of the model, and updates the operating efficiency score to the corresponding model;

[0049] The quantum security protection unit provides traceable quantum keys for data that needs to be protected and matches quantum encryption algorithms to achieve encryption and decryption of sensitive information;

[0050] The input and output unit checks the legitimacy of the input information and then sends it to the corresponding processing module. At the same time, it uses the quantum key provided by the quantum security protection unit and the matching quantum encryption algorithm to encrypt the data to be protected and then exchange it. It also completes data decryption and data verification before outputting it to the output terminal to output sensitive information to the output terminal.

[0051] ①The multivariate data recognition unit includes a data preprocessing module, a data classification recognition module and a data sample generation module;

[0052] The data preprocessing module preprocesses the multivariate data according to the data format and its multimodal characteristics, removes unqualified data with abnormal data format and missing information, and repairs the data errors that can be repaired;

[0053] The data classification and recognition module pre-reads the header information of the pre-processed data, matches the data based on its multimodal features, assigns a classification identification number to the matching data, and assigns a non-classification identification number to the unmatched data;

[0054] The data sample generation module matches the corresponding sample data extraction method by reading the classification identification number of each type of modal data, extracts samples from each type of modal data using the sample data extraction method, generates model training samples corresponding to each type of modal data, and appends the hash value of each data to the end of the corresponding original data in the model training sample.

[0055] Specifically, the data classification and identification module pre-reads the header information of the pre-processed data, matches the data in combination with the multimodal features to which the data belongs, assigns a classification identification number to the matching data, and assigns an unclassified identification number to the unmatched data, including:

[0056] S1. Read the header information of the specified length of the preprocessed data header and calculate the hash value of the header information:

[0057] IHH i =HSAH(DATA i );

[0058] Among them, IHHV i DATA is the header information of the i-th preprocessed data i HSAH(●) is the hash operation;

[0059] S2. Compare the hash value of the header information with the characteristic values ​​of each type of modal data. If they are equal, write the characteristic value of the corresponding category of modal data into the preprocessed data as the classification identification number.

[0060] ② The model training management unit includes a sample pre-verification module, a sample training implementation module and a sample training detection module;

[0061] The sample pre-verification module performs integrity verification on the model training samples sent by the multivariate data recognition unit through a hash algorithm, assigns a normal status flag to the model training samples that pass the verification, and discards the model training samples that fail the verification;

[0062] The sample training implementation module inputs the model training samples containing the normal state mark into the model corresponding to the corresponding category modal data for model training;

[0063] The sample training detection module tracks and monitors various indicators of the model training process in real time, checks the model training results, evaluates the model training quality, optimizes the model parameters based on the model training quality, and performs secondary optimization of the model parameters based on the model operation efficiency score provided by the model deployment unit.

[0064] ③The model deployment unit includes the model deployment conversion module, the model status monitoring module and the model efficiency evaluation module;

[0065] The model deployment conversion module packages the parameters, configurations and environments of the trained model into a model deployment package, and deploys the model deployment package to the operating environment;

[0066] The model status monitoring module monitors the status of models in each operating environment, marks abnormal status of models and presets recovery operations, sorts the running status scores of each model running instance from high to low, and responds to service requests based on the score ranking;

[0067] The model efficiency evaluation module regularly calculates the prediction accuracy of the models in each operating environment, calculates the model's operating efficiency score based on the model's prediction accuracy, and updates the operating efficiency score to the corresponding model.

[0068] ④The quantum security protection unit includes quantum monitoring module, quantum key management module and quantum algorithm management module;

[0069] The quantum monitoring module performs necessary initialization on the quantum security hardware and then monitors its operating status in real time, and issues an alarm when the operating status is abnormal;

[0070] The quantum key management module obtains the quantum random number from the quantum monitoring module, converts the quantum random number into a quantum key through a conversion algorithm, outputs the quantum key after assigning a unique identity, and tracks the use of each quantum key;

[0071] The quantum algorithm management module matches the optimal quantum encryption algorithm according to the type of encrypted data and encryption requirement level that the data requester needs to transmit, and dynamically updates the capability set and availability of the internal quantum encryption algorithm to support dynamic addition and reduction of algorithms.

[0072] ⑤The input and output unit includes an input information detection module, a data exchange module and a data output terminal management module;

[0073] The input information detection module performs a validity check on the input parameters and configuration information, identifies risky input information, stores it in the illegal input buffer, and notifies the information input party;

[0074] The data exchange module uses the quantum key and matching quantum encryption algorithm provided by the quantum security protection unit to encrypt the data exchanged between different modules and systems before exchanging them, and completes data decryption and data verification before outputting to the output terminal, thus building an effective data encryption and decryption coordination mechanism for the data exchange process;

[0075] The data output terminal management module enters and manages the attributes of the output terminal, attaches the output capability sets of different output terminals to the terminal information, and outputs the received sensitive information within the range of the output capability set.

[0076] The technical solution of this application proposes to use a quantum encryption algorithm with anti-quantum attack capabilities in conjunction with a quantum key, and at the same time combine it with multivariate data for model training, thereby greatly expanding the training data set of the community police information early warning model. Various types of sensitive information that could not be exported outside the public security domain due to security reasons are protected by quantum encryption and then model training is carried out outside the domain. Under the premise of ensuring data security, social computing resources are fully utilized to make up for the lack of public security proprietary computing power. In particular, for the situation where grassroots police work units have limited computing resources and insufficient technical personnel, the present invention greatly reduces the difficulty and application threshold of community police information analysis and early warning, and helps to promote the rapid development of intelligent community policing work.

[0077] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A community policing information early warning model training system based on quantum security protection, characterized by: It includes a multivariate data recognition unit, a model training management unit, a model deployment unit, a quantum security protection unit, and an input and output unit; A multivariate data recognition unit performs multivariate type classification after preprocessing the multivariate data, extracts samples from various types of modal data, and sends model training samples corresponding to various types of modal data to the model training management unit; The model training management unit verifies the model training samples sent by the multivariate data recognition unit, performs model training on the models corresponding to various modal data, and optimizes the models according to the model training quality and the model operation efficiency score provided by the model deployment unit; The model deployment unit deploys the trained model in the operating environment, monitors and evaluates the operating status and operating efficiency of the model, obtains the operating status score and operating efficiency score of the model corresponding to various modal data, optimizes the response service according to the operating status score of the model, and updates the operating efficiency score to the corresponding model; The quantum security protection unit provides traceable quantum keys for data that needs to be protected and matches quantum encryption algorithms to achieve encryption and decryption of sensitive information; The input and output unit checks the legitimacy of the input information and then sends it to the corresponding processing module. At the same time, it uses the quantum key provided by the quantum security protection unit and the matching quantum encryption algorithm to encrypt the data to be protected and then exchange it. It also completes data decryption and data verification before outputting it to the output terminal to output sensitive information to the output terminal.

2. The community policing information early warning model training system based on quantum security protection according to claim 1 is characterized by: The multivariate data recognition unit includes a data preprocessing module, a data classification recognition module and a data sample generation module; The data preprocessing module preprocesses the multivariate data according to the data format and its multimodal characteristics, removes unqualified data with abnormal data format and missing information, and repairs the data errors that can be repaired; The data classification and recognition module pre-reads the header information of the pre-processed data, matches the data based on its multimodal features, assigns a classification identification number to the matching data, and assigns a non-classification identification number to the unmatched data; The data sample generation module matches the corresponding sample data extraction method by reading the classification identification number of each type of modal data, extracts samples from each type of modal data using the sample data extraction method, generates model training samples corresponding to each type of modal data, and appends the hash value of each data to the end of the corresponding original data in the model training sample.

3. The community policing information early warning model training system based on quantum security protection according to claim 2 is characterized by: The data classification and identification module pre-reads the header information of the pre-processed data, matches the data in combination with the multimodal features to which the data belongs, assigns a classification identification number to the matching data, and assigns an unclassified identification number to the unmatched data, including: S1. Read the header information of the specified length of the preprocessed data header and calculate the hash value of the header information: IHHV i =VALID(DATA i ); Among them, IHHV i DATA is the header information of the i-th preprocessed data i Hash value, HSAH(●) is the hash operation; S2. Compare the hash value of the header information with the characteristic values ​​of each type of modal data. If they are equal, write the characteristic value of the corresponding category of modal data into the preprocessed data as the classification identification number.

4. The community policing information early warning model training system based on quantum security protection according to claim 2 is characterized by: The model training management unit includes a sample pre-verification module, a sample training implementation module and a sample training detection module; The sample pre-verification module performs integrity verification on the model training samples sent by the multivariate data recognition unit through a hash algorithm, assigns a normal status flag to the model training samples that pass the verification, and discards the model training samples that fail the verification; The sample training implementation module inputs the model training samples containing the normal state mark into the model corresponding to the corresponding category modal data for model training; The sample training detection module tracks and monitors various indicators of the model training process in real time, checks the model training results, evaluates the model training quality, optimizes the model parameters based on the model training quality, and performs secondary optimization of the model parameters based on the model operation efficiency score provided by the model deployment unit.

5. The community policing information early warning model training system based on quantum security protection according to claim 4 is characterized in that: The model deployment unit includes a model deployment conversion module, a model status monitoring module and a model efficiency evaluation module; The model deployment conversion module packages the parameters, configurations and environment of the trained model into a model deployment package, and deploys the model deployment package to the operating environment; The model status monitoring module monitors the status of models in each operating environment, marks abnormal status of models and presets recovery operations, sorts the running status scores of each model running instance from high to low, and responds to service requests based on the score ranking; The model efficiency evaluation module regularly calculates the prediction accuracy of the models in each operating environment, calculates the model's operating efficiency score based on the model's prediction accuracy, and updates the operating efficiency score to the corresponding model.

6. The community policing information early warning model training system based on quantum security protection according to claim 5 is characterized by: The quantum security protection unit includes a quantum monitoring module, a quantum key management module and a quantum algorithm management module; The quantum monitoring module performs necessary initialization on the quantum security hardware and then monitors its operating status in real time, and issues an alarm when the operating status is abnormal; The quantum key management module obtains the quantum random number from the quantum monitoring module, converts the quantum random number into a quantum key through a conversion algorithm, outputs the quantum key after assigning a unique identity, and tracks the use of each quantum key; The quantum algorithm management module matches the optimal quantum encryption algorithm according to the type of encrypted data and encryption requirement level that the data requester needs to transmit, and dynamically updates the capability set and availability of the internal quantum encryption algorithm to support dynamic addition and reduction of algorithms.

7. The community policing information early warning model training system based on quantum security protection according to claim 6 is characterized by: The input-output unit includes an input information detection module, a data exchange module and a data output terminal management module; The input information detection module performs a validity check on the input parameters and configuration information, identifies risky input information, stores it in the illegal input buffer, and notifies the information input party; The data exchange module uses the quantum key provided by the quantum security protection unit and the matching quantum encryption algorithm to encrypt the data exchanged between different modules and systems before exchanging them, and completes data decryption and data verification before outputting to the output terminal, thus building an effective data encryption and decryption coordination mechanism for the data exchange process; The data output terminal management module enters and manages the attributes of the output terminal, attaches the output capability sets of different output terminals to the terminal information, and outputs the received sensitive information within the range of the output capability set.

Images