WO2021068444A1

WO2021068444A1 - Data processing method and device, computer apparatus, and storage medium

Info

Publication number: WO2021068444A1
Application number: PCT/CN2020/076140
Authority: WO
Inventors: 唐虹刚; 李升林; 谢翔; 孙立林
Original assignee: 云图技术有限公司
Priority date: 2019-10-11
Filing date: 2020-02-21
Publication date: 2021-04-15
Also published as: CN110750801B; CN110750801A

Abstract

A data processing method and device, a computer apparatus, and a storage medium. The method comprises: receiving a data processing request sent by a client (S201); in response to the data processing request, calling, according to a preset MPC protocol, a convolution operator in a preset model to perform a convolution operation on ciphertext feature data, and obtaining ciphertext feature data (S202); and sending the ciphertext feature data and a preset operator in the preset model to an agent (S203), wherein the agent and the client call, according to the preset MPC protocol, the preset operator to process the ciphertext feature data, and obtaining a ciphertext processing result. The method enables processing of inputted ciphertext data, thereby effectively protecting data privacy, and saving bandwidth resources.

Description

Data processing method, device, computer equipment and storage medium

This application claims the priority of the Chinese patent application with the application number 201910962389.0 and the invention title of "data processing method, device, computer equipment and storage medium" filed on October 11, 2019, the entire content of which is incorporated into this application by reference .

Technical field

This application relates to the field of artificial intelligence technology, and in particular to a data processing method, device, computer equipment, and storage medium.

Background technique

In some application scenarios, the server can process the target data sent by the client according to the trained deep learning model to obtain the processing result. For example, the server can perform diagnosis processing on the medical image data sent by the client according to the trained diagnosis model, so as to obtain the diagnosis result. Among them, the aforementioned medical image data may often involve user privacy, and the aforementioned methods are prone to leakage of user privacy data during specific data processing.

Therefore, there is an urgent need for a method that can process data while protecting user data privacy.

Summary of the invention

The embodiments of the present application provide a data processing method, device, computer equipment, and storage medium to provide a data processing method that can protect user privacy.

The embodiment of the application provides a data processing method, including: receiving a data processing request sent by a client, wherein the data processing request carries ciphertext data, and the ciphertext data is processed by the client using a homomorphic encryption algorithm on the target data. Generated after encryption; in response to the data processing request, according to the preset MPC protocol, the convolution operator in the preset model is called to perform the convolution operation on the ciphertext data to obtain the ciphertext feature data; combine the ciphertext feature data with the preset The preset operator in the model is sent to the agent, where the agent is used to call the preset operator with the client according to the preset MPC protocol to process the ciphertext feature data to obtain the ciphertext processing result; wherein, the preset The model is compatible with the preset MPC protocol and is used to perform preset processing on ciphertext data under data encryption; the agent includes one of the following: a secure container running on the client, a virtual machine running on the client, A computer that establishes a local area network connection with the client.

The embodiment of the present application also provides a data processing method, which is applied to the agent, and includes: receiving the ciphertext feature data sent by the server and the preset operator in the preset model, wherein the ciphertext feature data is preset by the server according to The MPC protocol calls the convolution operator in the preset model to perform convolution operation on the ciphertext data. The ciphertext data is obtained by encrypting the target data by the client using a homomorphic encryption algorithm; according to the preset MPC protocol Call the preset operator in the preset model to process the ciphertext feature data in cooperation with the client to obtain the ciphertext processing result; send the ciphertext processing result to the client; where the preset model is compatible with the preset MPC Protocol, used to perform preset processing on ciphertext data under data encryption; the agent includes one of the following: a secure container running on the client, a virtual machine running on the client, and establishing a local area network connection with the client Computer.

The embodiment of the present application also provides a data processing method, including: a client obtains target data, and encrypts the target data using a homomorphic encryption algorithm to obtain ciphertext data; the client sends a data processing request to the server, where the data The processing request carries the ciphertext data; in response to the data processing request, the server and the client call the convolution operator in the preset model according to the preset MPC protocol to perform the convolution operation on the ciphertext data to obtain the ciphertext feature data; The server sends the ciphertext feature data and the preset operator in the preset model to the agent; the agent calls the preset operator according to the preset MPC protocol to process the ciphertext feature data in cooperation with the client to obtain the ciphertext feature data Text processing result; the agent sends the ciphertext processing result to the client; the client decrypts the ciphertext processing result to obtain the plaintext processing result of the target data; among them, the preset model is compatible with the preset MPC protocol, which is used in the data The ciphertext data is pre-processed in the case of encryption; the agent includes one of the following: a secure container running on the client, a virtual machine running on the client, and a computer that establishes a local area network connection with the client.

An embodiment of the present application also provides a data processing device, including: a receiving module for receiving a data processing request sent by a client, wherein the data processing request carries ciphertext data, and the ciphertext data is homomorphic by the client The encryption algorithm encrypts the target data and generates it; the convolution module is used to respond to the data processing request, according to the preset MPC protocol, call the convolution operator in the preset model to perform the convolution operation on the ciphertext data, and obtain the ciphertext data. Text feature data; the sending module is used to send the cipher text feature data and the preset operator in the preset model to the agent, where the agent is used to call the preset operator pair with the client according to the preset MPC protocol The ciphertext feature data is processed to obtain the ciphertext processing result; among them, the preset model is compatible with the preset MPC protocol, which is used to perform preset processing on the ciphertext data in the case of data encryption; the agent includes one of the following: Run A secure container on the client, a virtual machine running on the client, and a computer that establishes a LAN connection with the client.

An embodiment of the present application further provides a computer device including a processor and a memory for storing executable instructions of the processor, and the processor implements the steps of the data processing method described in any of the foregoing embodiments when the processor executes the instructions.

The embodiments of the present application also provide a computer-readable storage medium on which computer instructions are stored, and when the instructions are executed, the steps of the data processing method described in any of the foregoing embodiments are implemented.

In the embodiment of the present application, a data processing method is provided. The client uses a homomorphic encryption algorithm to encrypt target data to generate ciphertext data, and the server receives a data processing request carrying ciphertext data sent by the client, and In response to the data processing request, according to the preset MPC protocol, the convolution operator in the preset model is called to perform the convolution operation on the ciphertext data to obtain the ciphertext feature data. The server combines the ciphertext feature data with the preset model The preset operator is sent to the agent, and the agent and the client call the preset operator according to the preset MPC protocol to process the ciphertext feature data, and obtain the ciphertext processing result. Among them, the preset model is compatible with the preset MPC protocol, and is used to perform preset processing on the ciphertext data in the case of data encryption. In the above solution, the target data is homomorphically encrypted to obtain the corresponding ciphertext data, so as to prevent the specific content of the target data from being leaked. Further, the server calls the convolution operator in the preset model compatible with the MPC protocol to perform the convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data, and combine the ciphertext feature data with the preset model The preset operator of is sent to the agent, so that the agent and the client then call the preset operator in the preset model to perform specific processing on the ciphertext feature data according to the preset MPC protocol to obtain the ciphertext processing of the target data result. Realize that the target data can be specifically processed in the state of target data encryption, thereby effectively protecting data privacy. Furthermore, since the agent is a computer in the same local area network as the client or a secure container or virtual machine running on the client, it is more secure than directly having the server and the client perform ciphertext feature data. Calculation, through the agent and the client according to the preset MPC protocol to call the preset operator to perform secure multi-party calculation of the ciphertext feature data, which can effectively save the network bandwidth required for secure multi-party calculation, and greatly reduce the network delay. Effectively improve the efficiency of data processing.

Description of the drawings

The drawings described here are used to provide a further understanding of the application, constitute a part of the application, and do not constitute a limitation to the application. In the attached picture:

Figure 1 shows a schematic diagram of an application scenario of a data processing method in some embodiments of the present application;

Figure 2 shows a flowchart of a data processing method in some embodiments of the present application;

Figure 3 shows a flowchart of a data processing method in some embodiments of the present application;

Figure 4 shows a flowchart of a data processing method in some embodiments of the present application;

Figure 5 shows a sequence diagram of a data processing method in some embodiments of the present application;

Fig. 6 shows a schematic diagram of a data processing device in some embodiments of the present application;

Fig. 7 shows a schematic diagram of a data processing device in some embodiments of the present application;

Fig. 8 shows a schematic diagram of a computer device in some embodiments of the present application.

Detailed ways

The principle and spirit of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are only provided to enable those skilled in the art to better understand and then implement the application, but not to limit the scope of the application in any way. On the contrary, these embodiments are provided to make the disclosure of the present application more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

Those skilled in the art know that the implementation manners of this application can be implemented as a system, apparatus, method, or computer program product. Therefore, the disclosure of the present application can be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

Some embodiments of the present application provide a data processing method, and FIG. 1 shows a schematic diagram of an application scenario of the data processing method in an embodiment of the present application. As shown in Figure 1, the client can obtain target data. Among them, the target data may be data to be processed. For example, the target data may include but is not limited to at least one of the following: face image data, medical image data, human body motion data, and so on. The client can use the homomorphic encryption algorithm to encrypt the obtained target data to obtain ciphertext data. Among them, the data encrypted by the homomorphic encryption algorithm is processed to obtain an output, and the output is decrypted, and the result is the same as the output result obtained by processing the unencrypted original data in the same method.

The client can send a data processing request to the server. Among them, the data processing request may carry ciphertext data corresponding to the target data. In response to the data processing request, the server may call a convolution operator in the preset model to perform a convolution operation on the ciphertext data according to a preset MPC (Secure Multi-Party Computation) protocol to obtain the ciphertext feature data.

The server sends the ciphertext feature data and the preset operator in the preset model to the agent. The agent and the client call the preset operator according to the preset MPC protocol to process the ciphertext feature data, and obtain the ciphertext processing result. Wherein, the preset operator may be an operator specified in the preset model. For example, the preset operator may include but is not limited to at least one of the following: an activation operator, a pooling operator, and a fully connected operator. Among them, the preset model is on the server side. The preset model can be compatible with the preset MPC protocol, and is used for preset processing of ciphertext data in the case of data encryption. In other words, the preset MPC protocol can be used to implement the operations in the preset model when the input is ciphertext data, so as to complete the preset processing of the ciphertext data. Among them, the preset model may include a trained preset deep learning model. For example, the preset model may be a trained convolutional neural network model. In the case where the target data is medical image data, the preset processing may include performing diagnosis processing on the ciphertext data obtained after homomorphic encryption of the medical image data. In the case where the target data is face image data, the preset processing may include performing face recognition processing on the ciphertext data obtained after homomorphic encryption of the face image data. In the case where the target data is human body motion data, the preset processing may include performing motion recognition processing on the ciphertext data obtained after homomorphic encryption of the human body motion data.

In the case where the preset model includes multiple convolutional layers, the data obtained after processing the first ciphertext feature data on the agent and the client may need to perform a convolution operation through the second convolutional layer. The first ciphertext feature data may be obtained after the server calls the convolution operator of the first convolution layer to perform a convolution operation on the ciphertext data according to a preset MPC protocol. Therefore, the agent can send the processed data to the server. The server invokes the convolution operator of the second convolution layer to perform a convolution operation on the received data according to the preset MPC protocol to obtain the second ciphertext feature data. The server then sends the second ciphertext feature data to the agent, so that the agent and the client call the preset operator to process the second ciphertext feature data according to the preset MPC protocol, and so on, until the agent obtains the ciphertext feature data. Text processing results.

After the ciphertext processing result is obtained by the agent, the agent can send the ciphertext processing result to the client. The public key and private key generated by the homomorphic encryption algorithm are stored in the client. Among them, the public key is used to homomorphically encrypt the target data to obtain ciphertext data. The private key is used to decrypt the received ciphertext processing result to obtain the plaintext processing result.

Take a specific application scenario as an example for description. The client may be a client of a medical institution. The client can obtain the patient's medical image data (for example, X-ray film data). The client can use the homomorphic encryption algorithm to encrypt the X-ray film data to obtain the ciphertext X-ray film data. The client sends a diagnosis request to the medical optical film diagnosis server. Among them, the diagnosis request carries ciphertext X-ray film data. The server may be a medical optical film diagnosis server. The medical optical film diagnosis server calls the convolution operator in the preset deep learning model to perform the convolution operation on the ciphertext feature data according to the preset MPC protocol to obtain the ciphertext feature data. The medical optical film diagnosis server sends the ciphertext feature data and the preset operator in the preset model to the agent. Among them, the agent is a computer on the same intranet as the client or a secure container or virtual machine running on the client. The agent and the client call the preset operator according to the preset MPC protocol to process the ciphertext feature data, so as to obtain the ciphertext diagnosis result. The agent sends the ciphertext diagnosis result to the client. The client decrypts the ciphertext diagnosis result, and obtains the plaintext diagnosis result of the X-ray film data.

Among them, the above-mentioned client can be a desktop computer, notebook, mobile phone terminal, PDA, etc., as long as it is a device that can homomorphically encrypt the target data and cooperate with the agent to perform preset processing on the ciphertext data. The presentation form of the client , This application is not limited. The above-mentioned server can be a single server, a server cluster, or a cloud server, etc., as long as the device can perform convolution operations on ciphertext data according to the preset MPC protocol and preset model. The specific The composition form of this application is also not limited. The above-mentioned agent can be a computer that establishes a LAN connection with the client, a secure container running on the client, or a virtual machine running on the client, etc., as long as it can communicate with the server and can perform secure multi-party computing with the client. All are possible, and this application does not limit the presentation form of the agent.

Figure 2 shows a flowchart of a method in some embodiments of the present application, and the data processing method can be applied to the client. Although this application provides method operation steps or device structures as shown in the following embodiments or drawings, the method or device may include more or less operation steps or module units based on conventional or no creative labor. . In steps or structures where there is no necessary causal relationship logically, the execution order of these steps or the module structure of the device is not limited to the execution order or module structure shown in the description of the embodiments of this application and the drawings. When the described method or module structure is applied to an actual device or terminal product, it can be executed sequentially or in parallel according to the method or module structure connection shown in the embodiments or drawings (for example, parallel processors or multi-threaded processing Environment, even distributed processing environment).

Specifically, as shown in FIG. 2, the data processing method provided by some embodiments of the present application may include the following steps:

Step S201: Receive a data processing request sent by a client, where the data processing request carries ciphertext data, and the ciphertext data is generated by the client after encrypting target data using a homomorphic encryption algorithm.

The client can obtain the target data. Among them, the target data is the data to be processed. The client uses a homomorphic encryption algorithm to encrypt the target data to obtain ciphertext data. Specifically, the client can use a homomorphic encryption algorithm to generate a public key and a private key, and then use the public key to encrypt the target data to obtain ciphertext data. The client can send a data processing request to the server. Among them, the data processing request carries ciphertext data. In other embodiments, the data processing request may also carry the public key generated by the homomorphic encryption algorithm. The server receives the data processing request sent by the client.

Step S202: In response to the data processing request, according to the preset MPC protocol, the convolution operator in the preset model is called to perform a convolution operation on the ciphertext data to obtain the ciphertext feature data.

Step S203: Send the ciphertext feature data and the preset operator in the preset model to the agent, where the agent is used to call the preset operator with the client according to the preset MPC protocol to process the ciphertext feature data , Get the ciphertext processing result.

In response to the data processing request, the server calls the convolution operator in the preset model to perform a convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data. The server sends the obtained ciphertext feature data and the preset operator in the preset model to the agent. The agent and the client call the preset operator according to the preset MPC protocol to process the ciphertext feature data, and obtain the ciphertext processing result. Among them, the preset model is compatible with the preset MPC protocol, which is used to perform preset processing on ciphertext data in the case of data encryption; the agent includes one of the following: a secure container running on the client, and a secure container running on the client A virtual machine, a computer that establishes a LAN connection with the client.

Among them, the preset model may be a trained deep learning model. Among them, the deep learning model can be obtained by training with a large amount of plaintext data. In order to be compatible with the preset MPC protocol, the floating-point data in the trained deep learning model is converted into integer data. In other words, the preset MPC protocol can be used to implement various operations in the preset model when the input is ciphertext data, so as to realize the preset processing of ciphertext data. After obtaining the ciphertext processing result at the agent, the ciphertext processing result can be sent to the client. The client can decrypt the ciphertext processing result according to the private key generated by the homomorphic encryption algorithm to obtain the plaintext processing result of the target data.

In the data processing method in the foregoing embodiment, the target data is homomorphically encrypted to obtain the corresponding ciphertext data, so as to prevent the specific content of the target data from being leaked. Further, the server calls the convolution operator in the preset model compatible with the MPC protocol to perform the convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data, and combine the ciphertext feature data with the preset model The preset operator of is sent to the agent, so that the agent and the client then call the preset operator in the preset model to perform specific processing on the ciphertext feature data according to the preset MPC protocol to obtain the ciphertext processing of the target data result. Realize that the target data can be specifically processed in the state of target data encryption, thereby effectively protecting data privacy. Furthermore, since the agent is a computer in the same local area network as the client or a secure container or virtual machine running on the client, it is more secure than directly having the server and the client perform ciphertext feature data. Calculation, through the agent and the client according to the preset MPC protocol to call the preset operator to perform secure multi-party calculation of the ciphertext feature data, which can effectively save the network bandwidth required for secure multi-party calculation, and greatly reduce the network delay. Effectively improve the efficiency of data processing.

In some embodiments of the present application, when the preset model includes multiple convolutional layers, when the agent and the client call the preset operator according to the preset MPC protocol to process the ciphertext feature data, The obtained data that needs to be convolved is returned to the server. The server performs a convolution operation on the received data and sends the obtained data to the agent, and so on until the ciphertext processing result is obtained.

In some embodiments of the present application, after the agent and the client call the preset operator according to the preset MPC protocol to process the ciphertext feature data, and obtain the ciphertext processing result, it may further include: corresponding to the data processing request The transaction events are uploaded to the blockchain network for storage on the chain.

Exemplarily, after the agent and the client cooperate to complete the processing of the ciphertext data, the server may be notified that the preset processing of the ciphertext data has been completed. In response to the notification, the server uploads the transaction event data corresponding to the data processing request to the blockchain network for on-chain storage. The transaction event data may include data related to calculations performed by the server in response to the data processing request. Through the above method, the transaction event data is uploaded to the blockchain network for on-chain storage, which is convenient for subsequent query and backtracking.

In some embodiments of the present application, the target data may include but is not limited to at least one of the following: face image data, medical image data, and human body motion data. In some embodiments of the present application, when the target data includes medical image data, the preset processing may include performing diagnosis processing on the medical image data. When the target data includes face image data, the preset processing may include face recognition processing. When the target data includes human body motion data, the preset processing includes motion recognition processing. The above-mentioned embodiments are only exemplary, and the target data that can be processed by this application and the preset processing that can be performed are not limited thereto.

In some embodiments of the present application, according to the preset MPC protocol, invoking the convolution operator in the preset model to perform the convolution operation on the ciphertext data to obtain the ciphertext feature data may include: according to the preset MPC protocol With the homomorphic addition and multiplication characteristics of, the convolution operator in the preset model is called to perform the convolution operation on the ciphertext data to obtain the ciphertext feature data.

Specifically, the preset model may be a trained convolutional neural network model. Convolutional neural network models generally include: convolutional layer, excitation layer, pooling layer and fully connected layer. Among them, the number of each layer can be one or more. Among them, the convolution layer may include a convolution operator for extracting features of the data. After receiving the data processing request sent by the client, the server calls the convolution operator in the preset model to perform a private convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data. Among them, the plaintext feature data obtained after decrypting the ciphertext feature data obtained by performing the privacy convolution operation on the ciphertext data is the same as the feature data obtained by directly performing the plaintext convolution operation on the plaintext data. Among them, homomorphic encryption allows operations to be performed on the ciphertext, and the corresponding operations will be mapped to the plaintext. The homomorphic encryption can be refined into the following two homomorphic encryption algorithms:

1) Homomorphic addition

Given that C1=Enc(m1), C2=Enc(m2), then C1·C2=Enc(m1+m2);

2) Homomorphic multiplication

Given that C1=Enc(m1), t1, then (C1) ^t1 =Enc(m1·t1), which can also be derived by homomorphic addition.

The convolution operation on the plaintext data is to multiply the corresponding elements in the plaintext data matrix and the convolution operator matrix and then sum them. The convolution operation on the ciphertext data is obtained by multiplying the powers of the corresponding elements of the convolution operator matrix of the elements in the ciphertext data matrix. Exemplarily, A _3×3 is a plaintext data matrix, C _3×3 is a convolution operator matrix; a _3×3 is a ciphertext data matrix. The plaintext convolution operation and privacy convolution operation are as follows:

Some embodiments of the present application also provide a data processing method, which can be applied to an agent. Fig. 3 shows a flowchart of a data processing method provided by an embodiment of the present application.

As shown in Figure 3, the foregoing data processing method may include the following steps:

Step S301: Receive the ciphertext feature data and the preset operator in the preset model sent by the server.

Specifically, the server may receive the data processing request sent by the client. Among them, the data processing request may carry ciphertext data. The ciphertext data can be obtained by encrypting the target data by the client using a homomorphic encryption algorithm. The server may call the convolution operator in the preset model according to the preset MPC protocol in response to the data processing request to perform a convolution operation on the ciphertext data to obtain the ciphertext feature data. The server sends the generated ciphertext feature data and the preset operator in the preset model to the agent. Among them, the agent includes one of the following: a secure container running on the client, a virtual machine running on the client, and a computer that establishes a local area network connection with the client.

Step S302: Call the preset operator in the preset model according to the preset MPC protocol to process the ciphertext feature data in cooperation with the client to obtain the ciphertext processing result.

Step S303: Send the ciphertext processing result to the client.

After receiving the ciphertext feature data and the preset operator, the agent can call the preset operator in the preset model according to the preset MPC protocol, and cooperate with the client to process the ciphertext feature data to obtain the ciphertext processing result . Among them, the preset model is compatible with the preset MPC protocol, and is used to perform preset processing on the ciphertext data in the case of data encryption. Among them, the preset model may be a trained deep learning model. Among them, the deep learning model can be obtained by training using a large amount of plaintext data. In order to be compatible with the preset MPC protocol, the floating-point data in the trained deep learning model is converted into integer data. In other words, the preset MPC protocol can be used to implement various operations in the preset model when the input is ciphertext data, so as to realize the preset processing of ciphertext data. After obtaining the ciphertext processing result, the agent sends the ciphertext processing result to the client.

In some embodiments of the present application, after sending the ciphertext processing result to the client, it may further include: the client decrypts the ciphertext processing result to obtain the plaintext processing result. Specifically, the client can decrypt the ciphertext processing result according to the private key obtained by the homomorphic encryption algorithm to obtain the plaintext processing result of the target data. Through the above method, the plaintext processing result of the target data can be revealed on the client.

In some embodiments of the present application, after the client decrypts the ciphertext processing result to obtain the plaintext processing result, it may further include: the client obtains the verification result, where the verification result is used to indicate whether the plaintext processing result is correct; the client The verification result and the ciphertext processing result are sent to the blockchain network for on-chain storage.

Specifically, after the client decrypts the ciphertext processing result to obtain the plaintext processing result, the client obtains the verification result of the plaintext processing result. Among them, the verification result may be obtained after the business personnel verify the plaintext processing result. The verification result can also be inferred by the application in the client based on the target data and the plaintext processing result. The verification result is used to indicate whether the plaintext processing result is correct. The client sends the verification result and the ciphertext processing result to the blockchain network for on-chain storage, which is convenient for subsequent query, traceability, billing, verification, etc.

In some embodiments of the present application, before/while/after the client sends a data processing request to the server, the client may send the data processing resource data to the contract account of the blockchain network. Among them, the resource data may be fees that need to be paid for data processing, such as currency or Bitcoin. After the client sends the ciphertext processing result and the verification result to the blockchain network for on-chain storage, the blockchain network determines whether the verification result is correct. When it is determined that the verification result is correct, the blockchain network sends the resource data in the contract account to the server's account. Through the above method, the problem of service billing and settlement can be solved.

In some embodiments of the present application, calling the preset operator in the preset model according to the preset MPC protocol to process the ciphertext feature data in cooperation with the client to obtain the ciphertext processing result, which may include: The activation operator in the MPC protocol and the preset operator cooperates with the client to activate the ciphertext feature data to obtain the first activation data and the second activation data; according to the preset MPC protocol and the preset operator The pooling operator of the client cooperates with the client to perform the pooling operation on the first activation data and the second activation data to obtain the first pooling data and the second pooling data; according to the preset MPC protocol and the preset operator The fully connected operator cooperates with the client to perform a fully connected operation on the first pooled data and the second pooled data to obtain the ciphertext processing data.

Among them, the preset model may be a trained convolutional neural network model. Convolutional neural network models generally include: convolutional layer, excitation layer, pooling layer and fully connected layer. Among them, the number of each layer can be one or more. Among them, the convolution layer may include a convolution operator for extracting features of the data. The excitation layer can include activation operators to add non-linear factors to simulate more subtle changes. The pooling layer may include a pooling operator, which is used to compress the input features. On the one hand, the feature map is reduced and the network calculation complexity is simplified. On the other hand, the feature compression is performed to extract the main features. The fully connected layer may include a fully connected operator, which is used to connect all the features, send the output value to the classifier, and obtain the processing result. Among them, the operation corresponding to the paper operator requires more computing resources, so it can be completed on the server side. For incentives, pooling and fully connected operations that occupy less computing resources, the agent and the client can call the activation operator, the pooling operator and the fully connected operator according to a preset protocol to perform secure multi-party calculations. The preset operator may include but is not limited to at least one of the following: an activation operator, a pooling operator, and a fully connected operator.

After the server calls the convolution operator to perform the convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data, the server sends the ciphertext feature data and the preset operator to the agent. After obtaining the ciphertext feature data, the agent can call the activation operator in the preset operator according to the preset MPC protocol, and cooperate with the client to activate the ciphertext feature data to obtain the first activation data and the second activation data. Among them, the first activation data is held by the agent, and the second activation data is held by the client. Specifically, the agent performs a homomorphic encryption to secret sharing operation on the ciphertext feature data to obtain the first ciphertext data and the second ciphertext data. The agent sends the first ciphertext data to the client. The client decrypts the first ciphertext data to obtain the first plaintext data. The agent converts the activation operator into a Boolean circuit and obtains the truth table of each circuit gate of the Boolean circuit. Then the agent selects two evenly distributed strings as labels for each door, which are used to represent 0 or 1 respectively. After that, the agent randomly generates a first random number, and generates a first label according to the second ciphertext data and the first random number. After replacing the truth table with the first label as input, the agent obtains the label table. For each circuit gate, the agent uses the tag table to encrypt, and generates a first encryption circuit (GC, Garbled Circuit). The agent sends the first encryption circuit to the client. The client performs an inadvertent transfer (OT, Oblivious Transfer) to the first plaintext data to obtain the second label. The client decrypts the first encryption circuit according to the second tag to obtain activation data. Activation data is also feature data, which is used to characterize the features after adding nonlinear factors, and can simulate more subtle feature changes. Then, since the subsequent pooling operation is implemented through secure multi-party computing, the activation data can be secretly shared, so that the first activation data can be obtained on the agent side and the second activation data can be obtained on the client side. The activation data is the sum of the first activation data and the second activation data. That is, the agent and the client perform secure multi-party calculation on the ciphertext feature data according to the preset MPC protocol and activation operator to obtain the first activation data and the second activation data.

After the agent obtains the first activation data and the client obtains the second activation data, the agent calls the pooling operator in the preset operator according to the preset MPC protocol, and cooperates with the client to compare the first activation data and the second activation data. Activate the data to perform a pooling operation to obtain the first pooled data and the second pooled data. Among them, the first pooled data can be held by the agent. The second pooled data can be held by the client. Specifically, the agent converts the pooling operator into a Boolean circuit. The agent randomly generates a second random number. The agent generates a third label according to the first activation function and the random number. The agent uses the third label as input to replace the truth table of the Boolean circuit to obtain the label table. The agent uses the tag table to encrypt each gate of the Boolean circuit to generate a second encryption circuit. The agent sends the second encryption circuit to the client. The client inadvertently transmits the second activation data to obtain the fourth tag. The client uses the fourth tag to decrypt the second encryption circuit to obtain pooled data. Pooled data also belongs to feature data, which is the feature data obtained after the feature is compressed, which can characterize the main feature. After that, since the subsequent full connection operation is implemented through secure multi-party computing, the pooled data can be secretly shared, so that the agent obtains the first pooled data and the client obtains the second pooled data. The pooled data is the sum of the first pooled data and the second pooled data. That is, the agent and the client can perform secure multi-party calculation on the first activation data and the second activation data according to the preset MPC protocol and pooling operator to obtain the first pooling data and the second pooling data.

After the agent obtains the first pooled data and the client obtains the second pooled data, the agent performs the secret sharing to homomorphic encryption operation on the first pooled data, and the client performs secret sharing to the second pooled data. State encryption operation, so as to obtain pooled data on the agent side. Then the obtained pooled data is flattened into a one-dimensional array. The agent side multiplies each element in the one-dimensional array with the weight value on the connection line (that is, the element in the fully connected operator), adds and outputs the result, and obtains the ciphertext processing result. That is, the server cooperates with the client to perform a full connection operation on the first pooled data and the second pooled data according to the preset MPC protocol and the full connection operator in the preset operator to obtain the ciphertext processing data. The agent sends the obtained ciphertext processing result to the client. The client uses the private key of the homomorphic encryption algorithm to decrypt the ciphertext processing result to obtain the plaintext processing result of the target data.

In the data processing method of the above embodiment, the agent and the client call the activation operator, the pooling operator and the fully connected operator in the preset operator according to the preset MPC protocol to perform secure multi-party calculation of the ciphertext feature data , The ciphertext processing result is obtained, and the ciphertext data is processed or inferred through the preset model when the input is ciphertext, which can effectively protect the user's data privacy.

In the data processing method in the foregoing embodiment, the activation operator, the pooling operator, and the fully connected operator are all calculated by the agent and the client according to the preset MPC protocol, but the application is not limited to this. In some other embodiments of the present application, the calculation of operators that occupy more computing resources may be calculated by the server and the client according to a preset MPC protocol, such as an average pooling operator.

Some embodiments of the application also provide a data processing method. Fig. 4 shows a flowchart of a data processing method provided in an embodiment of the present application. As shown in Figure 4, the method may include:

In step S401, the client obtains the target data, and encrypts the target data using a homomorphic encryption algorithm to obtain ciphertext data.

Step S402: The client sends a data processing request to the server, where the data processing request carries ciphertext data.

Step S403: In response to the data processing request, the server and the client call the convolution operator in the preset model to perform a convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data.

In step S404, the server sends the ciphertext feature data and the preset operator in the preset model to the agent.

In step S405, the agent invokes a preset operator according to the preset MPC protocol to process the ciphertext feature data in cooperation with the client to obtain the ciphertext processing result.

In step S406, the agent sends the ciphertext processing result to the client.

In step S407, the client decrypts the ciphertext processing result to obtain the plaintext processing result of the target data.

In the data processing method in the foregoing embodiment, the target data is homomorphically encrypted to obtain the corresponding ciphertext data, so as to prevent the specific content of the target data from being leaked. Further, the server calls the convolution operator in the preset model compatible with the MPC protocol to perform the convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data, and combine the ciphertext feature data with the preset model The preset operator of is sent to the agent, so that the agent and the client then call the preset operator in the preset model to perform specific processing on the ciphertext feature data according to the preset MPC protocol to obtain the ciphertext processing of the target data As a result, the client can decrypt the ciphertext processing result to obtain the plaintext processing result of the target data. Realize that the target data can be specifically processed in the state of target data encryption, thereby effectively protecting data privacy. Furthermore, since the agent is a computer in the same local area network as the client or a secure container or virtual machine running on the client, it is more secure than directly having the server and the client perform ciphertext feature data. Calculation, through the agent and the client according to the preset MPC protocol to call the preset operator to perform secure multi-party calculation of the ciphertext feature data, which can effectively save the network bandwidth required for secure multi-party calculation, and greatly reduce the network delay. Effectively improve the efficiency of data processing.

In some embodiments of the present application, after the agent and the client call a preset operator according to the preset MPC protocol to process the ciphertext feature data, and after obtaining the ciphertext processing result, it may also include: the server will contact the data processing request The corresponding transaction event is uploaded to the blockchain network for storage on the chain.

In some embodiments of the present application, the target data may include but is not limited to at least one of the following: face image data, medical image data, and human body motion data.

In some embodiments of the present application, when the target data includes medical image data, the preset processing may include performing diagnosis processing on the medical image data.

In some embodiments of the present application, the server calls the convolution operator in the preset model to perform the convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data, which may include: the server according to the preset MPC The homomorphic addition and multiplication features in the protocol call the convolution operator in the preset model to perform the convolution operation on the ciphertext data to obtain the ciphertext feature data.

In some embodiments of the present application, the agent invokes the preset operator according to the preset MPC protocol to process the ciphertext feature data in cooperation with the client to obtain the ciphertext processing result, which may include: according to the preset MPC protocol The activation operator in the preset operator cooperates with the client to activate the ciphertext feature data to obtain the first activation data and the second activation data; according to the preset MPC protocol and the pooling in the preset operator Operator, cooperates with the client to perform pooling operations on the first activation data and the second activation data to obtain the first pooled data and the second pooled data; according to the preset MPC protocol and the full connection in the preset operator The operator cooperates with the client to perform a full connection operation on the first pooled data and the second pooled data to obtain the ciphertext processing data.

The above method will be described below in conjunction with a specific embodiment. However, it is worth noting that this specific embodiment is only for better describing the application, and does not constitute an improper limitation of the application.

Fig. 5 shows a sequence diagram of the data processing method in this specific embodiment. As shown in Figure 5, the data processing method may include the following steps:

Step 1. The client obtains the target data, and encrypts the target data using a homomorphic encryption algorithm to obtain the ciphertext data corresponding to the target data;

Step 2: The client sends a data processing request to the server, where the data processing request carries ciphertext data;

Step 3. The client mortgages the fee to the contract account of the blockchain network;

Step 4: The server invokes the convolution operator in the preset model to perform a private convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data;

Step 5. The server sends the ciphertext feature data and the preset operator in the preset model to the agent;

Step 6. The agent and the client call the activation operator in the preset operator according to the preset MPC protocol to perform secure multi-party calculation on the ciphertext feature data, and obtain the first activation data at the agent and the second activation at the client. Activation data;

Step 7. The agent and the client call the pooling operator in the preset operator according to the preset MPC protocol to perform a secure multi-party calculation on the first activation data and the second activation data, and obtain the first pooling data at the agent. And get the second pooled data on the client;

Step 8. The agent and the client call the full connection operator in the preset model according to the preset MPC protocol to perform the full connection operation on the first pooled data and the second pooled data, and the ciphertext processing is obtained on the agent side. result;

Step 9. The agent sends the ciphertext processing result to the client;

Step 10: The server sends the transaction event data corresponding to the data processing request to the blockchain network for storage on the chain;

Step 11. The client uses the private key of the homomorphic encryption algorithm to decrypt the ciphertext processing result to obtain the plaintext processing result of the target data;

Step 12. The client obtains the verification result, where the verification result is used to indicate whether the plaintext processing result of the target data is correct;

Step 13. The client sends the plaintext processing result and the verification result to the blockchain network for storage on the chain;

Step 14. The blockchain network determines whether the verification result is correct;

Step 15. In the case where it is determined that the verification result is correct, the blockchain network transfers the fees in the contract account to the server's account;

Step 16. In the case where it is determined that the verification result is incorrect, the blockchain network transfers the fees in the contract account to the client's account.

In the data processing method in the foregoing embodiment, the ciphertext data is obtained by homomorphic encryption of the target data, which facilitates subsequent direct processing of the ciphertext data, and prevents the specific content of the target data from being leaked. The server calls the convolution operator in the preset model to perform the convolution operation on the ciphertext data according to the preset MPC protocol, obtains the ciphertext feature data, and sends the ciphertext feature data and the preset operator in the preset model to Agent side. The agent and the client call the activation operator, pooling operator and fully connected operator in the preset operator according to the preset MPC protocol, and perform preset processing on the ciphertext feature data to obtain the ciphertext processing result of the target data , Which realizes the processing of data when the input is ciphertext data, which can effectively protect data privacy. After that, the server sends the data corresponding to this transaction event to the blockchain network for storage on the chain, so as to facilitate subsequent query, retrieval, verification, etc. The agent sends the ciphertext processing result to the client, and the client decrypts the ciphertext processing result to obtain the plaintext processing result of the target data. The client obtains the verification result of the plaintext processing result, and sends the ciphertext processing result and the verification result to the blockchain network for on-chain storage, which can further protect the privacy of the processing result and facilitate subsequent queries and backtracking. And in the case that the blockchain network determines that the verification result is correct, the cost of the client's mortgage in the contract account is transferred to the server's account. In the case that the blockchain network determines that the verification result is wrong, the cost of the client's mortgage to the contract account is returned to the client's account. In this way, the billing and clearing problem of the business can be effectively solved. Furthermore, since the agent is a computer in the same local area network as the client or a secure container or virtual machine running on the client, it is more secure than directly having the server and the client perform ciphertext feature data. Calculation, through the agent and the client according to the preset MPC protocol to call the preset operator to perform secure multi-party calculation of the ciphertext feature data, which can effectively save the network bandwidth required for secure multi-party calculation, and greatly reduce the network delay. Effectively improve the efficiency of data processing.

Based on the same inventive concept, an embodiment of the present application also provides a data processing device located on the server side, as described in the following embodiment. Since the problem-solving principle of the data processing device is similar to that of the data processing method, the implementation of the data processing device can refer to the implementation of the data processing method, and the repetition will not be repeated. As used below, the term "unit" or "module" can be a combination of software and/or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented by software, implementation by hardware or a combination of software and hardware is also possible and conceived. Fig. 6 is a structural block diagram of a data processing device according to an embodiment of the present application. As shown in Fig. 6, it includes: a receiving module 601, a convolution module 602, and a sending module 603. The structure is described below.

The receiving module 601 is configured to receive a data processing request sent by the client, where the data processing request carries ciphertext data, and the ciphertext data is generated by the client after encrypting the target data using a homomorphic encryption algorithm;

The convolution module 602 is configured to respond to the data processing request and call the convolution operator in the preset model to perform a convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data;

The sending module 603 is used to send the ciphertext feature data and the preset operator in the preset model to the agent, where the agent is used to call the preset operator to the ciphertext feature data according to the preset MPC protocol with the client. Perform processing to obtain the ciphertext processing result; among them, the preset model is compatible with the preset MPC protocol, and is used to perform preset processing on the ciphertext data in the case of data encryption; the agent includes one of the following: running on the client A secure container, a virtual machine running on the client, and a computer that establishes a local area network connection with the client.

In some embodiments of the present application, the data processing device may further include a storage module, and the storage module may be used for: invoking a preset operator on the agent side and the client side according to a preset MPC protocol to process the ciphertext feature data, After the ciphertext processing result is obtained, the transaction event corresponding to the data processing request is uploaded to the blockchain network for storage on the chain.

In some embodiments of the present application, the target data may include at least one of the following: face image data, medical image data, and human body motion data.

In some embodiments of the present application, the convolution module may be specifically used to: according to the homomorphic addition and multiplication characteristics in the preset MPC protocol, call the convolution operator in the preset model to perform convolution operations on the ciphertext data, Obtain ciphertext feature data.

Based on the same inventive concept, an embodiment of the present application also provides a data processing device located at the agent end, as described in the following embodiment. Since the problem-solving principle of the data processing device is similar to that of the data processing method, the implementation of the data processing device can refer to the implementation of the data processing method, and the repetition will not be repeated. As used below, the term "unit" or "module" can be a combination of software and/or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented by software, implementation by hardware or a combination of software and hardware is also possible and conceived. Fig. 7 is a structural block diagram of a data processing device according to an embodiment of the present application. As shown in Fig. 7, it includes: a receiving module 701, a processing module 702, and a sending module 703. The structure is described below.

The receiving module 701 is configured to receive the ciphertext feature data sent by the server and the preset operator in the preset model, where the ciphertext feature data is encrypted by the server calling the convolution operator in the preset model according to the preset MPC protocol. The text data is obtained by convolution operation, and the cipher text data is obtained by encrypting the target data by the client using a homomorphic encryption algorithm;

The processing die hole 702 is used to call the preset operator in the preset model according to the preset MPC protocol to process the ciphertext feature data in cooperation with the client to obtain the ciphertext processing result;

The sending module 703 is used to send the ciphertext processing result to the client; the preset model is compatible with the preset MPC protocol, and is used to perform preset processing on the ciphertext data in the case of data encryption; the agent includes one of the following : A secure container running on the client, a virtual machine running on the client, and a computer that establishes a LAN connection with the client.

In some embodiments of the present application, after sending the ciphertext processing result to the client, it may further include: the client decrypts the ciphertext processing result to obtain the plaintext processing result.

In some embodiments of the present application, the processing die hole can be specifically used to: according to the preset MPC protocol and the activation operator in the preset operator, cooperate with the client to perform an activation operation on the ciphertext feature data to obtain the first activation Data and second activation data; according to the preset MPC protocol and the pooling operator in the preset operator, cooperate with the client to perform a pooling operation on the first activation data and the second activation data to obtain the first pooled data And the second pooled data; according to the preset MPC protocol and the fully connected operator in the preset operator, cooperate with the client to perform the fully connected operation on the first pooled data and the second pooled data to obtain ciphertext processing data.

From the above description, it can be seen that the embodiments of the present application achieve the following technical effects: by homomorphically encrypting the target data, the corresponding ciphertext data is obtained, and the specific content of the target data is prevented from being leaked. Further, the server calls the convolution operator in the preset model compatible with the MPC protocol to perform the convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data, and combine the ciphertext feature data with the preset model The preset operator of is sent to the agent, so that the agent and the client then call the preset operator in the preset model to perform specific processing on the ciphertext feature data according to the preset MPC protocol to obtain the ciphertext processing of the target data result. Realize that the target data can be specifically processed in the state of target data encryption, thereby effectively protecting data privacy. Furthermore, since the agent is a computer in the same local area network as the client or a secure container or virtual machine running on the client, it is more secure than directly having the server and the client perform ciphertext feature data. Calculation, through the agent and the client according to the preset MPC protocol to call the preset operator to perform secure multi-party calculation of the ciphertext feature data, which can effectively save the network bandwidth required for secure multi-party calculation, and greatly reduce the network delay. Effectively improve the efficiency of data processing.

The embodiment of the present application also provides a computer device. For details, please refer to the schematic diagram of the structure of the computer device based on the data processing method provided by the embodiment of the present application shown in FIG. 8. The computer device may specifically include an input device 81 and a processor. 82. Memory 83. Wherein, the memory 83 is used to store processor executable instructions. The processor 82 implements the steps of the data processing method described in any of the foregoing embodiments when executing the instructions.

In this embodiment, the input device may specifically be one of the main devices for information exchange between the user and the computer system. The input device may include a keyboard, a mouse, a camera, a scanner, a light pen, a handwriting input board, a voice input device, etc.; the input device is used to input raw data and programs for processing these numbers into the computer. The input device can also obtain and receive data transmitted from other modules, units, and devices. The processor can be implemented in any suitable way. For example, the processor may take the form of a microprocessor or a processor and a computer readable medium, logic gates, switches, application specific integrated circuits ( Application Specific Integrated Circuit, ASIC), programmable logic controller and embedded microcontroller form, etc. The memory may specifically be a memory device used to store information in modern information technology. The memory can include multiple levels. In a digital system, as long as it can store binary data, it can be a memory; in an integrated circuit, a circuit with a storage function without a physical form is also called a memory, such as RAM, FIFO, etc.; In the system, storage devices in physical form are also called memory, such as memory sticks, TF cards, and so on.

In this embodiment, the specific functions and effects implemented by the computer device can be explained in comparison with other embodiments, and will not be repeated here.

The embodiment of the present application also provides a computer storage medium based on a data processing method, the computer storage medium stores computer program instructions, and when the computer program instructions are executed, the data processing method described in any of the above embodiments is implemented A step of.

In this embodiment, the above-mentioned storage medium includes, but is not limited to, random access memory (Random Access Memory, RAM), read-only memory (Read-Only Memory, ROM), cache (Cache), and hard disk (Hard Disk Drive, HDD). Or memory card (Memory Card). The memory can be used to store computer program instructions. The network communication unit may be an interface set up in accordance with a standard stipulated by the communication protocol and used for network connection communication.

In this embodiment, the specific functions and effects realized by the program instructions stored in the computer storage medium can be explained in comparison with other embodiments, and will not be repeated here.

Obviously, those skilled in the art should understand that the modules or steps of the embodiments of the present application described above can be implemented by a general computing device, and they can be concentrated on a single computing device or distributed among multiple computing devices. Optionally, they can be implemented by the program code executable by the computing device, so that they can be stored in the storage device for execution by the computing device, and in some cases, they can be different from here The steps shown or described are performed in the order of, or they are respectively fabricated into individual integrated circuit modules, or multiple modules or steps of them are fabricated into a single integrated circuit module to achieve. In this way, the embodiments of the present application are not limited to any specific combination of hardware and software.

It should be understood that the above description is for illustration and not for limitation. By reading the above description, many implementations and many applications beyond the examples provided will be obvious to those skilled in the art. Therefore, the scope of this application should not be determined with reference to the above description, but should be determined with reference to the foregoing claims and the full scope of equivalents possessed by these claims.

The foregoing descriptions are only preferred embodiments of the present application, and are not intended to limit the present application. For those skilled in the art, the embodiments of the present application may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included in the protection scope of this application.

Claims

A data processing method, characterized in that it comprises:

Receiving a data processing request sent by a client, where the data processing request carries ciphertext data, and the ciphertext data is generated by the client after encrypting target data using a homomorphic encryption algorithm;

In response to the data processing request, according to a preset MPC protocol, call a convolution operator in a preset model to perform a convolution operation on the ciphertext data to obtain ciphertext feature data;

The ciphertext feature data and the preset operator in the preset model are sent to the agent, where the agent is used to call the preset with the client according to the preset MPC protocol The operator processes the ciphertext feature data to obtain a ciphertext processing result;

Wherein, the preset model is compatible with the preset MPC protocol, and is used to perform preset processing on the ciphertext data in the case of data encryption; the agent includes one of the following: running on the client A secure container, a virtual machine running on the client, and a computer that establishes a local area network connection with the client.
The method according to claim 1, wherein the agent and the client call the preset operator according to the preset MPC protocol to process the ciphertext feature data to obtain the ciphertext feature data. After the document processing results, it also includes:

Upload the transaction event corresponding to the data processing request to the blockchain network for on-chain storage.
The method according to claim 1, wherein the target data includes at least one of the following: face image data, medical image data, and human body motion data.
The method according to claim 3, wherein when the target data includes medical image data, the preset processing includes performing diagnostic processing on the medical image data.
The method according to claim 1, wherein, according to a preset MPC protocol, calling a convolution operator in a preset model to perform a convolution operation on the ciphertext data to obtain ciphertext feature data, comprising:

According to the homomorphic addition and multiplication characteristics in the preset MPC protocol, the convolution operator in the preset model is called to perform a convolution operation on the ciphertext data to obtain ciphertext feature data.
A data processing method, which is characterized in that it is applied to an agent, and includes:

Receive the ciphertext feature data sent by the server and the preset operator in the preset model, wherein the ciphertext feature data is invoked by the server according to the preset MPC protocol to the convolution operator pair in the preset model The ciphertext data is obtained by performing a convolution operation, and the ciphertext data is obtained after the client encrypts the target data using a homomorphic encryption algorithm;

Calling a preset operator in the preset model according to a preset MPC protocol to process the ciphertext feature data in cooperation with the client to obtain a ciphertext processing result;

Sending the ciphertext processing result to the client;

Wherein, the preset model is compatible with the preset MPC protocol, and is used to perform preset processing on the ciphertext data in the case of data encryption; the agent includes one of the following: running on the client A secure container, a virtual machine running on the client, and a computer that establishes a local area network connection with the client.
The method according to claim 6, wherein after sending the ciphertext processing result to the client, the method further comprises:

The client decrypts the ciphertext processing result to obtain the plaintext processing result.
The method according to claim 7, wherein after the client decrypts the ciphertext processing result to obtain the plaintext processing result, the method further comprises:

The client obtains a verification result, where the verification result is used to indicate whether the plaintext processing result is correct;

The client sends the verification result and the ciphertext processing result to the blockchain network for uploading storage.
The method according to claim 6, wherein the preset operator in the preset model is called according to the preset MPC protocol to cooperate with the client to process the ciphertext feature data to obtain The results of ciphertext processing, including:

According to the preset MPC protocol and the activation operator in the preset operator, cooperate with the client to perform an activation operation on the ciphertext feature data to obtain first activation data and second activation data;

According to the preset MPC protocol and the pooling operator in the preset operator, cooperate with the client to perform a pooling operation on the first activation data and the second activation data to obtain the first Pooled data and second pooled data;

According to the preset MPC protocol and the fully connected operator in the preset operator, cooperate with the client to perform a fully connected operation on the first pooled data and the second pooled data to obtain Encrypted processing data.
A data processing method, characterized in that it comprises:

The client obtains the target data, and encrypts the target data using a homomorphic encryption algorithm to obtain ciphertext data;

The client sends a data processing request to the server, wherein the data processing request carries the ciphertext data;

In response to the data processing request, the server and the client invoke a convolution operator in a preset model according to a preset MPC protocol to perform a convolution operation on the ciphertext data to obtain ciphertext feature data;

The server sends the ciphertext feature data and the preset operator in the preset model to the agent;

The agent terminal invokes the preset operator according to the preset MPC protocol to process the ciphertext feature data in cooperation with the client to obtain a ciphertext processing result;

Sending the ciphertext processing result to the client by the agent;

The client decrypts the ciphertext processing result to obtain the plaintext processing result of the target data;

Wherein, the preset model is compatible with the preset MPC protocol, and is used to perform preset processing on the ciphertext data in the case of data encryption; the agent includes one of the following: running on the client A secure container, a virtual machine running on the client, and a computer that establishes a local area network connection with the client.
A data processing device, characterized in that it comprises:

The receiving module is configured to receive a data processing request sent by a client, wherein the data processing request carries ciphertext data, and the ciphertext data is generated by the client after encrypting target data using a homomorphic encryption algorithm ；

The convolution module is configured to respond to the data processing request and call the convolution operator in the preset model to perform the convolution operation on the ciphertext data according to the preset MPC protocol to obtain the ciphertext feature data;

The sending module is configured to send the ciphertext feature data and the preset operators in the preset model to the agent, where the agent is used to communicate with the client according to the preset MPC protocol Calling the preset operator to process the ciphertext feature data to obtain a ciphertext processing result;

Wherein, the preset model is compatible with the preset MPC protocol, and is used to perform preset processing on the ciphertext data in the case of data encryption; the agent includes one of the following: running on the client A secure container, a virtual machine running on the client, and a computer that establishes a local area network connection with the client.
A computer device, characterized by comprising a processor and a memory for storing executable instructions of the processor, and the processor implements the steps of the method according to any one of claims 1 to 5 when the processor executes the instructions.
A computer-readable storage medium having computer instructions stored thereon, characterized in that, when the instructions are executed, the steps of the method according to any one of claims 1 to 5 are realized.