CN114416829A

CN114416829A - Network training method based on machine learning and cloud authentication service system

Info

Publication number: CN114416829A
Application number: CN202210147068.7A
Authority: CN
Inventors: 周应凤
Original assignee: Individual
Current assignee: Individual
Priority date: 2021-01-26
Filing date: 2021-01-26
Publication date: 2022-04-29
Also published as: CN114416830A; CN112860759B; CN112860759A

Abstract

The embodiment of the disclosure provides a network training method based on machine learning and a cloud authentication service system, wherein service propagation associated parameters of labeled samples under at least two service hotspot propagation labels and corresponding first propagation linkage data, service interaction contact information among the service propagation associated parameters and corresponding second propagation linkage data, and actual service propagation associated parameters among the labeled samples are obtained, so that the machine learning network is repeatedly trained, the accuracy of the service propagation associated parameters among service data obtained by machine learning network prediction can be improved, and the determination accuracy of the service propagation associated parameters among the service data is further improved.

Description

Network training method based on machine learning and cloud authentication service system

The application is a divisional application of Chinese application with the name of 'big data mining method based on block chain security authentication and cloud authentication service system' invented and created by application number 202110104781.9, application date of 26/01/2021.

Technical Field

The disclosure relates to the technical field of cloud services, in particular to a network training method and a cloud authentication service system based on machine learning.

Background

In the related art, deep mining of push topics may be involved in a big data mining process, and in order to more accurately reflect the association relationship between business topics, currently, single-round information recommendation of a single topic is upgraded to linkage information recommendation of multiple topics, that is, currently, compound recommendation can be performed based on topic linkage information, so as to improve user experience. However, in the current determination method of theme linkage information (that is, linkage pushing is performed for different themes), based on the theme tags between the service data, the same content of the theme tags is determined as the theme linkage information, for example, information with the same theme is determined as the theme linkage information, so that the determined theme linkage information includes a large amount of information of the same type of theme tags, but the actual service propagation associated parameter is low, so that the determination accuracy of the theme linkage information is low.

Disclosure of Invention

In order to overcome at least the above disadvantages in the prior art, an object of the present disclosure is to provide a network training method and a cloud authentication service system based on machine learning.

In a first aspect, the present disclosure provides a big data mining method based on blockchain security authentication, which is applied to a cloud authentication service system, where the cloud authentication service system is in communication connection with a plurality of blockchain authentication terminals, and the method includes:

acquiring service big data which is acquired by distributing a corresponding target service acquisition process to the generated authentication service transmission channel after the block chain authentication terminal and the cloud authentication service system pass dynamic bidirectional authentication;

acquiring data of the service big data under at least two service hot spot propagation tags, and determining service hot spot propagation characteristics of the service big data under the at least two service hot spot propagation tags according to the data of the service big data under the at least two service hot spot propagation tags;

obtaining service propagation associated parameters among the service big data according to service hot spot propagation characteristics of the service big data under the at least two service hot spot propagation labels;

and determining target theme linkage information of each service big data according to service propagation associated parameters among the service big data, and pushing information to the block chain authentication terminal based on the target theme linkage information.

In a second aspect, an embodiment of the present disclosure further provides a big data mining device based on blockchain security authentication, which is applied to a cloud authentication service system, where the cloud authentication service system is in communication connection with a plurality of blockchain authentication terminals, and the cloud authentication service system is implemented based on a cloud computing platform, and the device includes:

the acquisition module is used for acquiring service big data which is generated after the block chain authentication terminal and the cloud authentication service system pass dynamic bidirectional authentication and used for distributing a corresponding target service acquisition process to the generated authentication service transmission channel to acquire big data of the authentication service transmission channel;

the first determining module is used for acquiring data of the service big data under at least two service hot spot propagation tags, and determining service hot spot propagation characteristics of the service big data under the at least two service hot spot propagation tags according to the data of the service big data under the at least two service hot spot propagation tags;

a second determining module, configured to obtain a service propagation association parameter between the service big data according to service hot spot propagation characteristics of the service big data under the at least two service hot spot propagation tags;

and the pushing module is used for determining target theme linkage information of each service big data according to service propagation associated parameters among all the service big data and pushing information to the block chain authentication terminal based on the target theme linkage information.

In a third aspect, an embodiment of the present disclosure further provides a big data mining system based on block chain security authentication, where the big data mining system based on block chain security authentication includes a cloud authentication service system and a plurality of block chain authentication terminals communicatively connected to the cloud authentication service system;

the cloud authentication service system is configured to:

In a fourth aspect, an embodiment of the present disclosure further provides a cloud authentication service system, where the cloud authentication service system includes a processor, a machine-readable storage medium, and a network interface, where the machine-readable storage medium, the network interface, and the processor are connected through a bus system, the network interface is configured to be in communication connection with at least one blockchain authentication terminal, the machine-readable storage medium is configured to store a program, an instruction, or a code, and the processor is configured to execute the program, the instruction, or the code in the machine-readable storage medium to perform the big data mining method based on blockchain security authentication in any one of the first aspect or any one of the first aspect possible design examples.

In a fifth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, where instructions are preset in the computer-readable storage medium, and when the instructions are executed, the computer executes the big data mining method based on the block chain security authentication in the first aspect or any one of the possible design examples of the first aspect.

Based on any one of the above aspects, the present disclosure determines service hotspot propagation characteristics of the service big data under the at least two service hotspot propagation labels according to the obtained data of the service big data under the at least two service hotspot propagation labels, so as to obtain service propagation associated parameters between the service big data, and determines target topic linkage information of each service big data based on the service propagation associated parameters between the service big data, and further may comprehensively determine the service propagation associated parameters between the service big data through the plurality of service hotspot propagation labels, so that the determination process of the service hotspot propagation between the service data is more accurate, thereby improving the determination accuracy of the service propagation associated parameters between the service data, and further improving the determination accuracy of the topic linkage information. Meanwhile, based on the accurately determined theme linkage information, the information push is favorably realized.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings that need to be called in the embodiments are briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present disclosure, and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.

Fig. 1 is a schematic view of an application scenario of a big data mining system based on block chain security authentication according to an embodiment of the present disclosure;

fig. 2 is a schematic flowchart of a big data mining method based on block chain security authentication according to an embodiment of the present disclosure;

fig. 3 is a functional module schematic diagram of a big data mining device based on block chain security authentication according to an embodiment of the present disclosure;

fig. 4 is a schematic block diagram of structural components of a cloud authentication service system for implementing the above-described big data mining method based on block chain security authentication according to an embodiment of the present disclosure.

Detailed Description

The present disclosure is described in detail below with reference to the drawings, and the specific operation methods in the method embodiments can also be applied to the device embodiments or the system embodiments.

Fig. 1 is a schematic application scenario diagram of a big data mining system 10 based on block chain security authentication according to an embodiment of the present disclosure. The big data mining system 10 based on the blockchain security authentication may include a cloud authentication service system 100 and a blockchain authentication terminal 200 communicatively connected to the cloud authentication service system 100. The big data mining system 10 based on block chain security authentication shown in fig. 1 is only one possible example, and in other possible embodiments, the big data mining system 10 based on block chain security authentication may also only include at least part of the components shown in fig. 1 or may also include other components.

In a possible design idea, the cloud authentication service system 100 and the blockchain authentication terminal 200 in the big data mining system 10 based on blockchain security authentication may cooperatively perform the big data mining method based on blockchain security authentication described in the following method embodiments, and the detailed description of the method embodiments may be referred to in the following steps of the cloud authentication service system 100 and the blockchain authentication terminal 200.

In order to solve the technical problem in the foregoing background art, fig. 2 is a schematic flowchart of a big data mining method based on block chain security authentication according to an embodiment of the present disclosure, where the big data mining method based on block chain security authentication according to the present embodiment may be executed by the cloud authentication service system 100 shown in fig. 1, and the big data mining method based on block chain security authentication is described in detail below.

Step S110, after the dynamic bidirectional authentication of the blockchain authentication terminal 200 and the cloud authentication service system 100 is passed, service big data, which is obtained by allocating a corresponding target service acquisition process to the generated authentication service transmission channel to perform big data acquisition on the authentication service transmission channel, is acquired.

Step S120, obtaining data of the service big data under the at least two service hot spot propagation tags, and determining service hot spot propagation characteristics of the service big data under the at least two service hot spot propagation tags according to the data of the service big data under the at least two service hot spot propagation tags.

The service big data refers to service data needing to count service propagation associated parameters; for example, the service big data may be payment service big data, live order service big data, or the like. The business big data can be counted by the installed application program, for example, the application program can be specifically a small program of various service terminals, a special application program, and the like.

The service hotspot propagation label is used for measuring the propagation type of service propagation associated parameters between service data; for example, the service hotspot propagation tag may specifically be an update service-operation service data tag, a service data-service data frequent item theme tag, an update service-operation service data frequent item theme tag, or the like; the updating service-operation business data label is used for describing whether the updating service covers the business data, the business data-business data frequent item theme label is used for describing the transmission linkage data occupied by the business data frequent item theme in the business data, and the updating service-operation business data frequent item theme label is used for describing whether the updating service covers the business data containing the business data frequent item theme. The data refers to data of the service big data under each service hotspot propagation label, for example, the data of the service big data under the update service-operation service data label refers to propagation linkage data between the update service and the service big data, for example, the update service covers the service big data, and the update service does not cover the service big data.

It should be noted that the service hotspot propagation tag may include other service hotspot propagation tags besides the listed update service-operation service data tag, service data-service data frequent item topic tag, and update service-operation service data frequent item topic tag, and the disclosure is not limited specifically. The at least two service hotspot propagation tags may be two or more service hotspot propagation tags, and the disclosure is not limited in this respect.

The service hotspot propagation characteristic refers to service data coding, and the service data coding is a real-value characteristic vector of a specified dimension.

For example, the cloud authentication service system 100 obtains service distribution of the service big data under the at least two service hotspot propagation tags according to the data of the service big data under the at least two service hotspot propagation tags; and extracting the theme mining characteristics of the service distribution to obtain the service hotspot propagation characteristics of the service big data under each service hotspot propagation label.

Step S130, obtaining service propagation associated parameters among the service big data according to service hot spot propagation characteristics of the service big data under at least two service hot spot propagation labels.

The service propagation correlation parameter is used for describing service propagation correlation among the service data; for example, the higher the service propagation association parameter between the business big data is, the higher the association of the topic linkage between the business big data is.

For example, the cloud authentication service system 100 determines service propagation associated parameters of two pairs of service big data under at least two service hotspot propagation labels according to service hotspot propagation characteristics of the service big data under the at least two service hotspot propagation labels; and determining the service propagation associated parameters between every two pieces of service big data according to the service propagation associated parameters of every two pieces of service big data under at least two service hotspot propagation labels, thereby obtaining the service propagation associated parameters between all the service big data.

Step S140, determining target topic linkage information of each service big data according to the service propagation association parameter between each service big data, and pushing information to the blockchain authentication terminal 200 based on the target topic linkage information.

The target topic linkage information refers to a service propagation topic which has a service propagation relation with each business big data; for example, the service propagation topic a1 and the service propagation topic a2, of which the service propagation associated parameter with each business big data is greater than 80%, and the service propagation relationship (such as a jump guide relationship, etc.) between the service propagation topic a1 and the service propagation topic a2, may obtain topic recommendation information of the service propagation relationship existing under the service propagation topic a1 and the service propagation topic a2 based on the service propagation relationship among the service propagation topic a1, the service propagation topic a2, the service propagation topic a1, and the service propagation topic a2, and push information to the block chain authentication terminal 200.

That is to say, the service propagation theme in which the service propagation associated parameter between the service big data is greater than the preset associated parameter may be obtained, the target theme linkage information of each service big data is determined based on each service propagation theme having the service propagation relationship, and the information is pushed to the blockchain authentication terminal 200 based on the target theme linkage information.

Based on the steps, the service hot spot propagation characteristics of the big service data under the at least two service hot spot propagation labels are determined according to the obtained big service data under the at least two service hot spot propagation labels, so that service propagation associated parameters among the big service data are obtained, and target subject linkage information of the big service data is determined based on the service propagation associated parameters among the big service data; the method and the device achieve the purpose of comprehensively determining the service propagation associated parameters among the large service data through a plurality of service hotspot propagation labels, so that the content service hotspot propagation among the service data is more accurate, the determination accuracy of the service propagation associated parameters among the service data is improved, and the determination accuracy of the theme linkage information is further improved. Meanwhile, based on the accurately determined theme linkage information, the information push is favorably realized.

In a possible design idea, for the data obtained in step S120 under the propagation labels of the service big data in at least two service hotspots, the following steps are specifically included:

step S121, collecting the service big data in the target service range and updating the data of the service subscription service big data.

The data of the service big data subscribed by the update service refers to historical record data of whether the update service covers the service big data.

For example, the cloud authentication service system 100 determines service data in a target service range from the locally cached historical service big data as service big data; collecting update service subscription data reported by a plurality of update service terminals, and determining data of big data of update service subscription services within a target service range from the update service subscription data; therefore, the method is beneficial to obtaining the service big data in the target service range and updating the specific information of the service subscription service big data.

For example, the cloud authentication service system 100 may screen out the service data in the target service range from the historical service data cached in the local database, collect the update service subscription data reported by the update service terminal as the service data to be counted, and count the data of the service data to be counted subscribed by the update service in the target service range from the update service subscription data.

Step S122, acquiring data of the service big data under at least two service hot spot propagation labels in the target service range according to the service big data in the target service range and the data of the service subscription service big data.

For example, the cloud authentication service system 100 may obtain a plurality of service hotspot propagation tags, and perform statistical analysis on the service big data in the target service range and the data of the updated service subscription service big data according to the service hotspot propagation tags to obtain data of the service big data in each service hotspot propagation tag in the target service range.

In the embodiment, the data of the service big data in the target service range under each service hotspot propagation label is obtained by performing statistical analysis on the service big data in the target service range and the data of the service big data subscribed by the update service; the method and the device achieve the purpose of acquiring the data of the large service data under the plurality of service hot spot propagation tags in the target service range, facilitate the subsequent analysis of service propagation associated parameters among the large service data from the plurality of service hot spot propagation tags, and further improve the accuracy of determining the theme linkage information.

For example, in one possible design approach, if the service hotspot propagation tag includes a first service hotspot propagation tag, a second service hotspot propagation tag, and a third service hotspot propagation tag; then, in step S122, according to the service big data in the target service range and the data of the service subscription service big data, data of the service big data under each service hotspot propagation label in the target service range is obtained, which specifically includes the following steps:

step S1221, according to the data of the service big data subscribed by the update service, determining the transmission linkage data between the update service and the service big data in the target service range, and using the transmission linkage data as the data of the service big data under the first service hotspot transmission label.

The first service hotspot propagation tag is an update service-operation service data tag and is used for describing propagation linkage data between the update service and the service big data, namely, whether the update service covers the service big data or not is represented, and the relevance of the service big data in the update service type is reflected; the data of the service big data under the first service hotspot propagation label can be represented by an array (i, j), and the array (i, j) represents whether the update service i covers the service big data j; if the update service i already covers the service big data j, the array value corresponding to the array (i, j) is equal to 1; if the update service i does not cover the service big data j, the array value corresponding to the array (i, j) is equal to 0.

For example, the cloud authentication service system 100 analyzes data of the update service subscription service big data in the target service range to determine whether the update service covers the service big data, so as to obtain propagation linkage data between the update service and the service big data; and mapping the transmission linkage data between the update service and the service big data into a corresponding array, and taking the array as the data of the service big data under the first service hotspot transmission label in a target service range.

For example, the cloud authentication service system 100 represents an update service by i, represents service data to be counted by j, and counts a relationship between the update service i and the service data to be counted by j; and mapping the relation between the updated service i and the service data j to be counted into a corresponding array (i, j), and taking the array (i, j) as the data of the service data to be counted under the first service hotspot propagation label in a target service range.

Step S1222, extracting frequent item topics from the business big data.

The frequent item theme refers to frequent item information used for describing business big data; for example, in a service big data describing a certain payment process, the frequent theme refers to a payment object, a payment scene, and the like. Each business big data comprises one or more frequent item topics.

And step S1223, acquiring the propagation linkage data of the frequent item theme in the service big data in the target service range, and identifying the propagation linkage data as the data of the service big data under the second service hotspot propagation label.

The second service hotspot propagation tag is a service data-service data frequent item theme tag, is used for describing propagation linkage data occupied by a service data frequent item theme in service big data, and reflects the relevance of the service big data in the aspect of content semantics; data of the service big data under the second service hotspot propagation label can be represented by an array (i, j), and the array (i, j) represents propagation linkage data of a frequent theme j in the service big data i; if the propagation linkage data of the frequent theme j in the service big data i is k, the array value corresponding to the array (i, j) is equal to k; and if the propagation linkage data of the frequent theme j in the service big data i is 0, the array value corresponding to the array (i, j) is equal to 0.

For example, the cloud authentication service system 100 analyzes the business big data in the target business range to count the propagation linkage data of the extracted frequent theme in the business big data; and mapping the propagation linkage data occupied by the frequent item subject in the service big data into a corresponding array, and taking the array as the data of the service big data under the second service hotspot propagation label in a target service range.

For example, the cloud authentication service system 100 represents the service data to be counted by i, represents the frequent theme by j, and counts the propagation linkage data occupied by the frequent theme j in the service data to be counted i; and mapping the propagation linkage data occupied in the frequent theme j in the service data i to be counted into a corresponding array (i, j), and taking the array (i, j) as data of the service data to be counted under the second service hotspot propagation label in a target service range.

Step S1224, according to the propagation linkage data between the updated service and the service big data and the propagation linkage data of the frequent theme in the service big data in the target service range, determining the propagation linkage data of the updated service and the frequent theme in the service big data as the data of the service big data under the third service hotspot propagation tag.

The third service hotspot propagation tag is an update service-operation service data frequent item theme tag and is used for describing propagation linkage data of the update service and a frequent item theme in service big data, namely, whether the update service covers the service big data containing the frequent item theme or not is represented, and the relevance of the frequent item theme in the update service type is reflected; the data of the business big data under the third service hotspot propagation label can be represented by an array (i, j), and the array (i, j) represents whether the update service i covers the business big data containing the frequent item topic j; if the updating service i covers the business big data containing the frequent item theme j, the array value corresponding to the array (i, j) is equal to the propagation linkage data occupied by the frequent item theme j in the covered business big data; if the update service i does not cover the business big data containing the frequent item subject j, the array value corresponding to the array (i, j) is equal to 0.

It should be noted that, if the propagation linkage data occupied by the corresponding frequent item topic j is different in the service big data covered by the update service, the maximum value of the propagation linkage data occupied by the frequent item topic j is adopted as the array value corresponding to the array (i, j).

For example, the cloud authentication service system 100 analyzes the propagation linkage data between the update service and the business big data in the target business range and the propagation linkage data of the frequent theme in the business big data to determine whether the update service covers the business big data containing the frequent theme, so as to obtain the propagation linkage data of the update service and the frequent theme in the business big data; and mapping the updated service and the propagation linkage data of the frequent theme in the service big data into a corresponding array, and taking the array as the data of the service big data under the propagation label of the third service hotspot in a target service range.

For example, the cloud authentication service system 100 represents an update service by i, represents a frequent theme by j, and counts propagation linkage data of the update service i and the frequent theme j in the service data to be counted; and mapping the propagation linkage data of the updated service i and the frequent item subject j in the service data to be counted into a corresponding array (i, j), and taking the array (i, j) as data of the service data to be counted under a third service hotspot propagation label in a target service range.

In this embodiment, by analyzing the service big data in the target service range and the data of the update service subscription service big data, the data of the service big data under each service hotspot propagation label in the target service range can be obtained; the method is beneficial to analyzing the service propagation associated parameters among the service big data from a plurality of service hotspot propagation labels subsequently, and further improves the accuracy rate of determining the theme linkage information.

In a possible design idea, the step S120 determines service hotspot propagation characteristics of the service big data under the at least two service hotspot propagation labels according to the data of the service big data under the at least two service hotspot propagation labels, and specifically may include the following steps:

step S123, constructing service distribution of the service big data under the at least two service hot spot propagation tags according to the data of the service big data under the at least two service hot spot propagation tags.

The service distribution refers to service distribution formed by array values of an array corresponding to data of service big data under a service hotspot propagation label; for example, under the label of update service-operation service data, if the update service i is used as a horizontal array and the service big data j is used as a vertical array, the service distribution obtained based on the array (i, j) is | (i, j) |.

For example, the cloud authentication service system 100 analyzes data of the service big data under at least two service hotspot propagation labels to obtain an array of the service big data under at least two service hotspot propagation labels; the method comprises the steps of obtaining array values corresponding to arrays of the large service data under the same service hotspot propagation label, combining the array values to obtain service distribution of the large service data under the service hotspot propagation label, and thus obtaining the service distribution of the large service data under at least two service hotspot propagation labels.

For example, under the update service-operation traffic data label, if the update service 1 covers the traffic data 1 and does not cover the traffic data 2, and the update service 2 covers the traffic data 2 and does not cover the traffic data 1, then (1,1) ═ 1, (1,2) ═ 0, (2,1) ═ 0, and (2,2) ═ 1, then the traffic distribution composed of (1,1), (1,2), (2,1), (2,2) is: the service distribution is identified as service distribution of the service data 1 and the service data 2 under the update service-operation service data label. It should be noted that the process of service distribution in the corresponding type based on the data composition under the service data-service data frequent item theme label or the update service-operation service data frequent item theme label is the same as the above, and is not described herein again.

Step S124, performing theme mining feature extraction on the service distribution to obtain service hotspot propagation features of the service big data under at least two service hotspot propagation labels.

For example, considering the traffic distribution of the traffic big data under each service hotspot propagation label, the traffic big data is composed of array values of an array composed of two objects; then, the cloud authentication service system 100 performs service distribution topic mining feature extraction on service distribution of the service big data under each service hotspot propagation label respectively to obtain a plurality of object codes of the service big data under each service hotspot propagation label, screens out service hotspot propagation features from the plurality of object codes or obtains the service hotspot propagation features by processing the object codes, and uses the service hotspot propagation features as service hotspot propagation features of the service big data under corresponding service hotspot propagation labels.

For example, under the update service-operation service data label, the obtained object code is an update service description and a service data code, and both the update service description and the service data code are real-valued vectors with a specified dimension; and the service data codes are used as service hotspot propagation characteristics of the service big data under the service updating-operation service data label. Under the service updating-operation business data frequent item theme label, the obtained object codes are an updating service description and a business data frequent item theme description, and the updating service description and the business data frequent item theme description are real-valued vectors with a specified dimensionality; and performing weighting processing on the service data frequent item theme description to obtain a corresponding service data code, and taking the service data code as a service hotspot propagation characteristic of service big data under the service updating-operation service data frequent item theme label.

In this embodiment, by analyzing the data of the service big data under the at least two service hot spot propagation tags, the service hot spot propagation characteristics of the service big data under the at least two service hot spot propagation tags can be obtained, so that the service propagation associated parameters between the service big data can be obtained subsequently based on the service hot spot propagation characteristics.

In a possible design idea, in step S123, according to data of the service big data under the propagation tags of the at least two service hotspots, a service distribution of the service big data under the propagation tags of the at least two service hotspots is constructed, which specifically includes the following steps:

step S1231, constructing a first service distribution, a second service distribution and a third service distribution according to the data of the service big data under the at least two service hot spot propagation labels; the first business distribution is used for representing the relationship between the service description and the service hotspot propagation characteristics, the second business distribution is used for representing the relationship between the service hotspot propagation characteristics and the frequent item topic description, and the third business distribution is used for representing the relationship between the service description and the frequent item topic description.

For example, the cloud authentication service system 100 constructs a first service distribution according to data of the service big data under the first service hotspot propagation label; constructing a second service distribution according to the data of the service big data under the second service hotspot propagation label; and constructing a third service distribution according to the data of the service big data under the third service hotspot propagation label.

Step S1232, respectively identifying the first service distribution, the second service distribution, and the third service distribution as service distributions of the service big data under at least two service hotspot propagation labels.

In this embodiment, according to the data of the service big data under the at least two service hotspot propagation tags, the service distribution of the service big data under the at least two service hotspot propagation tags is constructed, which is beneficial to performing topic mining feature extraction on the service distribution subsequently, so as to obtain the service hotspot propagation features of the service big data under the at least two service hotspot propagation tags.

In a possible design idea, if the service distribution is the first service distribution or the second service distribution, then, in step S1232, the subject mining feature extraction is performed on the service distribution to obtain service hotspot propagation features of the service big data under at least two service hotspot propagation labels, which includes: and extracting the theme mining characteristics of the service distribution to obtain service hotspot propagation characteristics which are used as service hotspot propagation characteristics of the service big data under the corresponding service hotspot propagation label.

For example, the cloud authentication service system 100 performs topic mining feature extraction on the first service distribution to obtain an updated service description and a service hotspot propagation feature, and uses the service hotspot propagation feature as a service hotspot propagation feature of the service big data under the first service hotspot propagation label; and performing theme mining feature extraction on the second service distribution to obtain service hotspot propagation features and frequent theme descriptions, and taking the service hotspot propagation features as service hotspot propagation features of the service big data under a second service hotspot propagation label. Therefore, the purpose of obtaining the service hotspot propagation characteristics of the service big data under the first service hotspot propagation label and the second service hotspot propagation label is achieved.

In another possible design idea, if the service distribution is the third service distribution, then, in step S1232, the subject mining feature extraction is performed on the service distribution to obtain service hotspot propagation features of the service big data under at least two service hotspot propagation labels, which specifically includes the following steps:

(1) and performing theme mining feature extraction on the service distribution to obtain frequent item theme description in the service big data.

For example, the cloud authentication service system 100 performs topic mining feature extraction on the third business distribution to obtain an updated service description and a frequent topic description in the business big data.

(2) And acquiring the propagation linkage data of the frequent item theme in the service big data as the propagation linkage data corresponding to the frequent item theme description.

(3) And obtaining service hotspot propagation characteristics according to the frequent theme description and the corresponding propagation linkage data, wherein the service hotspot propagation characteristics are used as service hotspot propagation characteristics of the service big data under the corresponding service hotspot propagation labels.

For example, the service big data includes a frequent item topic 1, a frequent item topic 2, and a frequent item topic 3, propagation linkage data corresponding to the frequent item topic 1, the frequent item topic 2, and the frequent item topic 3 are w1, w2, and w3, respectively, and descriptions of the corresponding frequent item topics are v1, v2, and v3, respectively, and then service hotspot propagation characteristics of the service big data under a corresponding service hotspot propagation label are as follows: w1 xv 1+ w2 xv 2+ w3 xv 3.

In this embodiment, under the condition that the service distribution is the third service distribution, the service hotspot propagation characteristics corresponding to the service big data can be obtained by performing weighted combination on the frequent item theme descriptions extracted from the service distribution theme mining characteristics, so that the purpose of processing the frequent item theme descriptions corresponding to the service big data to obtain the service hotspot propagation characteristics corresponding to the single service big data is achieved.

In a possible design idea, the step S130 obtains service propagation associated parameters between the service big data according to service hot spot propagation characteristics of the service big data under at least two service hot spot propagation tags, and specifically includes the following steps:

step S131, calculating service propagation associated parameters of two pairs of service big data under at least two service hot spot propagation labels according to service hot spot propagation characteristics of the service big data under at least two service hot spot propagation labels.

The service propagation association parameter refers to cosine similarity between the service big data, is used for describing content association of any two service big data under the corresponding service hotspot propagation label, and can be represented by cosine similarity between the service big data.

For example, the service hotspot propagation characteristic of the service data 1 under the first service hotspot propagation label is v1, and the service hotspot propagation characteristic of the service data 2 under the first service hotspot propagation label is v2, then the service propagation associated parameters of the service data 1 and the service data 2 under the first service hotspot propagation label are:

it should be noted that the service propagation associated parameters of the two pieces of service big data under the propagation tags of other service hotspots may be calculated based on the above manner, and are not described herein again.

Step S132, obtaining service propagation associated parameters between each two pieces of service big data according to the service propagation associated parameters of each two pieces of service big data under at least two service hotspot propagation labels.

In this embodiment, the service propagation associated parameters between the service big data are obtained through the service propagation associated parameters of every two pieces of service big data under the at least two service hotspot propagation tags, and the purpose of calculating the service propagation associated parameters between the service big data through the plurality of service hotspot propagation tags is achieved, so that the determination accuracy of the service propagation associated parameters between the service data is improved, the subsequent accurate determination of the theme linkage information of the service big data is facilitated, and the determination accuracy of the theme linkage information is further improved.

In a possible design idea, the step S132 obtains service propagation associated parameters between the service big data according to the service propagation associated parameters of every two service big data under at least two service hotspot propagation labels, and specifically includes the following steps:

step S1321, service interaction contact information between the service propagation associated parameters is calculated according to the service propagation associated parameters of the two pieces of service big data under the at least two service hotspot propagation labels.

The service interaction contact information refers to a cross item of two service propagation associated parameters, for example, the service interaction contact information of the service propagation associated parameter a and the service propagation associated parameter b is a × b.

Step S1322 is to acquire the propagation linkage data of each service propagation associated parameter and the propagation linkage data of each service interaction contact information.

And the propagation linkage data of each service propagation associated parameter and the propagation linkage data of each service interaction contact information are obtained based on historical business big data training.

And step S1323, calculating to obtain service propagation associated parameters among the service big data according to the service propagation associated parameters, the corresponding propagation linkage data, the service interaction contact information and the corresponding propagation linkage data.

For example, the cloud authentication service system 100 calculates a product of each service propagation associated parameter and corresponding propagation linkage data and a product of each service interaction contact information and corresponding propagation linkage data, and adds the product of each service propagation associated parameter and corresponding propagation linkage data and the product of each service interaction contact information and corresponding propagation linkage data to obtain a service propagation associated parameter between every two service big data, thereby obtaining a service propagation associated parameter between each two service big data.

Further, in order to facilitate calculation of service propagation associated parameters between the service big data and improve accuracy of determination of the service propagation associated parameters between the service big data, service interaction contact information between the service propagation associated parameters is calculated according to the service propagation associated parameters of every two service big data under each service hotspot propagation label, and the method may further include: discretizing the service propagation associated parameters of the two pairs of service big data under the at least two service hotspot propagation labels to obtain service propagation associated parameter discrete values of the two pairs of service big data under the at least two service hotspot propagation labels; and calculating service interaction contact information discrete values among the service propagation associated parameter discrete values.

For example, service propagation associated parameters of the service data 1 and the service data 2 under the three service hotspot propagation labels are respectively 0.6, 0.7 and 0.5, and the rounding value of the service propagation associated parameter after the service propagation associated parameter is enlarged by 10 times is calculated through the following ten-equal discretization formula, so that the obtained service propagation associated parameter discrete values are respectively 6, 7 and 5. Then, the discrete value of service interaction contact information between the discrete value of service propagation associated parameter 6 and the discrete value of service propagation associated parameter 7 is 6 × 7.

For example, the ten-point discretization formula is as follows:

round(max(X×10,0))；

wherein, X is the service propagation associated parameter, and round is the rounding operation. It should be noted that the discrete value of the service propagation associated parameter smaller than 0 is 0, and then the service propagation associated parameters of the two service data under one service hotspot propagation label have 11 values, which are 0 to 10 respectively.

In another possible design idea, in step S132, the service propagation associated parameter between the service big data is obtained according to the service propagation associated parameter of every two service big data under the at least two service hotspot propagation labels, and the method specifically includes the following steps:

(1) and inputting service propagation associated parameters of the two pairs of business big data under at least two service hotspot propagation labels into a pre-trained machine learning network.

Wherein, the machine learning network refers to a factorization model taking machine; compared with a linear regression model, the method introduces a binary cross term, and considers the influence of cross combination of two service propagation associated parameters on regression accuracy.

(2) Confirming service interaction contact information among service propagation associated parameters through a machine learning network; acquiring the propagation linkage data of each service propagation associated parameter and the propagation linkage data of each service interaction contact information; and calculating to obtain the service propagation associated parameters among the service big data according to the service propagation associated parameters, the corresponding propagation linkage data, the service interaction contact information and the corresponding propagation linkage data.

In this embodiment, the cloud authentication service system 100 inputs the service propagation associated parameters of every two pieces of service big data under each service hotspot propagation label into a machine learning network trained in advance, obtains the service propagation associated parameters between each piece of service big data through calculation of the machine learning network, fully considers the service interaction contact information of the service propagation associated parameters, and further improves the determination accuracy of the service propagation associated parameters between the pieces of service big data.

In one possible design approach, the machine learning network is trained by:

(1) and acquiring service propagation associated parameters of the marked samples under at least two service hotspot propagation labels, corresponding first propagation linkage data, service interaction contact information among the service propagation associated parameters, corresponding second propagation linkage data and actual service propagation associated parameters among the marked samples.

The marked sample refers to historical business data, and the actual service propagation associated parameter can be represented by the true co-occurrence (i.e. business tendency value) of the two business data in the current business range. The service tendency value may be represented by a support degree, for example, in 1000 service data sets obtained through statistics, if the service data 1 and the service data 2 are propagated and linked together 800 times, the service tendency value is 800/1000 ═ 0.8, then the service tendency value of the two service data in the current time period is 0.8, that is, the actual service propagation associated parameter between the two service data is 0.8.

(2) And training the machine learning network according to the service propagation associated parameters of the marked samples under the at least two service hotspot propagation labels, the corresponding first propagation linkage data, the service interaction contact information among the service propagation associated parameters and the corresponding second propagation linkage data to obtain the trained machine learning network.

For example, the cloud authentication service system 100 trains the machine learning network through service propagation associated parameters of different labeled samples under at least two service hotspot propagation labels, corresponding first propagation linkage data, service interaction contact information between the service propagation associated parameters, and corresponding second propagation linkage data.

(3) And obtaining loss parameter distribution between the service propagation associated parameters output by the trained machine learning network and the corresponding actual service propagation associated parameters.

(4) And when the loss parameter distribution is larger than or equal to the preset threshold, adjusting the first transmission linkage data and the second transmission linkage data according to the loss parameter distribution, and repeatedly training the machine learning network according to the adjusted first transmission linkage data and second transmission linkage data until the loss parameter distribution obtained according to the trained machine learning network is smaller than the preset threshold.

For example, when the test error is greater than or equal to the preset threshold, the cloud authentication service system 100 continuously adjusts the first propagation linkage data and the second propagation linkage data in the machine learning network to repeatedly train the machine learning network until the loss parameter distribution obtained according to the trained machine learning network is less than the preset threshold, and uses the current machine learning network as the trained machine learning network.

In this embodiment, the cloud authentication service system 100 may improve the accuracy of service propagation associated parameters between the service data predicted by the machine learning network by repeatedly training the machine learning network, and further improve the determination accuracy of the service propagation associated parameters between the service data.

In one possible design approach, step S110 can be implemented by the following exemplary specific steps, which are described in detail below.

Step S111, obtain a first authentication event feature distribution of the target authentication event in the first service dynamic environment.

In one possible design example, the target authentication event may correspond to different authentication event feature distributions in different business dynamic environments, where the business dynamic environments may specifically be the business application services, and the target authentication event may specifically be represented by a business authentication resource of the target authentication event in the business application services at this time, that is, when the target authentication event uses different business application services, the feature distributions of the features used for representing the target authentication event in the business application services are different, where the authentication event feature distribution may specifically be a feature distribution representation of an encryption invocation feature of the authentication event in the business dynamic environments, and when the business dynamic environments are the business application services, the encryption invocation feature may specifically be an access encryption invocation feature of the authentication event to each encryption policy node in the business application services, and specifically may be based on access encryption times, access times, and access encryption times of the authentication event to different types of objects, Access encryption protocol, etc. The first authentication event feature distribution is feature distribution used for representing encryption calling features of the target authentication event in the first service dynamic environment, and the first authentication event feature distribution can be specifically used for subsequently determining the encryption calling feature degrees of the authentication event for each encryption policy node to be called by the encryption policy in the first service dynamic environment, and calling the encryption policy for the object with high encryption calling feature degree in the first service dynamic environment.

The first service dynamic environment may specifically refer to any one of the service dynamic environments, and is used to form a distinction with the second service dynamic environment, in the first service dynamic environment, the first authentication event feature distribution is specifically used to represent a target authentication event, and the cloud authentication service system 100 may obtain the first authentication event feature distribution of the target authentication event in the first service dynamic environment, for example, the first service dynamic environment is a first service application service, and the cloud authentication service system 100 may obtain the first authentication event feature distribution corresponding to the target authentication event from a background server of the first service application service, where the service application service may be an online new media application service, a payment application service, a live application service, and the like.

Step S112, a policy decision network is invoked to make a decision on the first authentication event feature distribution, so as to obtain a second authentication event feature distribution of the target authentication event in the second service dynamic environment.

In a possible design example, after obtaining the first authentication event feature distribution of the target authentication event in the first business dynamic environment, the cloud authentication service system 100 may invoke a policy decision network to make a decision on the first authentication event feature distribution, so as to obtain the second authentication event feature distribution of the target authentication event in the second business dynamic environment.

In a possible design example, the target authentication event may have no or only a small amount of service dynamic associated data in the second service dynamic environment, that is, data analysis cannot be performed based on access information of the target authentication event in the second service dynamic environment, so that the authentication event feature distribution of the target authentication event in the second service dynamic environment is accurately determined. At this time, the cloud authentication service system 100 may invoke a policy decision network to make a decision on the first authentication event feature distribution of the target authentication event in the first business dynamic environment, so as to obtain the second authentication event feature distribution of the target authentication event in the second business dynamic environment. The second authentication event feature distribution is specifically used for representing the encryption calling feature of the target authentication event in the second service dynamic environment, and for subsequently determining the encryption calling feature degree of the authentication event for each encryption policy node to be called by the encryption policy in the second service dynamic environment, and performing encryption policy calling on the object with high encryption calling feature degree in the second service dynamic environment.

In one possible design example, if the target authentication event has no service dynamic associated data in the second service dynamic environment, the cloud authentication service system 100 may invoke the policy decision network to make a decision on the first authentication event feature distribution, and use the feature distribution obtained by the decision as the second authentication event feature distribution of the target authentication event in the second service dynamic environment. In a possible design example, if the target authentication event has dynamic service association data in the second dynamic service environment, the cloud authentication service system 100 may invoke a policy decision network to make a decision on the feature distribution of the first authentication event to obtain a third authentication event feature distribution, and obtain a fourth authentication event feature distribution obtained based on the decision on the dynamic service association data, and the cloud authentication service system 100 fuses the third authentication event feature distribution and the fourth authentication event feature distribution to obtain a second authentication event feature distribution of the target authentication event in the second dynamic service environment. The Fusion (Fusion) refers to combining together, and the Fusion mode may be referred to a common Fusion algorithm in the prior art, which is not limited herein. For example, at least one of a concatenation or a weighted addition. For example, the splicing may be performed, and the result obtained by the splicing is input into a fusion model for processing, where the fusion model may be, for example, a multilayer perceptron model, a recurrent neural network model, or a convolutional neural network model. For example, the fusion model may include P aggregators, each aggregator mixes two different heterogeneous features (i.e., target text encoding vectors and target knowledge representation vectors corresponding to target entities) through MLP (multi-layer perceptron), and the number of P may be set as required.

It should be noted that the policy decision network is obtained by training a target policy decision network through example feature distribution, where the example feature distribution includes a first example feature distribution of an example authentication event list in a first business dynamic environment and a target example feature distribution in a second business dynamic environment; the target policy decision network is obtained by training an initial policy decision network through M target feature distributions, wherein the M target feature distributions comprise target feature distributions of a collection authentication event list in each service dynamic environment of M service dynamic environments, and M is a positive integer. The training process of the target strategy decision network comprises iterative updating of network propagation linkage data information of the target strategy decision network for multiple times, each iterative updating process of the network propagation linkage data information is a decision process of the network propagation linkage data information of the target strategy decision network and N groups of updating reference parameters, the updating reference parameters are obtained by training based on a group of sub-example authentication event feature distribution sequences, each group of sub-example authentication event feature distribution sequences are feature distributions corresponding to any group of sub-example authentication event lists in the N groups of sub-example authentication event lists, and the N groups of sub-example authentication event lists are obtained by searching based on the example authentication event lists.

It should be noted that, in the training process of the policy decision network, specifically, the cloud authentication service system 100 obtains an example authentication event list, finds N groups of sub-example authentication event lists from the example authentication event list, and determines, from the example feature distribution, a feature distribution corresponding to each group of sub-example authentication event lists, where the feature distribution corresponding to each group of sub-example authentication event lists includes a first sub-example feature distribution of each group of sub-example authentication event lists in a first business dynamic environment and a target sub-example feature distribution of each group of sub-example authentication event lists in a second business dynamic environment; the cloud authentication service system 100 performs network configuration on the target policy decision network based on the feature distribution corresponding to each group of sub-example authentication event lists to obtain N update reference parameters; updating network propagation linkage data information in the target policy decision network based on the N updating reference parameters; and if the target policy decision network after the network transmission linkage data information is updated meets the preset condition, determining the target policy decision network after the network transmission linkage data information is updated as the policy decision network, wherein the preset condition comprises that the policy decision precision of each feature distribution in the feature distributions corresponding to each group of sub-example authentication event lists of the target policy decision network is higher than the preset precision.

In one possible design example, the way for the cloud authentication service system 100 to perform network configuration on the target policy decision network based on the feature distribution corresponding to any one target sub-example authentication event list in the N groups of sub-example authentication event lists includes: the cloud authentication service system 100 splits a target sub-example authentication event list to obtain a first target sub-example authentication event list and a second target sub-example authentication event list, the target sub-example authentication event list is any one group of sub-example authentication event lists in N groups of sub-example authentication event lists, the cloud authentication service system 100 obtains a first feature distribution corresponding to the first target sub-example authentication event list and a second feature distribution corresponding to the second target sub-example authentication event list, the cloud authentication service system 100 performs network configuration on a target policy decision network based on the first feature distribution to obtain a first adjustment reference parameter, and updates network propagation linkage data information of the target policy decision network from initial propagation linkage data information to target propagation linkage data information based on the first adjustment reference parameter; performing network configuration on the target policy decision network after the network transmission linkage data information is updated to the target transmission linkage data information based on the second characteristic distribution to obtain a second adjustment reference parameter; the cloud authentication service system 100 determines a target update reference parameter corresponding to the target sub-example authentication event list based on the second adjustment reference parameter and the initial propagation linkage data information.

The first characteristic distribution comprises a first target authentication event characteristic distribution sequence of the first target sub-example authentication event list in the first business dynamic environment and a third authentication event characteristic distribution sequence of the first target sub-example authentication event list in the second business dynamic environment; the cloud authentication service system 100 performs network configuration on the target policy decision network based on the first feature distribution, and a specific manner of obtaining the first adjustment reference parameter may be that the cloud authentication service system 100 calls the target policy decision network to make a decision on the first target authentication event feature distribution sequence to obtain a first decision example feature distribution, calculates a mean square error between the first decision example feature distribution and the third authentication event feature distribution sequence, and uses the mean square error as the first adjustment reference parameter.

The second profile includes a second target authentication event profile sequence of the second target sub-example authentication event manifest in the first business dynamic environment, and a fourth authentication event feature distribution sequence in the second service dynamic environment, the cloud authentication service system 100 performs network configuration on the target policy decision network after updating the network propagation linkage data information to the target propagation linkage data information based on the second feature distribution to obtain the second adjustment reference parameter in such a manner that the cloud authentication service system 100 calls the target policy decision network after updating the parameter to make a decision on the second target authentication event feature distribution sequence to obtain a second decision example feature distribution, and calculating a mean square error between the second decision example feature distribution and the fourth authentication event feature distribution sequence, and taking the mean square error as a second adjustment reference parameter.

The specific way for the cloud authentication service system 100 to determine the target update reference parameter corresponding to the target sub-example authentication event list based on the second adjustment reference parameter and the initial propagation linkage data information may be that the cloud authentication service system 100 performs second-order partial derivation processing on the initial propagation linkage data information based on the second adjustment reference parameter to obtain the target update reference parameter corresponding to the target sub-example authentication event list, where the target update reference parameter is one of N update reference parameters for updating the network propagation linkage data information in the target policy decision network.

The specific way of updating the network propagation linkage data information in the target policy decision network by the cloud authentication service system 100 based on the N update reference parameters may be that the cloud authentication service system 100 fuses the N update reference parameters to obtain a fusion reference parameter; and acquiring propagation linkage data corresponding to the fusion reference parameter, and performing weighting processing on the fusion reference parameter by using the propagation linkage data to obtain a weighted reference parameter, and updating network propagation linkage data information in the target policy decision network from initial propagation linkage data information to a difference value between the initial propagation linkage data information and the weighted reference parameter by the cloud authentication service system 100. In the above manner, different combination manners are adopted to combine the example authentication events to obtain a plurality of sub-example authentication event lists, and the sub-example authentication event lists are used as training units to complete the updating of the network transmission linkage data information, so that the number of example data can be increased, the example expansion is realized, and the problem that the example data with strong correlation is less in the network configuration process is solved.

It should be further noted that, the specific way in which the cloud authentication service system 100 performs network configuration on the initial policy decision network by collecting M target feature distributions of the authentication event list in M service dynamic environments to obtain a target initialization network may be that the cloud authentication service system 100 obtains the collected authentication event list; determining target feature distribution of a collected authentication event list in each service dynamic environment in M service dynamic environments to obtain M target feature distributions; the cloud authentication service system 100 combines the M target feature distributions to obtain M feature distribution combinations, each feature distribution combination includes any two groups of target feature distributions in the M target feature distributions, and performs network configuration on the initial policy decision network through each feature distribution combination in the M feature distribution combinations to obtain a target policy decision network, where M is a positive integer.

In a possible design example, the cloud authentication service system 100 performs network configuration on the initial policy decision network through each feature distribution combination in M feature distribution combinations to obtain a target policy decision network, where the cloud authentication service system 100 performs network configuration on the initial policy decision network through a first feature distribution combination in the M feature distribution combinations to update parameters in the initial policy decision network, the first feature distribution combination includes a first target feature distribution and a first pair feature distribution, the first target feature distribution is feature distribution for collecting an authentication event list in a first training service dynamic environment, and the first pair feature distribution is feature distribution for collecting an authentication event list in a first test service dynamic environment; if the first strategy decision precision of the initial strategy decision network after the parameter updating for the first target feature distribution is higher than the preset precision, performing network configuration on the initial strategy decision network after the parameter updating through a second feature distribution combination in the M feature distribution combinations to obtain a first initial strategy decision network, wherein the first strategy decision precision is determined by a correlation parameter between the first decision feature distribution obtained by the initial strategy decision network making a decision for the first target feature distribution and the first pair feature distribution, the second feature distribution combination comprises the second target feature distribution and a second pair bit feature distribution, the second target feature distribution is the feature distribution of a collection authentication event list in a second training service dynamic environment, and the second pair bit distribution is the feature distribution of the collection authentication event list in a second testing service dynamic environment; and if the second strategy decision precision of the first initial strategy decision network on the second target characteristic distribution is higher than the preset precision, determining the first initial strategy decision network as the target strategy decision network, wherein the second strategy decision precision is determined by a correlation parameter between the second decision characteristic distribution and the second alignment characteristic distribution, which are obtained by the decision of the first initial strategy decision network on the second target characteristic distribution. The correlation parameter between the feature distributions may be specifically determined by a mean square error between the feature distributions, and the correlation parameter specifically decreases as the mean square error increases. It should be noted that the number of feature distributions in the second feature distribution combination may be much lower than the number of examples in the first feature distribution combination, and whether the model has better migration capability may be verified by the above-described secondary training method, that is, whether the initial policy decision network after parameter update is configured by using a small number of examples, so that the model may also achieve higher policy decision accuracy.

In practical application, the training of the M feature distribution combinations may be completed in the above manner, for example, after it is determined that the second policy decision precision of the first initial policy decision network on the second target feature distribution is higher than the preset precision, the first initial policy decision network may be configured through the third feature distribution combination of the M feature distribution combinations, so as to obtain the first initial policy decision network with updated parameters. The above manner can make part of parameters in the initial policy decision network to be more optimal.

The method comprises the steps that in the process of configuring an initial network through any one of M characteristic distribution combinations, iterative updating of network transmission linkage data information of the initial strategy decision network is carried out for multiple times, each iterative updating process of the network transmission linkage data information is carried out for the network transmission linkage data information of the initial strategy decision network and N groups of updating reference parameter decision processes, updating reference parameters are obtained by training based on a group of sub-collection authentication event characteristic distribution sequences, each group of sub-collection authentication event characteristic distribution sequences are characteristic distributions corresponding to any one group of sub-collection authentication event lists in the N groups of sub-collection authentication event lists, and the N groups of sub-collection authentication event lists are obtained by searching based on the collection authentication event lists.

In step S113, the cloud authentication service system 100 obtains encryption policy invocation information for the target authentication event according to the second authentication event feature distribution, and invokes the encryption policy for the target authentication event in the second service dynamic environment.

In one possible design example, after the cloud authentication service system 100 calls the policy decision network to make a decision on the first authentication event feature distribution to obtain a second authentication event feature distribution of the target authentication event in the second service dynamic environment, the cloud authentication service system may obtain encryption policy call information for the target authentication event according to the second authentication event feature distribution, and call the encryption policy call information in the second service dynamic environment.

In a possible design example, the specific way for the cloud authentication service system 100 to obtain the second authentication event feature distribution of the target authentication event in the second service dynamic environment may be that the cloud authentication service system 100 obtains the encryption policy parameters of each encryption policy node in the encryption policy node set called by the policy to be encrypted in the second service dynamic environment, and determines the encryption policy calling index for each encryption policy node in the encryption policy node set called by the policy to be encrypted based on the second authentication event feature distribution and each encryption policy parameter; searching out at least one target encryption strategy node from an encryption strategy node set to be encrypted strategy called based on the encryption strategy calling index of each encryption strategy node as encryption strategy calling information aiming at a target authentication event; and determining a target encryption policy calling mode for the target encryption policy node in the encryption policy calling information based on the corresponding relationship between the encryption policy calling index and the encryption policy calling mode, and the cloud authentication service system 100 calls the encryption policy for the target encryption policy node in the second service dynamic environment based on the target encryption policy calling mode. The encryption policy calling mode comprises at least one of an encryption policy calling sequence, an encryption policy calling time and an encryption policy calling frequency, and the corresponding relation between the encryption policy calling index and the encryption policy calling mode can be that the higher the encryption policy calling index is, the earlier the encryption policy calling sequence is, the longer the encryption policy calling time is and the higher the encryption policy calling frequency is.

It should be noted that, a specific determination manner of the encryption policy invocation index of each encryption policy node in the encryption policy node set to be invoked by the encryption policy may be that the cloud authentication service system 100 performs a dot product decision on the second authentication event feature distribution and each encryption policy parameter, so as to obtain a dot product value corresponding to each encryption policy node; and determining the dot product value corresponding to each encryption strategy node as an encryption strategy calling index of each encryption strategy node in the encryption strategy node set called by the strategy to be encrypted. By the method, the corresponding information can be more intelligently called for the authentication event encryption strategy, so that the information called by the encryption strategy is more matched with the authentication event encryption calling characteristics.

In one possible design example, the cloud authentication service system 100 obtains encryption policy invocation information for a target authentication event according to second authentication event feature distribution, and obtains reference authentication event feature distribution of the target authentication event in a reference service dynamic environment before the encryption policy invocation information is subjected to encryption policy invocation in a second service dynamic environment, the cloud authentication service system 100 invokes a reference policy decision network to make a decision on the reference authentication event feature distribution to obtain third authentication event feature distribution of the target authentication event in the second service dynamic environment, and the reference policy decision network is obtained by training a reference example feature distribution of an example authentication event list in the reference service dynamic environment and a target example feature distribution in the second service dynamic environment to a target policy decision network; and if the correlation parameter between the third authentication event characteristic distribution and the second authentication event characteristic distribution is greater than the preset correlation parameter, acquiring encryption strategy calling information for the target authentication event according to the second authentication event characteristic distribution. In one possible design example, if a correlation parameter between the third authentication event feature distribution and the second authentication event feature distribution is smaller than a preset correlation parameter, the third authentication event feature distribution and the second authentication event feature distribution are fused to obtain a target feature distribution, and encryption policy invocation information is obtained for the target authentication event based on the target feature distribution.

Wherein, the reference service dynamic environment can be an existing service dynamic environment having relevance with the second service dynamic environment, a consistency check mode for the second authentication event feature distribution is provided by checking a third authentication event feature distribution obtained based on the reference service dynamic environment decision and a second authentication event feature distribution obtained based on the first service dynamic environment decision, when the correlation parameter between different feature distribution representations of the target authentication event obtained based on different service dynamic environment decisions in the second service dynamic environment is low, the confidence coefficient of the second authentication event feature distribution obtained based on the first service dynamic environment decision is determined to be lower, the second authentication event feature distribution needs to be obtained again by adopting other modes, when the correlation parameter between different feature distribution representations of the target authentication event obtained based on different service dynamic environment decisions in the second service dynamic environment is high, and determining that the confidence coefficient of the second authentication event feature distribution is higher, namely acquiring encryption strategy calling information for the target authentication event according to the second authentication event feature distribution, thereby realizing the confidence coefficient check of the second authentication event feature distribution.

In one possible design example, the cloud authentication service system 100 obtains a first authentication event feature distribution of a target authentication event in a first service dynamic environment, calls a policy decision network to make a decision on the first authentication event feature distribution, obtains a second authentication event feature distribution of the target authentication event in a second service dynamic environment, obtains encryption policy calling information for the target authentication event according to the second authentication event feature distribution, and calls the encryption policy calling information for an encryption policy in the second service dynamic environment. The training process of the strategy decision network comprises iterative updating of network transmission linkage data information for multiple times, the iterative updating process of the network transmission linkage data information for each time and the decision process of the network transmission linkage data information and N updating reference parameters of the strategy decision network, wherein one updating reference parameter is obtained by training based on a group of sub-example authentication event feature distribution sequences. The strategy decision network is obtained through a grouping iterative training mode, and an object for carrying out encryption strategy calling on the authentication event is determined based on the strategy decision network, so that the accuracy of information encryption strategy calling is improved.

Based on the above description, the embodiments of the present disclosure provide a method for training a policy decision network, and the policy decision network training process may be implemented by the following exemplary steps.

In step S210, the cloud authentication service system 100 acquires an example authentication event list.

In one possible design example, each example authentication event in the example authentication event list has service dynamic association data in both the first service dynamic environment and the second service dynamic environment, and the service dynamic environment may be a service application service, a service microservice service, or the like, and if the service dynamic environment is a service application service, each example authentication event in the example authentication event list uses both the first service application service and the second service application service, that is, the service dynamic association data is stored in both the first service application service and the second service application service. The cloud authentication service system 100 may acquire an authentication event list Us in a first business dynamic environment and an authentication event list Ut in a second business dynamic environment, and use an intersection between Us and Ut as an example authentication event list Uo.

In a possible design example, to further ensure that example authentication events are representative, after the cloud authentication service system 100 acquires an intersection between Us and Ut, the cloud authentication service system may search for the authentication events in the intersection based on a preset search condition, determine the authentication events in the intersection that satisfy the search condition as example authentication events, and construct an example authentication event list Uo based on each searched example authentication event. The search condition may specifically be that data amounts of the service dynamic association data in the first service dynamic environment and the second service dynamic environment are both greater than a preset threshold, and the service dynamic association data may specifically be a service dynamic access parameter, an access frequency, and the like to the service dynamic environment, and if the service dynamic environment is a service application service, the cloud authentication service system 100 determines an authentication event that the access frequency to the first service application service and the second service application service is both greater than a preset frequency as an example authentication event.

In step S220, the cloud authentication service system 100 finds N groups of sub-example authentication event lists from the example authentication event list, and determines a feature distribution corresponding to each group of sub-example authentication event lists from the example feature distribution.

In one possible design example, after the cloud authentication service system 100 obtains the example authentication event list, N groups of sub-example authentication event lists may be found from the example authentication event list, and the feature distribution corresponding to each group of sub-example authentication event lists may be determined from the example feature distribution. The feature distribution corresponding to each group of the sub-example authentication event lists comprises a first sub-example feature distribution of each group of the sub-example authentication event lists in a first business dynamic environment and a target sub-example feature distribution in a second business dynamic environment. In one possible design example, for the example authentication event manifest Uo, the cloud authentication service system 100 looks up N child example authentication event manifests { U1, … … Un } from the example authentication event manifest Uo.

In step S230, the cloud authentication service system 100 performs network configuration on the target policy decision network based on the feature distribution corresponding to each group of sub-example authentication event lists, so as to obtain N update reference parameters.

In a possible design example, after the cloud authentication service system 100 determines the feature distribution corresponding to each group of sub-example authentication event lists from the example feature distribution, the target policy decision network is configured on the basis of the feature distribution corresponding to each group of sub-example authentication event lists, so as to obtain N update reference parameters.

In a possible design example, the cloud authentication service system 100 determines the update reference parameter corresponding to each group of sub-example authentication event lists in the N sub-example authentication event lists in the same manner, and specifically, a manner of determining the update reference parameter ω i corresponding to any target sub-example authentication event list Ui in the N sub-example authentication event lists is described in detail below. The cloud authentication service system 100 splits the target sub-example authentication event list Ui to obtain a first target sub-example authentication event list Ua and a second target sub-example authentication event list Ub, and obtains a first feature distribution Da corresponding to the first target sub-example authentication event list and a second feature distribution Db corresponding to the second target sub-example authentication event list, the cloud authentication service system 100 performs network configuration on the target policy decision network based on the first feature distribution Da to obtain a first adjustment reference parameter L θ, and updates the network propagation linkage data information of the target policy decision network from the initial propagation linkage data information θ to the target propagation linkage data information θ 1 based on the first adjustment reference parameter, the cloud authentication service system 100 performs network configuration on the target policy decision network after updating the network propagation linkage data information to the target propagation linkage data information θ 1 based on the second feature distribution Db, and obtaining a second adjustment reference parameter L θ 1, and determining, by the cloud authentication service system 100, an update reference parameter ω i corresponding to the target sub-example authentication event list Ui based on the second adjustment reference parameter L θ 1 and the initial propagation linkage data information θ. Wherein the first profile comprises a first target authentication event profile Da1 of the first target sub-example authentication event manifest in the first business dynamic environment, and a third authentication event signature distribution sequence Da2 in the second business dynamic environment, the second signature distribution comprising a second target authentication event signature distribution sequence Db1 of the second target sub-instance authentication event manifest in the first business dynamic environment, in the fourth authentication event feature distribution sequence Db2 in the second service dynamic environment, the specific manner of obtaining the first adjustment reference parameter L θ by the cloud authentication service system 100 performing network configuration on the target policy decision network based on the first feature distribution Da may be that the cloud authentication service system 100 calls the target policy decision network to make a decision on the first target authentication event feature distribution sequence Da1 to obtain a first decision example feature distribution Qa 1; a mean square error between the first decision example signature distribution Qa1 and the third authentication event signature distribution sequence Da2 is calculated, and the mean square error is taken as the first adjustment reference parameter L θ. Similarly, the cloud authentication service system 100 calls the target policy decision network with updated parameters to decide the target policy decision network on the second target authentication event feature distribution sequence Db1, so as to obtain a second decision example feature distribution Qb 1; a mean square error between the second decision example signature distribution Qb1 and the fourth authentication event signature distribution sequence Db2 is calculated, and the mean square error is taken as the first adjustment reference parameter L θ 1.

It should be noted that, the specific manner in which the cloud authentication service system 100 updates the network propagation linkage data information of the target policy decision network from the initial propagation linkage data information θ to the target propagation linkage data information θ 1 based on the first adjustment reference parameter L θ may be that a partial derivative value of L θ with respect to θ is calculated, the partial derivative value is weighted based on the preset parameter λ to obtain a weighted partial derivative value, and the cloud authentication service system 100 determines the difference between the initial propagation linkage data information θ and the weighted partial derivative value as the target propagation linkage data information θ 1.

The specific way of determining the updated reference parameter ω i corresponding to the target sub-example authentication event list Ui by the cloud authentication service system 100 based on the second adjustment reference parameter L θ 1 and the initial propagation linkage data information θ may be to calculate a second-order partial derivative value of L θ 1 with respect to θ, and use the second-order partial derivative value as the updated reference parameter ω i corresponding to the target sub-example authentication event list Ui.

Through the scheme, the cloud authentication service system 100 can calculate the feature distribution corresponding to each group of sub-example authentication event lists to perform network configuration on the target policy decision network, so as to obtain an updated reference parameter set { ω 1, … …, ω N }, where ω i in the set corresponds to the updated reference parameter of Ui corresponding to any one group of sub-example authentication event lists in the N groups of sub-example authentication event lists.

By the method for obtaining the updated reference parameter based on the second-order partial derivative, the relation between different parameters obtained by decision can be constructed, so that the parameters obtained by different training before and after the parameter form influence on the final network transmission linkage data information, and the accuracy of the strategy decision network on the characteristic distribution decision is improved.

In step S240, the cloud authentication service system 100 updates the network propagation linkage data information in the target policy decision network based on the N update reference parameters.

In one possible design example, after the cloud authentication service system 100 calculates N update reference parameters, the network propagation linkage data information θ in the target policy decision network may be updated based on the N update reference parameters.

For example, N updated reference parameters ω i may be fused to obtain a fusion reference parameter, propagation linkage data corresponding to the fusion reference parameter is obtained, the propagation linkage data is used to perform weighting processing on the fusion reference parameter to obtain a weighted reference parameter, and the cloud authentication service system 100 updates network propagation linkage data information in the target policy decision network from initial propagation linkage data information to a difference value between the initial propagation linkage data information and the weighted reference parameter.

Step S250, if the target policy decision network after the network propagation linkage data information update meets the preset condition, determining the target policy decision network after the network propagation linkage data information update as the trained policy decision network.

In a possible design example, the preset condition includes that the policy decision precision of each feature distribution in the feature distribution corresponding to each group of the sub-example authentication event lists by the target policy decision network is higher than the preset precision, wherein the cloud authentication service system 100 obtains any first feature distribution in the feature distributions corresponding to the sub-example authentication event lists to make a decision to obtain a first decision feature distribution, calculates a mean square error between the first decision feature distribution and a second feature distribution, determines that the decision on the first feature distribution is accurate when the mean square error is smaller than the preset error, and the cloud authentication service system 100 determines the policy decision precision of each feature distribution in the feature distributions in this way. The first characteristic distribution is the characteristic distribution of any one authentication event in the sub-example authentication event list in the first service dynamic environment, the second characteristic distribution is the characteristic distribution of the authentication event in the second service dynamic environment, and the first decision characteristic distribution is the characteristic distribution obtained by calling a policy decision network to make a decision on the first characteristic distribution.

In one possible design example, the training process of the policy decision network includes a plurality of iterative updates of network propagation linkage data information of the target policy decision network, each iterative update process of the network propagation linkage data information, the network propagation linkage data information of the target policy decision network and N groups of updated reference parameter decision processes, the updated reference parameters are obtained by training based on a group of sub-example authentication event feature distribution sequences, each group of sub-example authentication event feature distribution sequences is the feature distribution corresponding to any group of sub-example authentication event lists in the N groups of sub-example authentication event lists, the N groups of sub-example authentication event lists are obtained by searching based on the example authentication event list, one update of the network propagation linkage data information is completed through a plurality of updated reference parameters, and the parameters are updated by combining the features of the plurality of groups of examples, the iterative process is more regular, parameters updated by combining multiple groups of characteristics can be close to the optimal network transmission linkage data information, and the training efficiency of the strategy decision network is improved. Moreover, different combination modes are adopted to combine the example authentication events to obtain a plurality of sub-example authentication event lists, and the updating of the network transmission linkage data information is completed by taking the sub-example authentication event lists as training units.

Based on the above description, the information encryption policy calling procedure may include the following steps S310 to S370:

in step S310, the cloud authentication service system 100 acquires a second dynamic service environment for performing information encryption policy invocation on the target authentication event.

In a possible design example, the second business dynamic environment may specifically be a target business application service, and the like, and the cloud authentication service system 100 may obtain the second business dynamic environment that needs to perform information encryption policy invocation on the authentication event in advance, in a possible design example, the cloud authentication service system 100 may obtain the second business dynamic environment when a trigger condition is detected, and the trigger condition may be that the second business dynamic environment is installed on a terminal used by the target authentication event or that the terminal is registered in the second business dynamic environment. If the second service dynamic environment is a target service application service and the target service application service has a function of performing information encryption policy invocation on the authentication event, the cloud authentication service system 100 acquires the target service application service when detecting that the target service application service is installed on a terminal used by the target authentication event, so as to further perform information encryption policy invocation on the target authentication event in the target service application service.

In step S320, the cloud authentication service system 100 finds the first dynamic service environment matching the second dynamic service environment from the dynamic service environment list.

In one possible design example, after the cloud authentication service system 100 obtains the second business dynamic environment, the first business dynamic environment matching the second business dynamic environment may be found from a business dynamic environment list, where the business dynamic environment list includes at least one target business dynamic environment.

In a possible design example, the specific way of the cloud authentication service system 100 finding the first service dynamic environment matched with the second service dynamic environment from the service dynamic environment list may be that the cloud authentication service system 100 obtains the second service dynamic environment characteristic of the second service dynamic environment and the target service dynamic environment characteristic of each target service dynamic environment in the service dynamic environment list, and each target service dynamic environment stores service dynamic associated data with a target authentication event; the cloud authentication service system 100 determines correlation parameters between the second service dynamic environment characteristics and the target service dynamic environment characteristics of each target service dynamic environment, and finds out target service dynamic environment characteristics of which the correlation parameters with the second service dynamic environment characteristics meet preset conditions; the cloud authentication service system 100 determines a business dynamic environment corresponding to the target business dynamic environment feature as a first business dynamic environment matched with a second business dynamic environment. The service dynamic environment characteristics comprise at least one of service dynamic environment types, service dynamic environment resources and service dynamic environment applicable users, the service dynamic environment types comprise payment dynamic environment types, multimedia dynamic environment types and the like, the service dynamic environment resources are the size of resources of a terminal for installing the service dynamic environment occupied by the service dynamic environment, the service dynamic environment applicable users are divided into individual users, enterprise users, private users, public users and the like, correlation parameters corresponding to different service dynamic environment characteristics can be preset, for example, the correlation parameters of multimedia and new media are 50%, the correlation parameters are 80% when the difference of the occupied memory size is within 100 megabytes, and the correlation parameters of the individual users and the private users are 80%. If the service dynamic environment characteristic is a service dynamic environment type, the type of the second service dynamic environment is a new media, the type of the target service dynamic environment 1 is a multimedia, the type of the target service dynamic environment 2 is a live broadcast, the cloud authentication service system 100 determines that the correlation parameter of the multimedia and the new media is 50% from the preset correlation parameter corresponding relation, the correlation parameter of the live broadcast and the new media is 30%, and the preset condition is that the correlation parameter is the highest, the characteristic of the target service dynamic environment 1 meets the preset condition, and the cloud authentication service system 100 determines the target service dynamic environment 1 as a first service dynamic environment matched with the second service dynamic environment. In a possible design example, the first business dynamic environment may also be a set of multiple target business dynamic environments, if the preset condition is that the correlation parameter is greater than 20%, then both the characteristics of the target business dynamic environment 1 and the characteristics of the target business dynamic environment 2 satisfy the preset condition, and the cloud authentication service system 100 determines the target business dynamic environment 1 and the target business dynamic environment 2 as the first business dynamic environment matching the second business dynamic environment.

In step S330, the cloud authentication service system 100 obtains an example authentication event list dynamically associated with the second service dynamic environment and the first service dynamic environment, and determines an example feature distribution corresponding to the example authentication event list.

In one possible design example, after the cloud authentication service system 100 determines the second business dynamic environment and the first business dynamic environment, an example authentication event dynamically associated with the second business dynamic environment and the first business dynamic environment is searched, and an example authentication event list is constructed based on the searched example authentication events.

Further, the cloud authentication service system 100 determines an example feature distribution corresponding to the example authentication event list, where the example feature distribution includes a first example feature distribution of the example authentication event list in the first business dynamic environment and a target example feature distribution in the second business dynamic environment, the first example feature distribution includes a first example feature distribution of each example authentication event in the first business dynamic environment, and the target example feature distribution includes a target example feature distribution of each example authentication event in the second business dynamic environment.

In step S340, the cloud authentication service system 100 performs target-oriented policy decision network training through the example feature distribution corresponding to the example authentication event list to obtain a policy decision network.

In one possible design example, the example feature distributions include a first example feature distribution of the example authentication event list in the first business dynamic environment and a target example feature distribution in the second business dynamic environment, and the cloud authentication service system 100 performs network configuration on the target policy decision network specifically through the first example feature distribution and the target example feature distribution, so that a correlation parameter between the target policy decision network and the target example feature distribution obtained by processing the first example feature distribution by the target initialization network is higher than a preset threshold, that is, the policy decision network has the capability of deciding the feature distribution in the first business dynamic environment to the feature distribution in the second business dynamic environment. The correlation parameter between the feature distributions may be specifically determined by a mean square error value between the feature distributions.

In one possible design example, the first business dynamic environment is a collection of multiple business dynamic environments, such as including a first target business dynamic environment and a second target business dynamic environment, the policy decision network also includes a first policy decision network and a second policy decision network, the cloud authentication service system 100 may perform network configuration on the target policy decision network through the first target example feature distribution of the example authentication event list in the first target business dynamic environment and the target example feature distribution in the second business dynamic environment, so as to obtain the first policy decision network, and performing network configuration on the target policy decision network through second target example feature distribution of the example authentication event list in a second target service dynamic environment and target example feature distribution in the second service dynamic environment to obtain a second policy decision network.

In step S350, the cloud authentication service system 100 obtains a first authentication event feature distribution of the target authentication event in the first service dynamic environment.

In one possible design example, the target authentication event has a business dynamic association with the first business dynamic environment in a history, so the cloud authentication service system 100 may obtain the first authentication event feature distribution generated based on the business dynamic association data of the target authentication event and the first business dynamic environment. The first authentication event feature distribution may be specifically used to represent an encryption calling feature of the target authentication event to the first service dynamic environment, and if the first service dynamic environment is the first service application service, the first authentication event feature distribution is a feature distribution generated based on an access encryption calling feature when the target authentication event accesses the first service application service. The cloud authentication service system 100 may obtain a first authentication event feature distribution of the target authentication event in the first business dynamic environment.

In one possible design example, the first business dynamic environment is a set of multiple business dynamic environments, such as a first target business dynamic environment and a second target business dynamic environment, and the first authentication event feature distribution of the target authentication event in the first business dynamic environment includes a first authentication event feature distribution of the target authentication event in the first target business dynamic environment and a second authentication event feature distribution of the target authentication event in the second target business dynamic environment.

In step S350, the cloud authentication service system 100 invokes the policy decision network to make a decision on the first authentication event feature distribution, so as to obtain a second authentication event feature distribution of the target authentication event in the second service dynamic environment.

In a possible design example, the target authentication event may have no or only a small amount of service dynamic associated data in the second service dynamic environment, that is, data analysis cannot be performed based on access information of the target authentication event in the second service dynamic environment, so that the authentication event feature distribution of the target authentication event in the second service dynamic environment is accurately determined. The cloud authentication service system 100 may invoke a policy decision network to make a decision on the feature distribution of the first authentication event when detecting that the target authentication event has no or only a small amount of service dynamic associated data in the second service dynamic environment, so as to obtain the second authentication event feature distribution of the target authentication event in the second service dynamic environment.

In one possible design example, if there is one first business dynamic environment, the cloud authentication service system 100 may directly invoke the policy decision network to make a decision on the first authentication event feature distribution of the target authentication event in the first business dynamic environment, so as to obtain the second authentication event feature distribution of the target authentication event in the second business dynamic environment.

In one possible design example, the first business dynamic environment is plural, such as comprising a first target business dynamic environment and a second target business dynamic environment, the policy decision network comprises a first policy decision network for making a decision on feature distribution in the first target business dynamic environment and a second policy decision network for making a decision on feature distribution in the second target business dynamic environment, the cloud authentication service system 100 may invoke a first policy decision network to make a decision on the first target authentication event feature distribution, resulting in a third authentication event feature distribution, and invoking a second policy decision network to make a decision on the second target authentication event feature distribution to obtain a fourth authentication event feature distribution, and fusing the third authentication event characteristic distribution and the fourth authentication event characteristic distribution to obtain a second authentication event characteristic distribution of the target authentication event in the second service dynamic environment. The strategy decision network decides the first authentication event feature distribution of the target authentication event in the first service dynamic environment into the second authentication event feature distribution in the second service dynamic environment, so that the encryption calling feature of the target authentication event can be conveniently analyzed in the second service dynamic environment, and a corresponding object is called to the target authentication event encryption strategy in the second service dynamic environment.

In step S370, the cloud authentication service system 100 obtains encryption policy invocation information for the target authentication event according to the second authentication event feature distribution, and invokes the encryption policy for the target authentication event in the second service dynamic environment.

In one possible design example, after the cloud authentication service system 100 obtains the second authentication event characteristic distribution of the target authentication event in the second business dynamic environment, encryption policy invocation information may be obtained for the target authentication event based on the second authentication event profile, in one possible design example, the encryption policy invocation information is one or more encryption policy nodes in the encryption policy node set to be encrypted policy invoked, the cloud authentication service system 100 may obtain encryption policy parameters of each encryption policy node in the encryption policy node set to be encrypted policy invoked in the second business dynamic environment, and determining an encryption policy invocation index for each encryption policy node in the set of encryption policy nodes to be invoked for the encryption policy based on the second authentication event feature distribution and each encryption policy parameter, and finding out encryption strategy calling information based on the encryption strategy calling indexes of the encryption strategy nodes. The cloud authentication service system 100 obtains the encryption policy parameters and determines the encryption policy invocation indexes for the encryption policy nodes in the encryption policy node set to be invoked by the encryption policy to be invoked based on the second authentication event characteristic distribution and the encryption policy parameters.

In a possible design example, the encryption policy invocation index of each encryption policy node in the encryption policy node set called by the policy to be encrypted may be determined in such a manner that the cloud authentication service system 100 performs a dot-product decision on the second authentication event feature distribution and each encryption policy parameter, respectively, to obtain a dot-product value corresponding to each encryption policy node, and determines the dot-product value corresponding to each encryption policy node as the encryption policy invocation index for each encryption policy node in the encryption policy node set called by the policy to be encrypted, where the higher the encryption policy invocation index is, the more matched the encryption invocation feature of the object and the target authentication event is.

Further, after the cloud authentication service system 100 determines the encryption policy invocation index of each encryption policy node, an object in the encryption policy node set to be invoked by the encryption policy to be invoked may be searched based on the encryption policy invocation index, so as to obtain the encryption policy invocation information for the target authentication event.

In a possible design example, the specific way for the cloud authentication service system 100 to search for the objects in the encryption policy node set to be called by the encryption policy based on the encryption policy calling index may be that the cloud authentication service system 100 sorts the objects in the encryption policy node set to be called by the encryption policy according to the order of the encryption policy calling index from high to low, determines the object sorted to the first I bit as the target encryption policy node, and uses the I target encryption policy nodes as the encryption policy calling information for the target authentication event, where I is a positive integer.

In a possible design example, the specific way for the cloud authentication service system 100 to search for an object in the set of encryption policy nodes to be called by the encryption policy based on the encryption policy invocation index may be that the cloud authentication service system 100 detects whether the encryption policy invocation index of each encryption policy node is greater than a preset threshold, and determines the object whose encryption policy invocation index is greater than the preset threshold as a target encryption policy node, and the target encryption policy node is used as the encryption policy invocation information for the target authentication event.

After the cloud authentication service system 100 determines the encryption policy invocation information for the target authentication event, the encryption policy invocation information may be invoked in the second service dynamic environment.

In one possible design example, a feature distribution representation in a second business dynamic environment of an authentication event may be determined based on a feature distribution representation of the authentication event in multiple first business dynamic environments, making the feature distribution decision more accurate, and for objects of different encryption policy invocation indexes, the cloud authentication service system 100 may have different encryption policy invocation modes, specifically, a target encryption policy invocation mode for each target encryption policy node in the encryption policy invocation information is determined based on a corresponding relationship between the encryption policy invocation index and the encryption policy invocation mode, the encryption policy invocation mode includes at least one of an encryption policy invocation sequence, an encryption policy invocation duration, and an encryption policy invocation frequency, and the cloud authentication service system 100 invokes the encryption policy invocation information in the second service dynamic environment based on the target encryption policy invocation mode. The corresponding relation between the encryption policy calling index and the encryption policy calling mode can be that the higher the encryption policy calling index is, the more advanced the encryption policy calling sequence is, the longer the encryption policy calling time is, and the higher the encryption policy calling frequency is. By the method, the corresponding information can be more intelligently called for the authentication event encryption strategy, so that the information called by the encryption strategy is more matched with the authentication event encryption calling characteristics.

In one implementation scenario, the solution may be used between different business dynamic environments, such as a first business dynamic environment and a second business dynamic environment. The method is mainly used for linking encryption strategy calling across service dynamic environments, namely, the encryption calling characteristics of the authentication event in the second service dynamic environment are determined based on the encryption calling characteristics of the authentication event in the first service dynamic environment. And the method does not change the original encryption strategy calling system and influence the encryption strategy calling of the common authentication event. Aiming at a newly added authentication event, firstly inquiring service dynamic associated data of the authentication event in a first service dynamic environment to obtain first authentication event characteristic distribution of the authentication event in the first service dynamic environment, then obtaining second authentication event characteristic distribution of the authentication event in a second service dynamic environment through a policy decision network, and determining information needing to carry out encryption policy calling on the authentication event by using the second authentication event characteristic distribution. For example, assuming that a target authentication event has dynamic service related data in a first dynamic service environment, the cloud authentication service system 100 may obtain a first authentication event feature distribution corresponding to the target authentication event in the first dynamic service environment from a background server in the first dynamic service environment, call a policy decision network to make a decision on the first authentication event feature distribution, obtain a second authentication event feature distribution in a second dynamic service environment, find out encryption policy invocation information based on the second authentication event feature distribution, and send the encryption policy invocation information to a terminal used by the target authentication event, so that the target authentication event refers to the encryption policy invocation information. In the calling process of the policy decision network, in a possible design example, assuming that there is a model in each of the first business dynamic environment and the second business dynamic environment, the first model (R1) in the first business dynamic environment is used for making a decision on a first feature distribution U1 of the target authentication event in the first business dynamic environment to obtain a first encryption policy calling information V1 of the target authentication event in the first business dynamic environment, the target model (R2) in the second business dynamic environment is used for making a decision on a target feature distribution U2 of the target authentication event in the second business dynamic environment to obtain a target encryption policy calling information V2 of the target authentication event in the second business dynamic environment, when the first feature distribution U1 of the target authentication event in the first business dynamic environment and the target feature distribution U2 in the second business dynamic environment are not present, the first signature distribution U1 may be used as an input to the policy decision network, which makes a decision on U1, i.e., a target signature distribution U2 of the target authentication event in the second business dynamic environment may be output. Then, the target feature distribution U2 is obtained by making a decision on the first feature distribution U1, so that a decision on the target feature distribution U2 through a target model (R2) can be achieved, and the target encryption policy invoking information V2 of the target authentication event in the second service dynamic environment is obtained, so as to achieve a decision between the first encryption policy invoking information V1 and the target encryption policy invoking information V2.

Further, in a possible design concept, the above scheme may further include the following steps.

Step S114, acquiring a block chain security authentication library generated after performing authentication encryption after performing encryption policy invocation of a target authentication event in the second service dynamic environment, where an authentication digital certificate of an authentication service object for various dynamic access behaviors is stored in the block chain security authentication library.

Step S115, based on the block chain security authentication library, performing dynamic bidirectional authentication on each received dynamic access behavior, and after the dynamic bidirectional authentication is passed, generating an authentication service transmission channel for each dynamic access behavior.

Step S116, acquiring control information for acquiring service data of the authentication service transmission channel, where the acquisition control information is control model information for acquiring service data of each authentication service transmission channel in the sequence of authentication service transmission channels to be processed.

Step S117, obtaining target distributed control information corresponding to the acquisition control information, where the target distributed control information corresponding to the acquisition control information includes service acquisition node distribution corresponding to the acquisition control information.

Step S118, performing service acquisition decision analysis on target distributed control information corresponding to the acquisition control information according to the target service acquisition decision network, to obtain service acquisition node distribution corresponding to the acquisition control information.

And step S119, determining a target service data acquisition project corresponding to the acquisition control information according to the service acquisition node distribution corresponding to the acquisition control information, and allocating a corresponding target service acquisition process to the authentication service transmission channel to acquire big data of the authentication service transmission channel.

Based on the steps, the service data acquisition items required by the authentication service transmission channels are subjected to system analysis through the acquisition control information, so that the large data acquisition is performed on the corresponding authentication service transmission channels according to the target service data acquisition items corresponding to the authentication service transmission channels in the authentication service transmission channel sequence, and the large data acquisition efficiency and the large data acquisition precision can be improved.

For example, in a possible design concept, before step S118, the embodiment may further obtain the first acquisition control information example, the service acquisition decision reference information corresponding to the first acquisition control information example, and the second acquisition control information example. And acquiring target distributed control information corresponding to the first acquisition control information example and target distributed control information corresponding to the second acquisition control information example. And training a preset service acquisition decision network according to the target distributed control information corresponding to the first acquisition control information example and the service acquisition decision reference information corresponding to the first acquisition control information example to obtain a first service acquisition decision network. And performing service acquisition decision analysis on target distributed control information corresponding to the second acquisition control information example according to the first service acquisition decision network to obtain first to-be-determined service acquisition node distribution corresponding to the second acquisition control information example. And performing service acquisition decision analysis on target distributed control information corresponding to the first acquisition control information example according to a preset service acquisition decision network to obtain undetermined service acquisition node distribution corresponding to the first acquisition control information example, calculating a difference parameter according to the undetermined service acquisition node distribution corresponding to the first acquisition control information example and service acquisition decision reference information corresponding to the first acquisition control information example, and reversely updating the parameter of the preset service acquisition decision network by using the difference parameter. And (3) iteratively executing the process until a termination condition of the supervised training is met, obtaining a first service acquisition decision network, wherein the termination condition of the supervised training comprises at least one of the following conditions: the iterative training times reach the set times, the difference parameter is smaller than the set threshold value, and the difference parameter is converged. And adjusting parameters of the first service acquisition decision network after parameter adjustment again according to the target distributed control information corresponding to the second acquisition control information example, the target distributed control information corresponding to the first acquisition control information example and the service acquisition decision reference information corresponding to the first acquisition control information example until a training end condition is reached, and taking the first service acquisition decision network obtained when the training end condition is reached as a target service acquisition decision network.

For example, in a possible design concept, the service acquisition node distribution corresponding to the acquisition control information includes a service acquisition confidence corresponding to a reference service acquisition decision manner and service acquisition confidence corresponding to a plurality of pending service acquisition decision manners, and step S119 may be implemented by the following implementation manners.

And a substep S1191 of acquiring a service acquisition confidence corresponding to each pending service acquisition decision mode and a service acquisition confidence corresponding to a reference service acquisition decision mode according to pre-recorded historical service acquisition control data.

And a substep S1192 of comparing the service acquisition confidence corresponding to each pending service acquisition decision mode with the service acquisition confidence corresponding to the reference service acquisition decision mode.

And a substep S1193, when the service acquisition confidence corresponding to each pending service acquisition decision mode is less than or equal to the service acquisition confidence corresponding to the reference service acquisition decision mode, taking the reference service acquisition decision list corresponding to the reference service acquisition decision mode as the target service data acquisition item corresponding to the acquisition control information.

And a substep S1194, when the service acquisition confidence corresponding to each of the pending service acquisition decision modes is greater than the service acquisition confidence corresponding to the reference service acquisition decision mode, taking the pending service acquisition decision list corresponding to the pending service acquisition decision modes as a target service data acquisition item corresponding to the acquisition control information, wherein the pending service acquisition decision list corresponding to the pending service acquisition decision modes is the service acquisition decision list corresponding to the service acquisition decision mode with the highest service acquisition confidence.

Fig. 3 is a schematic functional module diagram of a big data mining device 300 based on blockchain security authentication according to an embodiment of the present disclosure, and in this embodiment, the big data mining device 300 based on blockchain security authentication may be divided into functional modules according to the method embodiment executed by the cloud authentication service system 100, that is, the following functional modules corresponding to the big data mining device 300 based on blockchain security authentication may be used to execute the method embodiments executed by the cloud authentication service system 100. The big data mining device 300 based on blockchain security authentication may include an obtaining module 310, a first determining module 320, a second determining module 330, and a pushing module 340, and the functions of the functional modules of the big data mining device 300 based on blockchain security authentication are described in detail below.

The obtaining module 310 is configured to obtain service big data, where after the dynamic bidirectional authentication of the blockchain authentication terminal 200 and the cloud authentication service system 100 is passed, the service big data is obtained by allocating a corresponding target service acquisition process to the generated authentication service transmission channel to perform big data acquisition on the authentication service transmission channel. The obtaining module 310 may be configured to perform the step S110, and the detailed implementation of the obtaining module 310 may refer to the detailed description of the step S110.

The first determining module 320 is configured to obtain data of the service big data under the at least two service hotspot propagation tags, and determine service hotspot propagation characteristics of the service big data under the at least two service hotspot propagation tags according to the data of the service big data under the at least two service hotspot propagation tags. The first determining module 320 may be configured to perform the step S120, and for a detailed implementation of the first determining module 320, reference may be made to the detailed description of the step S120.

The second determining module 330 is configured to obtain a service propagation association parameter between each service big data according to service hot spot propagation characteristics of the service big data under at least two service hot spot propagation tags. The second determining module 330 may be configured to perform the step S130, and the detailed implementation of the second determining module 330 may refer to the detailed description of the step S130.

The pushing module 340 is configured to determine target topic linkage information of each service big data according to the service propagation association parameter between each service big data, and perform information pushing on the blockchain authentication terminal 200 based on the target topic linkage information. The pushing module 340 may be configured to perform the step S140, and the detailed implementation manner of the pushing module 340 may refer to the detailed description of the step S140.

It should be noted that the division of each module of the above apparatus is only a division of a logic function, and when the actual implementation is implemented, all or part of the division may be integrated into one physical entity, or may be physically separated. And these modules may all be implemented in software invoked by a processing element. Or may be implemented entirely in hardware. And part of the modules can be realized in the form of calling software by the processing element, and part of the modules can be realized in the form of hardware. For example, the obtaining module 310 may be a processing element separately set up, or may be implemented by being integrated into a chip of the apparatus, or may be stored in a memory of the apparatus in the form of program code, and the processing element of the apparatus calls and executes the functions of the obtaining module 310. Other modules are implemented similarly. In addition, all or part of the modules can be integrated together or can be independently realized. The processing element described herein may be an integrated circuit having signal processing capabilities. In implementation, each step of the above method or each module above may be implemented by an integrated logic circuit of hardware in a processor element or an instruction in the form of software.

Fig. 4 is a schematic diagram illustrating a hardware structure of a cloud authentication service system 100 for implementing the above big data mining method based on block chain security authentication according to an embodiment of the present disclosure, and as shown in fig. 4, the cloud authentication service system 100 may include a processor 110, a machine-readable storage medium 120, a bus 130, and a transceiver 140.

In a specific implementation process, at least one processor 110 executes computer-executable instructions stored in the machine-readable storage medium 120 (for example, the obtaining module 310, the first determining module 320, the second determining module 330, and the pushing module 340 included in the big data mining apparatus 300 based on blockchain security authentication shown in fig. 3), so that the processor 110 may execute the big data mining method based on blockchain security authentication according to the above method embodiment, where the processor 110, the machine-readable storage medium 120, and the transceiver 140 are connected through the bus 130, and the processor 110 may be configured to control the transceiving action of the transceiver 140, so as to perform data transceiving with the blockchain authentication terminal 200.

For a specific implementation process of the processor 110, reference may be made to the above-mentioned method embodiments executed by the cloud authentication service system 100, which implement principles and technical effects are similar, and details of this embodiment are not described herein again.

In addition, the embodiment of the disclosure also provides a readable storage medium, wherein a computer execution instruction is preset in the readable storage medium, and when a processor executes the computer execution instruction, the big data mining method based on the block chain security authentication is implemented.

Finally, it should be understood that the examples in this specification are only intended to illustrate the principles of the examples in this specification. Other variations are also possible within the scope of this description. Thus, by way of example, and not limitation, alternative configurations of the embodiments of the specification can be considered consistent with the teachings of the specification. Accordingly, the embodiments of the present description are not limited to only those embodiments explicitly described and depicted herein.

Claims

1. A network training method based on machine learning is applied to a cloud authentication service system, the cloud authentication service system is in communication connection with a plurality of block chain authentication terminals, and the method comprises the following steps:

acquiring service propagation associated parameters of marked samples under at least two service hotspot propagation labels, corresponding first propagation linkage data, service interaction contact information among the service propagation associated parameters, corresponding second propagation linkage data and actual service propagation associated parameters among the marked samples;

training the machine learning network according to service propagation associated parameters of the marked samples under at least two service hotspot propagation labels, corresponding first propagation linkage data, service interaction contact information among the service propagation associated parameters and corresponding second propagation linkage data to obtain a trained machine learning network;

obtaining loss parameter distribution between the service propagation associated parameters output by the trained machine learning network and the corresponding actual service propagation associated parameters;

and when the loss parameter distribution is larger than or equal to a preset threshold value, adjusting the first transmission linkage data and the second transmission linkage data according to the loss parameter distribution, and repeatedly training the machine learning network according to the adjusted first transmission linkage data and second transmission linkage data until the loss parameter distribution obtained according to the trained machine learning network is smaller than the preset threshold value.

2. The machine learning based network training method of claim 1, the method further comprising:

calculating service propagation associated parameters of every two pieces of service big data under at least two service hot spot propagation labels according to service hot spot propagation characteristics of the service big data under at least two service hot spot propagation labels;

inputting service propagation associated parameters of the two service big data under at least two service hotspot propagation labels into a pre-trained machine learning network;

confirming service interaction contact information among the service propagation associated parameters through the machine learning network;

acquiring propagation linkage data of each service propagation associated parameter and propagation linkage data of each service interaction contact information;

calculating to obtain service propagation associated parameters among the service big data according to the service propagation associated parameters, the corresponding propagation linkage data, the service interaction contact information and the corresponding propagation linkage data;

3. The machine learning-based network training method of claim 2, wherein the step of obtaining the data of the traffic big data under at least two service hotspot propagation labels comprises:

collecting service big data in a target service range and updating data of service subscription of the service big data;

and acquiring data of the service big data under at least two service hotspot propagation labels in the target service range according to the service big data in the target service range and the data of the service big data subscribed by the update service.

4. The machine-learning-based network training method of claim 3, wherein the service hotspot propagation tags comprise a first service hotspot propagation tag, a second service hotspot propagation tag, and a third service hotspot propagation tag;

the step of acquiring data of the service big data under at least two service hot spot propagation labels in the target service range according to the service big data in the target service range and the data of the service big data subscribed by the update service comprises the following steps:

according to the data of the service big data subscribed by the updating service, determining the transmission linkage data between the updating service and the service big data in a target service range, wherein the transmission linkage data is used as the data of the service big data under the first service hotspot transmission label;

extracting a frequent theme from the service big data, acquiring the propagation linkage data of the frequent theme in the service big data within the target service range, and identifying the propagation linkage data as the data of the service big data under the second service hotspot propagation label;

and according to the propagation linkage data between the updating service and the service big data and the propagation linkage data of the frequent theme in the service big data in the target service range, confirming the propagation linkage data of the updating service and the frequent theme in the service big data as the data of the service big data under the third service hotspot propagation label.

5. The machine learning-based network training method according to claim 2, wherein the step of determining service hotspot propagation characteristics of the traffic big data under at least two service hotspot propagation labels according to the data of the traffic big data under the at least two service hotspot propagation labels comprises:

according to the data of the service big data under at least two service hot spot propagation labels, constructing service distribution of the service big data under the at least two service hot spot propagation labels;

and extracting the theme mining characteristics of the service distribution to obtain the service hotspot propagation characteristics of the service big data under the at least two service hotspot propagation labels.

6. The machine learning-based network training method of claim 5, wherein the step of constructing the traffic distribution of the traffic big data under at least two service hotspot propagation labels according to the data of the traffic big data under the at least two service hotspot propagation labels comprises:

respectively constructing a first service distribution, a second service distribution and a third service distribution according to the data of the service big data under the at least two service hot spot propagation labels; the first business distribution is used for representing the relationship between the updated service description and the service hotspot propagation characteristics, the second business distribution is used for representing the relationship between the service hotspot propagation characteristics and the frequent item topic description, and the third business distribution is used for representing the relationship between the updated service description and the frequent item topic description;

and respectively identifying the first service distribution, the second service distribution and the third service distribution as the service distribution of the service big data under the at least two service hotspot propagation labels.

7. The machine learning-based network training method of claim 6, wherein if the service distribution is the first service distribution or the second service distribution, the step of performing topic mining feature extraction on the service distribution to obtain service hotspot propagation features of the service big data under the at least two service hotspot propagation labels comprises:

performing theme mining feature extraction on the service distribution to obtain service hotspot propagation features serving as service hotspot propagation features of the service big data under corresponding service hotspot propagation labels;

if the service distribution is the third service distribution, the step of extracting the subject mining characteristics of the service distribution to obtain the service hotspot propagation characteristics of the service big data under the at least two service hotspot propagation labels comprises the following steps:

performing theme mining feature extraction on the service distribution to obtain frequent item theme description in the service big data;

acquiring propagation linkage data of the frequent item theme in the service big data as propagation linkage data corresponding to the frequent item theme description;

and obtaining service hotspot propagation characteristics according to the frequent theme description and the corresponding propagation linkage data, wherein the service hotspot propagation characteristics are used as the service hotspot propagation characteristics of the service big data under the corresponding service hotspot propagation label.

8. The machine learning-based network training method according to any one of claims 2 to 7, wherein the step of determining target topic linkage information of each piece of business big data according to service propagation associated parameters between each piece of business big data, and pushing information to the blockchain authentication terminal based on the target topic linkage information includes:

acquiring a service propagation theme of which the service propagation associated parameter between the service big data is greater than a preset associated parameter, and determining target theme linkage information of the service big data based on each service propagation theme with a service propagation relation;

and carrying out information pushing on the block chain authentication terminal based on the target theme linkage information.

9. A big data mining system based on block chain security authentication comprises a cloud authentication service system and a plurality of block chain authentication terminals in communication connection with the cloud authentication service system;

the cloud authentication service system is configured to:

10. A cloud authentication service system, characterized in that the cloud authentication service system includes a processor, a machine-readable storage medium, and a network interface, the machine-readable storage medium, the network interface, and the processor are connected through a bus system, the network interface is configured to be connected to at least one blockchain authentication terminal in a communication manner, the machine-readable storage medium is configured to store a program, an instruction, or a code, and the processor is configured to execute the program, the instruction, or the code in the machine-readable storage medium to perform the machine learning based network training method according to any one of claims 1 to 9.