Data privacy-protectable classification grading method and system based on integrated decision learning

Abstract

A data privacy protection classification grading method and system based on integrated decision learning, the user side equipment data category attribute parameters required for training the decision tree model are sent to the cloud server for data integration in a scrambled manner through symmetric encryption; the cloud server integrates all the category attributes uploaded by the user side equipment without decryption, and then sends the integrated ciphertext to each user side equipment, and the user side equipment completes the training of the decision tree model of the integrated decision learning by using the decrypted data; the trained decision tree model is used for classification and grading identification of the local data of the user side equipment. The application can realize automatic classification and grading identification of data in the cloud service scenario under encryption protection.

Description

Technical Field

[0001] This invention belongs to the field of power data security technology, and relates to a data privacy-protectable classification and grading method and system based on integrated decision learning. Background Technology

[0002] With the development of digital transformation of the power grid, power data, as a core production factor, has become a key strategic resource for promoting energy digital transformation and building a new power system. Strengthening data security protection is urgently needed, and data classification and grading is a crucial foundation for this work. The promulgation of the "Data Security Law" and the "Personal Information Protection Law" has clearly established requirements for data classification and grading. In recent years, power grid companies have actively conducted a series of preliminary explorations regarding data classification and grading, initially forming classification and grading methods and practical operational foundations. However, at present, classification and grading work mainly relies on manual methods, which suffers from low efficiency, poor accuracy, and difficulty in implementation.

[0003] Electricity data is characterized by its large volume, diverse types, and high level of confidentiality, including a large amount of commercially confidential data and personal information data that could impact the interests of the nation, society, and companies. To better clarify important data, identify data risks, and strengthen data security protection, the power industry has made a series of preliminary explorations in recent years regarding data classification and grading. It has successively issued documents such as the "Typical Architecture Design for Internet Business Data Security of State Grid Corporation of China (State Grid Interconnect

[2019] No. 806)," the "Notice of State Grid Corporation of China on Further Standardizing Data Security Work (State Grid Interconnect

[2020] No. 745)," and the "Notice of State Grid Corporation of China on Strengthening the Application of the Negative List for Data Sharing (State Grid Interconnect

[2020] No. 280)," which put forward overall requirements for the classification and grading of electricity data. Based on data tables and field descriptions, a negative list catalog for sharing was released through manual sorting to promote the internal sharing and integration of electricity data.

[0004] Currently, power data involves 11 business domains, including production, infrastructure, and marketing, which are then aggregated into three main categories: production data, marketing data, and management data. This presents a diverse range of power data types and complex scenarios. Traditional manual sorting methods are insufficient for effectively classifying and grading large volumes of power business data. Generally, automated big data classification processing based on cloud services is adopted. User-side devices can rely on cloud service resources to aggregate data from various user-side devices for integrated decision-making training to obtain a data classification and grading recognition model. However, traditional data classification and grading machine learning methods require cloud servers to aggregate raw data information from user-side devices for training and recognition, leading to data privacy leaks. The leakage of large amounts of customer privacy data, such as electricity consumption, overdue payments, and payment records, could harm customers' legitimate rights and interests. Furthermore, even after encrypting the data using traditional encryption algorithms, existing machine learning methods are unable to achieve data classification and grading recognition processing.

[0005] Therefore, how to simultaneously meet the two needs of data classification and grading identification in cloud service scenarios and data privacy protection of user-side devices is a key issue that data classification and grading solutions based on cloud services urgently need to address. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a data privacy-protected classification and grading method and system based on integrated decision learning, which can realize automatic classification and grading of data in cloud service scenarios under encrypted protection.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A data privacy-preserving classification and grading method based on ensemble decision learning includes the following steps:

[0009] Let D be the set of power data samples with known classification and grading information for each user-side device. Let c = {c1, c2, ..., c...} be the set corresponding to the category attributes of the sample data. k}, c i (i = 1, 2, ..., k) represents the i-th category attribute, and the frequency set P of the sample data corresponding to each category attribute is calculated. D ={P1,P2,…,P k};

[0010] The user-side device generates a random number z and adds z to P. D On each element, we obtain P. D ′={P1+z,P2+z,…,P k +z};

[0011] The random number z is encrypted using the public key of the cloud server, and the ciphertext of the random number z is sent to the cloud server;

[0012] The user-side equipment uses a symmetric encryption algorithm to secure P. D Each element in ' is symmetrically encrypted, and P is... D Each element in the ciphertext E(P) i Send +z) to the cloud server;

[0013] The random number z is obtained by decrypting the ciphertext of the random number z using the private key of the cloud server, and then from E(P) i Subtracting the random number z from +z) yields the ciphertext E(P) corresponding to the original frequency value of the sample data for the i-th attribute category. i ):

[0014] Receive the encrypted summation value returned by the cloud server, which is obtained by summing the ciphertext corresponding to the original values ​​of the sample data frequency of all user-side devices;

[0015] The user-side device decrypts the encrypted summation value to obtain the summation value, and calculates the information gain of each category attribute based on the summation value;

[0016] The user-side device trains the decision tree model using the information gain values ​​of each category attribute and the corresponding hierarchical classification of its original sample data as the training set, thus completing the integrated decision learning.

[0017] The trained decision tree model is used to classify and classify local data from user-side devices.

[0018] The present invention further includes the following preferred embodiments:

[0019] Preferably, the frequency set P D element P i The value of (i = 1, 2, ..., k) is the proportion of the number of sample data corresponding to the i-th category attribute of the user-side device to the total amount of data in the user-side device power sample dataset D.

[0020] Preferably, a random number r is selected, and the symmetric encryption algorithm public key pk is used to encrypt P according to the following formula. D 'Element P i +z is used for symmetric encryption to obtain ciphertext E(P). i +z):

[0021]

[0022] Where n and g are parameter values ​​in the public key pk = (n, g) of the symmetric encryption algorithm, and mod is the modulo expression.

[0023] Preferably, the public and private keys for the symmetric encryption algorithm are generated in the following manner:

[0024] Public key pk = (n, g), private key sk = (p, q, μ);

[0025] Where the Euler totient function of n is λ;

[0026] Let n be a random integer, and let n divide the order of g.

[0027] p and q are two randomly selected large prime numbers. Represented as a field of positive integers;

[0028]

[0029] λ = (p-1)(q-1).

[0030] Preferably, the ciphertext E(P) corresponding to the original value of the sample data frequency of the i-th attribute category is... i )for:

[0031]

[0032] Preferably, assuming there are m user-side devices, P ji Let E represent the raw frequency value of the sample data for the i-th attribute category of the j-th (j = 1, 2, ..., m) user-side device. The ciphertext corresponding to the raw frequency value of the sample data for the i-th attribute category of the j-th user-side device is E. j (P ji ),but The encrypted summation value is:

[0033]

[0034] Preferably, the decryption process is as follows:

[0035]

[0036] This is the summation value obtained after decryption.

[0037] Preferably, the information gain value IG(i) of the i-th category attribute is calculated using the following formula:

[0038] IG(i) = Entropy(D) - Entropy i (D)(4)

[0039] in,

[0040]

[0041] This is the summation value obtained after decryption;

[0042] D represents the set of power data samples, and the operator || represents the number of sets;

[0043] a l Let L be the l-th sample data value corresponding to the i-th category attribute of the j-th user-side device, 1≤l≤L, where L is the number of sample data corresponding to the i-th category attribute;

[0044] For a l The proportion of the number of corresponding sample data to the number of data for the i-th category attribute.

[0045] Preferably, the user-side device trains a decision tree model using the information gain values ​​of each category attribute and the hierarchical classification corresponding to its original sample data as the training set to complete ensemble decision learning, specifically as follows:

[0046] Create a root node for the decision tree. Starting from the root node, select the category attribute with the largest attribute gain value from all categories as the child node of the decision tree. Create branches according to the different values ​​of the category attribute and mark the corresponding hierarchical classification of the branch. Then, select the category attribute with the largest attribute gain value from the remaining category attributes as the next child node of the decision tree. Repeat the above steps until the remaining category attributes are zero to obtain the final decision tree model.

[0047] The present invention also provides a data privacy-protected classification and grading system based on ensemble decision learning, the system being used to implement the data privacy-protected classification and grading method based on ensemble decision learning.

[0048] The present invention also provides a terminal, including a processor and a storage medium; the storage medium is used to store instructions;

[0049] The processor is configured to operate according to the instructions to perform steps according to the data privacy-protected classification and grading method based on ensemble decision learning.

[0050] This invention also provides a data privacy-protected classification and grading method based on ensemble decision learning, wherein the cloud server receives the ciphertext of the random number z and E(P) i After +z), the random number z is decrypted using the private key of the cloud server, and then decrypted from E(P) to obtain the random number z. i Subtracting the random number z from +z) yields the ciphertext E(P) corresponding to the original frequency value of the sample data for the i-th attribute category. i ); where E(P) i +z) is P D Each element in the encrypted text, P D ′={P1+z,P2+z,…,P k +z},P1,P2,…,P kThe power data sample set is D, and the sample data frequencies correspond to each category attribute.

[0051] The cloud server aggregates the original frequency values ​​of sample data from all user-side devices, sums them to obtain an encrypted sum value, and returns the encrypted sum value to each user-side device.

[0052] The present invention also provides a cloud server that executes the data privacy-protected classification and grading method based on integrated decision learning.

[0053] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that the program, when executed by a processor, implements the steps of the method.

[0054] The beneficial effects of this invention are compared with those of the prior art:

[0055] This invention integrates user-side device data category attribute parameters required for training a decision tree model by scrambling and symmetric encryption before sending them to a cloud server. The cloud server integrates the category attributes uploaded by all user-side devices without decryption, and then distributes the resulting ciphertext to each user-side device. The user-side devices use the decrypted data to train the ensemble decision tree model. The trained decision tree model is then used for classification and grading of the local data on the user-side devices. By combining a symmetric encryption algorithm with an ensemble decision tree model, this invention achieves data integration in a cloud service scenario without decrypting the encrypted data from user-side devices. This integration is then pushed to each user-side device for local training of the classification and grading model, thus protecting privacy during the distributed automatic classification and grading process of user-side device data in a cloud service scenario. Attached Figure Description

[0056] Figure 1 This is a flowchart of the data privacy-protected classification and grading method based on integrated decision learning, as described in this invention. Detailed Implementation

[0057] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, other embodiments obtained by those skilled in the art without creative effort are all within the protection scope of this invention.

[0059] like Figure 1 As shown, Embodiment 1 of the present invention provides a data privacy-protected classification and grading method based on ensemble decision learning. In a preferred but non-limiting embodiment of the present invention, the method includes the following steps 1-9:

[0060] Step 1: Let D be the set of power data samples with known classification and grading for each user-side device. The sample data includes information such as user-side device name, ID number, electricity consumption, and electricity cost. Define the set corresponding to these category attributes as c = {c1, c2, ..., c...} k}, c i (i = 1, 2, ..., k) represents the i-th category attribute, and the frequency set P of the sample data corresponding to each category attribute is calculated. D ={P1,P2,…,P k};

[0061] More preferably, the frequency set P D element P in i The value of (i = 1, 2, ..., k) is the proportion of the number of sample data corresponding to the i-th category attribute of the user-side device to the total amount of data in the user-side device power sample dataset D.

[0062] More preferably, the classification and grading are illustrated with the following examples:

[0063] 1. Classification: Based on the current status of power industry operation and management and the management characteristics of power companies themselves, user-side equipment power data samples are integrated into two main categories:

[0064] 1) User-side device related data refers to identity-related data, service content data, and user-side device service-derived data related to personal user-side devices and enterprise customers;

[0065] 2) Enterprise-owned data refers to data held by power companies that is unrelated to user-side equipment, including network and system data, enterprise management data, partner data, etc.

[0066] Network and system data mainly involves information on the construction, operation and maintenance of networks and systems, information on hardware and software resources, and information on security management.

[0067] Enterprise management data mainly involves data related to enterprise strategy, planning and construction, business analysis, and office automation. This is further subdivided into several subcategories based on specific application scenarios.

[0068] 2. Classification: Based on the importance of the power data of user-side equipment and the impact and harm that leakage would cause to national security, social order, business operations and management and public interests, it is divided into four sensitivity levels: serious, high, medium and low.

[0069] Step 2: The user-side device generates a random number z and adds z to P. D On each element, we obtain P. D ′={P1+z,P2+z,…,P k +z};

[0070] Step 3: Encrypt the random number z using the cloud server's public key, and send the ciphertext of the random number z to the cloud server;

[0071] More preferably, the random number z is encrypted using the RSA algorithm.

[0072] Step 4: The user-side device uses a symmetric encryption algorithm to secure P. D Each element in ' is symmetrically encrypted, and P is... D Each element in the message is sent in encrypted form to the cloud server;

[0073] More preferably, a random number r is selected, and the symmetric encryption algorithm public key pk is used to encrypt P according to the following formula. D 'Element P i +z is used for symmetric encryption to obtain ciphertext E(P). i +z):

[0074]

[0075] Where n and g are parameter values ​​in the public key pk = (n, g) of the symmetric encryption algorithm, and mod is the modulo expression.

[0076] Generate the public and private keys for the symmetric encryption algorithm as follows:

[0077] Public key pk = (n, g), private key sk = (p, q, μ);

[0078] Where the Euler totient function of n is λ;

[0079] Let n be a random integer, and let n divide the order of g.

[0080] p and q are two randomly selected large prime numbers. Represented as a field of positive integers, g is modulo n 2 A random positive integer;

[0081]

[0082] λ = (p-1)(q-1);

[0083] There is g λ =1 mod n,g nλ =1 mod n 2 Because of g nλ This can be represented as (1+kn). n In the form of , k is any integer. Therefore, the design of μ in this invention can ensure that the random number r in step 4 is eliminated during decryption in step 7.

[0084] The ciphertext E(P) i +z) Send to the cloud server.

[0085] This invention encrypts the machine models trained by each user-side device and sends them to a server. The server can then integrate the encrypted models uploaded by all user-side devices without decryption, resulting in an integrated model that is sent back to each user-side device. The user-side devices then decrypt the integrated model to obtain the final model.

[0086] Therefore, the symmetric encryption algorithm here differs from existing technology Pailier without specific details: The value is different from Pailier's. Data encryption is done by adding a random number z. The design of μ can ensure that the random number z does not affect the symmetric encryption result during the symmetric encryption process.

[0087] Furthermore, the information entropy and decision tree in steps 7 and 8 are chosen because this combination allows for ensemble learning without decryption. Conventional machine learning methods such as SVM cannot achieve this. In other words, the core innovation of this invention is to achieve ensemble learning even when the model (i.e., equation (1)) is encrypted. Therefore, a decision tree structure is adopted because the decision tree model can be expressed as a set of probability variables, which can be symmetrically encrypted and used for ensemble computation. SVM and other machine learning models do not possess this property.

[0088] Step 5: The cloud server decrypts the ciphertext of the random number z using its private key to obtain the random number z, and then extracts it from E(P) i Subtracting the random number z from +z) yields the ciphertext E(P) corresponding to the original frequency value of the sample data for the i-th attribute category. i ):

[0089]

[0090] That is, divide g by equation (1). z Equation (2) is obtained.

[0091] Step 6: The cloud server gathers the ciphertext corresponding to the original frequency values ​​of sample data from all user-side devices, sums them to obtain an encrypted sum value, and returns the encrypted sum value to each user-side device.

[0092] The cloud server aggregates encrypted P data uploaded by all user-side devices. D 'ciphertext E(P) i +z);

[0093] Assume there are m user-side devices, P ji Let E represent the raw frequency value of the sample data for the i-th attribute category of the j-th (1≤j≤m) user-side device. Then, the ciphertext corresponding to the raw frequency value of the sample data for the i-th attribute category of the j-th (1≤j≤m) user-side device is E. j (P ji According to equation (2), we can obtain r n mod n 2 Summing all m user-side devices yields an encrypted sum value, which is then returned to all user-side devices.

[0094] The summation process is as follows:

[0095]

[0096] This is for encrypting the summation value.

[0097] Step 7: The user-side device decrypts the encrypted summation value to obtain the summation value, and calculates the information gain of each category attribute based on the summation value;

[0098] The decryption process is as follows:

[0099]

[0100] To find the sum.

[0101] The formula for calculating the information gain value IG(i) of the i-th category attribute is:

[0102] IG(i) = Entropy(D) - Entropy i (D) (4)

[0103] in,

[0104]

[0105] D represents all sample data, and the operator || represents the number of sets;

[0106] a l Let be the value of the l-th sample data in the i-th category attribute, 1≤l≤L, where L is the number of sample data in the i-th category attribute;

[0107] Paj For a l The proportion of the number of corresponding sample data to the number of data for the i-th category attribute.

[0108] Step 8: The user-side device trains the decision tree model using the information gain value of each category attribute and the corresponding hierarchical classification of its original sample data as the training set. Specifically, it includes the following steps: Starting from the root node, select the category attribute with the largest attribute gain value from the attribute gain values ​​of each category obtained in Step 7 as the decision tree child node. Create branches according to the different values ​​of the category attribute and label the corresponding hierarchical classification of the branch. Then, select the category attribute with the largest attribute gain value from the remaining category attributes as the next decision tree child node. Repeat the above steps until the remaining category attributes are zero to obtain the final decision tree model.

[0109] Step 9: Use the trained decision tree model to classify and classify local data from user-side devices. For example, electricity payment data includes category attributes such as: name, ID number, user-side device address, user-side device payment account, bank account, electricity consumption, and electricity cost. Based on the information gain values ​​obtained in Steps 2-7, input them into the decision tree model. When the gain value corresponding to the category attribute meets the decision tree's judgment, the classification and grading can be automatically identified and output. For example, if the electricity payment data belongs to the severely sensitive level of user-side device related data, the classification and grading can be automatically identified and output.

[0110] This invention employs a privacy-preserving ensemble decision learning algorithm. Each user-side device extracts a set of category attribute frequencies from its own sample data set, encrypts it, and uploads it to the server. The server integrates and sums the category attribute frequency sets uploaded by all user-side devices without decryption, generating ciphertext corresponding to the integration sum. This ciphertext is then sent back to all user-side devices. After decryption, the user-side devices obtain the integrated category attribute frequency set. This set is used to extract the gain value required for decision tree discrimination, and the decision tree is trained to obtain the final decision tree model. Then, the decision tree model can be used to automatically classify and grade power data based on the content of the category attributes.

[0111] Embodiment 2 of the present invention provides a data privacy-preserving classification and grading system based on ensemble decision learning, the system being used to implement the above-described data privacy-preserving classification and grading method based on ensemble decision learning.

[0112] Embodiment 3 of the present invention provides a terminal, including a processor and a storage medium, wherein the storage medium is used to store instructions; and the processor is used to perform operations according to the instructions to execute the steps of the above method.

[0113] Embodiment 4 of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method.

[0114] The beneficial effects of this invention are compared with those of the prior art:

[0115] This invention integrates user-side device data category attribute parameters required for training a decision tree model by scrambling and symmetric encryption before sending them to a cloud server. The cloud server integrates the category attributes uploaded by all user-side devices without decryption, and then distributes the resulting ciphertext to each user-side device. The user-side devices use the decrypted data to train the ensemble decision tree model. The trained decision tree model is then used for classification and grading of the local data on the user-side devices. By combining a symmetric encryption algorithm with an ensemble decision tree model, this invention achieves data integration in a cloud service scenario without decrypting the encrypted data from user-side devices. This integration is then pushed to each user-side device for local training of the classification and grading model, thus protecting privacy during the distributed automatic classification and grading process of user-side device data in a cloud service scenario.

[0116] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0117] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0118] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0119] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0120] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0121] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0122] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0123] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A data privacy-preserving classification and grading method based on ensemble decision learning, characterized in that: The method includes the following steps: The set of power data samples with known classification and grading for each user-side device is as follows: Define the set corresponding to the category attributes of the sample data as , For the i-th category attribute, calculate the sample data frequency set corresponding to each category attribute. ; User-side device generates random numbers and will Add to On each element, we obtain ; For random numbers Encrypt using the public key of the cloud server, and then use the random number. The encrypted message was sent to the cloud server. The user-side equipment uses a symmetric encryption algorithm to... Encrypt each element in the middle, and Ciphertext of various elements Send to the cloud server; Receive the encrypted summation value returned by the cloud server, which is obtained by summing the ciphertext corresponding to the original values ​​of the sample data frequency of all user-side devices; The user-side device decrypts the encrypted summation value to obtain the summation value, and calculates the information gain of each category attribute based on the summation value; The user-side device trains the decision tree model using the information gain values ​​of each category attribute and the corresponding hierarchical classification of its original sample data as the training set, thus completing the integrated decision learning. The trained decision tree model is used to classify and classify local data from user-side devices.

2. The data privacy-preserving classification and grading method based on ensemble decision learning according to claim 1, characterized in that: Frequency set elements in The value is the user-side device's first The number of sample data corresponding to each category attribute accounts for a certain percentage of the user-side device power sample dataset. The proportion of data volume.

3. The data privacy-preserving classification and grading method based on ensemble decision learning according to claim 1, characterized in that: Select random number Using a symmetric encryption algorithm with public key Apply the following formula medium elements Symmetric encryption yields ciphertext : （1） in, All are public keys for symmetric encryption algorithms. The parameter values ​​in For the remainder calculation formula.

4. The data privacy-preserving classification and grading method based on ensemble decision learning according to claim 3, characterized in that: Generate the public and private keys for the symmetric encryption algorithm as follows: Public Key private key ; in, The Euler totient function value is ; It is a random integer, and Divisible The order; For a randomly selected prime number, Represented as a field of positive integers; 5. The data privacy-preserving classification and grading method based on ensemble decision learning according to claim 4, characterized in that: No. The ciphertext corresponding to the original values ​​of sample data frequencies for each attribute category for: （2）。 6. The data privacy-preserving classification and grading method based on ensemble decision learning according to claim 1, characterized in that: Assume there is Each user-side device Represented as the first The first user-side device The original values ​​of the sample data frequency for each attribute category, the first... The first user-side device The ciphertext corresponding to the original values ​​of the sample data frequencies for each attribute category is: ,but The encrypted summation value is: = （3）。 7. The data privacy-preserving classification and grading method based on ensemble decision learning according to claim 4, characterized in that: The decryption process is as follows: To encrypt the summation value; This is the summation value obtained after decryption.

8. The data privacy-preserving classification and grading method based on ensemble decision learning according to claim 1, characterized in that: No. Information gain value of each category attribute The calculation formula is: （4） in, (5) （6） This is the summation value obtained after decryption; This represents a set of power data samples, and the operator | represents the number of sets. For the first The first user-side device The first category attribute corresponding to the Each sample data value, , L No. The number of sample data corresponding to each category attribute; For the first User-side devices The corresponding number of sample data accounts for the first The proportion of data for each category attribute.

9. The data privacy-preserving classification and grading method based on ensemble decision learning according to claim 1, characterized in that: The user-side device trains a decision tree model using the information gain values ​​of each category attribute and the corresponding hierarchical classification of its original sample data as the training set, thereby completing ensemble decision learning, specifically as follows: Create a root node for the decision tree. Starting from the root node, select the category attribute with the largest attribute gain value from all categories as the child node of the decision tree. Create branches according to the different values ​​of the category attribute and mark the corresponding hierarchical classification of the branch. Then, select the category attribute with the largest attribute gain value from the remaining category attributes as the next child node of the decision tree. Repeat the above steps until the remaining category attributes are zero to obtain the final decision tree model.

10. A data privacy-preserving classification and grading system based on ensemble decision learning, characterized in that: The system is used to implement the data privacy-protected classification and grading method based on ensemble decision learning as described in any one of claims 1-9.

11. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the data privacy-protected classification and grading method based on ensemble decision learning according to any one of claims 1-9.

12. A data privacy-preserving classification and grading method based on ensemble decision learning, characterized in that: Cloud server receives random numbers Ciphertext and Then, for random numbers The encrypted text is decrypted using the private key on the cloud server to obtain a random number. and from Subtracting the random number z from the middle, we get the first... The ciphertext corresponding to the original values ​​of sample data frequencies for each attribute category ;in, for Ciphertext of each element , The power data sample set is The frequency of sample data corresponding to each category attribute; The cloud server aggregates the original frequency values ​​of sample data from all user-side devices, sums them to obtain an encrypted sum value, and then returns the encrypted sum value to each user-side device.

13. A cloud server, characterized in that: The cloud server executes the data privacy-protected classification and grading method based on integrated decision learning as described in claim 12.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-9 or 12.

Images