Method and device for determining computing architecture for cross-cloud environment data sharing

Abstract

The invention provides a computing architecture determination method and device for cross-cloud environment data sharing. The method comprises the following steps: establishing an initial calculation architecture, and performing gradient calculation on multiple pieces of cloud environment data on the initial calculation architecture to obtain gradient data corresponding to each piece of cloud environment data; any two gradient data are compared, unsafe data are determined according to the comparison result, the unsafe data are removed, safe data are obtained, the unsafe data represent unsafe data existing in the cloud environment data, and the safe data represent safe data existing in the cloud environment data; and performing reliability verification on the security data, and determining the computing architecture shared by the cross-cloud environment data as the initial computing architecture under the condition that the verification is passed. According to the method and the device, the security of the data in the calculation process of the cross-cloud environment data is improved.

Description

Technical Field

[0001] The present application relates to the field of cloud computing technology, and more specifically, to a method, device, computer-readable storage medium, and electronic device for determining a computing architecture for cross-cloud environment data sharing. Background Art

[0002] With the rapid development of cloud computing technology, the demand for data sharing across cloud environments is growing. This demand comes from many aspects: Data island problem: Different organizations or enterprises often have their own independent data storage and processing systems, forming data islands. This makes it difficult to share and use data, limiting the development of technologies such as big data analysis and artificial intelligence; Dispersed computing resources: Different cloud platforms have different computing resources, such as CPU (central processing unit, CPU for short), GPU (graphics processing unit, GPU for short) memory, etc. Cross-cloud environment data sharing can achieve the coordinated use of computing resources, improve resource utilization, and reduce costs; Business collaboration needs: Cross-industry and cross-field data sharing can promote business collaboration. For example, data sharing in the fields of medical care, finance, transportation, etc. can promote the development of innovative applications.

[0003] Therefore, how to collaboratively perform computing without leaking each other's data, thereby achieving the dual goals of data sharing and privacy protection, is a technical problem that urgently needs to be solved. Summary of the invention

[0004] The main purpose of the present application is to provide a method, device, computer-readable storage medium and electronic device for determining a computing architecture for sharing data across cloud environments, so as to at least solve the problem of data insecurity in the computing process of data across cloud environments in the prior art.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for determining a computing architecture for sharing data across cloud environments is provided, including: establishing an initial computing architecture, performing gradient calculation on multiple cloud environment data on the initial computing architecture, and obtaining gradient data corresponding to each of the cloud environment data; obtaining an ideal gradient, comparing the ideal gradient with the gradient data, determining unsafe data based on the comparison result, and eliminating the unsafe data to obtain safe data, wherein the ideal gradient is the gradient corresponding to the case where all the cloud environment data are safe, the unsafe data represents the unsafe data existing in the cloud environment data, and the safe data represents the safe data existing in the cloud environment data; performing reliability verification on the safe data, and if the verification passes, determining that the computing architecture for sharing data across cloud environments is the initial computing architecture.

[0006] Optionally, establishing an initial computing architecture includes: obtaining user identity authentication information, verifying the user identity authentication information, and sending a login permission identification to the user if the verification is successful, so as to establish an access control policy for the initial computing architecture; in the process of data transmission between users, encrypting the target transmission data and calculating the encrypted data, so as to establish an encryption communication strategy for the initial computing architecture; allocating computing resources of the cloud environment through an intelligent scheduler, so as to establish an intelligent scheduling strategy for the initial computing architecture.

[0007] Optionally, performing gradient calculation on multiple cloud environment data on the initial computing architecture includes: using the formula A gradient calculation is performed on the cloud environment data, wherein x is an independent variable, y is a dependent variable, f(x, y) represents a functional relationship between x and y, and ▽(f(x, y)) represents a partial derivative of the function f(x, y).

[0008] Optionally, comparing the ideal gradient with the gradient data, and determining unsafe data according to the comparison result, comprises: using the formula The ideal gradient is compared with the gradient data, where x a represents the ideal gradient, x b represents the gradient data, sgn(·) represents the sign extraction function, and I(·) represents the judgment function.

[0009] Optionally, the reliability of the security data is verified, including: The reliability of the security data is verified, wherein: is the reliability index, x a represents the ideal gradient, x b represents the gradient data, l is the number of dimensions, M is the number of users, and dis(·) is the distance function.

[0010] Optionally, the method further comprises: Determine the distance function, where x a =(x1,y1),x b =(x2,y2).

[0011] Optionally, when the verification is passed, the method further includes: using the reliability index of the reliability verification as a weight, updating the gradient data, so as to update the initial calculation architecture.

[0012] According to another aspect of the present application, a device for determining a computing architecture for sharing data across cloud environments is provided, comprising: a computing unit, used to establish an initial computing architecture, and perform gradient calculation on cloud environment data on the initial computing architecture to obtain gradient data; a first determination unit, used to obtain an ideal gradient, compare the ideal gradient with the gradient data, determine unsafe data, and eliminate the unsafe data to obtain safe data, wherein the ideal gradient is the gradient corresponding to the case where all cloud environment data are safe, the unsafe data represents unsafe data existing in the cloud environment data, and the safe data represents safe data existing in the cloud environment data; a second determination unit, used to perform reliability verification on the safe data, and if the verification passes, determine that the computing architecture for sharing data across cloud environments is the initial computing architecture.

[0013] According to another aspect of the present application, a computer-readable storage medium is provided, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the computing architecture determination methods for cross-cloud environment data sharing.

[0014] According to another aspect of the present application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a computing architecture determination method for executing any one of the cross-cloud environment data sharing methods.

[0015] Apply the technical solution of this application, establish an initial computing architecture, perform gradient calculation on multiple cloud environment data on the initial computing architecture, and obtain gradient data corresponding to each cloud environment data; obtain an ideal gradient, compare the ideal gradient with the gradient data, determine unsafe data based on the comparison result, and eliminate unsafe data to obtain safe data, wherein unsafe data represents unsafe data existing in cloud environment data, and safe data represents safe data existing in cloud environment data; perform reliability verification on safe data, and if the verification passes, determine the computing architecture for sharing data across cloud environments as the initial computing architecture. Compared with the prior art, which cannot both guarantee the security of data on each platform and coordinate data resources on each platform for big data computing, this application evaluates and verifies the security data of the initial computing architecture, and further determines the computing architecture for sharing data across cloud environments. Therefore, the security of data in the calculation process of cross-cloud environment data can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings constituting part of the present application are used to provide a further understanding of the present application. The exemplary embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 A hardware structure block diagram of a mobile terminal for performing a computing architecture determination method for cross-cloud environment data sharing provided by an embodiment of the present application is shown;

[0018] Figure 2 A schematic diagram of a process for determining a computing architecture for cross-cloud environment data sharing provided by an embodiment of the present application is shown;

[0019] Figure 3 A schematic diagram of data arrangement for a relatively high ratio of unreliable users provided by an embodiment of the present application is shown;

[0020] Figure 4 A schematic diagram of data arrangement of a medium-level unreliable user ratio provided by an embodiment of the present application is shown;

[0021] Figure 5 A schematic diagram of data arrangement of a high level of unreliable user ratio provided by an embodiment of the present application is shown;

[0022] Figure 6 A schematic diagram showing a test result of a system running time provided by an embodiment of the present application is shown;

[0023] Figure 7 A schematic diagram of a communication overhead test result provided by an embodiment of the present application is shown;

[0024] Figure 8 A schematic diagram of a specific method for determining a computing architecture for cross-cloud environment data sharing provided by an embodiment of the present application is shown;

[0025] Fig. 9 A schematic diagram of a computing architecture for cross-cloud environment data sharing provided by an embodiment of the present application is shown;

[0026] Fig.10 A structural block diagram of a computing architecture determination device for cross-cloud environment data sharing provided by an embodiment of the present application is shown.

[0027] The above drawings include the following reference numerals:

[0028] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0029] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0030] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0031] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0032] As introduced in the background technology, the existing technology is unable to ensure the security of data on each platform and coordinate the data resources of each platform to perform big data computing. In order to solve the problem of data insecurity when the data resources of each platform perform big data computing, the embodiments of the present application provide a method, device, computer-readable storage medium and electronic device for determining a computing architecture for sharing data across cloud environments.

[0033] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0034] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 1 is a hardware structure block diagram of a mobile terminal of a method for determining a computing architecture for cross-cloud environment data sharing according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0035] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the computing architecture determination method for cross-cloud environment data sharing in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The above-mentioned specific network examples may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0036] In this embodiment, a method for determining a computing architecture for cross-cloud environment data sharing running on a mobile terminal, a computer terminal, or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0037] Figure 2 1 is a flow chart of a method for determining a computing architecture for cross-cloud environment data sharing according to an embodiment of the present application. Figure 2As shown, the method comprises the following steps:

[0038] Step S201, establishing an initial computing architecture, performing gradient calculation on multiple cloud environment data on the initial computing architecture, and obtaining gradient data corresponding to each of the cloud environment data;

[0039] Specifically, the computing architecture for cross-cloud data sharing aims to solve data privacy protection and security issues, and is particularly suitable for scenarios where sensitive data needs to be shared in multiple cloud environments. The architecture is based on secure multi-party computing technology and combines key functions such as encrypted communication, access control, and intelligent scheduler. In terms of access control, the architecture adopts strict authentication and authorization policies to ensure that only authorized users and services can access and process specific data. Participants establish secure connections by using protocols such as TLS / SSL, and encrypt communications during transmission to prevent eavesdropping and tampering. For distributed structures, the traceability and immutability of the process are ensured, and the gradient descent algorithm is used locally to calculate the loss function and optimize it. The architecture performs end-to-end encryption and distributed computing of data to ensure that data remains encrypted during transmission and storage. The architecture uses an intelligent scheduler to optimize the utilization and dynamic allocation of computing resources in a multi-cloud environment, ensuring the efficient execution of the secure multi-party computing process.

[0040] Step S202, obtaining an ideal gradient, comparing the ideal gradient with the gradient data, determining unsafe data according to the comparison result, and eliminating the unsafe data to obtain safe data, wherein the ideal gradient is a gradient corresponding to the case where the cloud environment data are all safe, the unsafe data represents unsafe data existing in the cloud environment data, and the safe data represents safe data existing in the cloud environment data;

[0041] Specifically, after the gradient data is calculated, the unsafe data is removed. The unsafe data can be unsafe user information. In order to cope with the distributed training method involving multiple users, the ideal gradient and the local gradient submitted by a specific user can be verified by comparing the vector dimensions, thereby ensuring that irrelevant components can be accurately removed before the global ideal gradient is aggregated. Specifically, it uses the dimension of the gradient of the previous round rather than the current round as the comparison object. Although this cannot adapt to the optimization results of the previous iteration that change rapidly, it can avoid the current comparison waiting for all uploads to complete as much as possible on the basis of a small loss of convergence, which can significantly reduce the algorithm time consumption.

[0042] Step S203, performing reliability verification on the security data, and if the verification passes, determining the computing architecture for sharing the cross-cloud environment data as the initial computing architecture.

[0043] Specifically, in order to obtain the components of each dimension in the ideal gradient, it is necessary to first consider the reliability of the components of the local dimension in each dimension. Common reliability verification methods include: consistency index: compare the signs and values ​​of the local gradients submitted by different users in each dimension. If there are large differences, it means that these gradients may come from unreliable users; convergence speed index: evaluate the rate of decrease of the local gradient to the model loss function. If the convergence speed is too slow, it means that these gradients may come from unreliable users; distance index: calculate the distance between the local gradients submitted by different users. If the distance is too large, it means that these gradients may come from unreliable users; establish a game theory model to analyze the behavioral motivations of users in the distributed training process, so as to evaluate the reliability of the local gradients submitted by users; this application will construct specific reliability indicators below.

[0044] Through this embodiment, an initial computing architecture is established, and gradient calculations are performed on multiple cloud environment data on the initial computing architecture to obtain gradient data corresponding to each cloud environment data; any two gradient data are compared, unsafe data is determined based on the comparison result, and unsafe data is eliminated to obtain safe data, wherein unsafe data represents unsafe data existing in cloud environment data, and safe data represents safe data existing in cloud environment data; reliability verification is performed on safe data, and when the verification passes, the computing architecture for sharing data across cloud environments is determined to be the initial computing architecture. Compared with the prior art, which is unable to both guarantee the security of data on each platform and coordinate data resources on each platform for big data computing, the present application evaluates and verifies the security data of the initial computing architecture, and further determines the computing architecture for sharing data across cloud environments. Therefore, the security of data during the calculation process of data across cloud environments can be improved.

[0045] In the specific implementation process, the above step S201 establishes the initial computing architecture, which can be implemented through the following steps: Step S2011: obtain user identity authentication information, verify the user identity authentication information, and send a login permission identification to the user if the verification is passed, so as to establish the access control policy of the initial computing architecture; Step S2012: in the process of data transmission between users, encrypt the target transmission data, and calculate the encrypted data to establish the encryption communication strategy of the initial computing architecture; Step S2013: allocate the computing resources of the cloud environment through the intelligent scheduler to establish the intelligent scheduling strategy of the initial computing architecture. The method establishes the initial computing structure through the above steps, and performs data calculation and data security strategy layout on the initial computing architecture to ensure data security.

[0046] Specifically, in terms of access control, the architecture adopts strict authentication and authorization policies to ensure that only authorized users and services can access and process specific data. Participants need to register and obtain a unique identifier. During the registration process, they will provide user authentication information and generate a login ID for encrypted communication and authentication, which can be a public-private key pair. In order to prevent identity information leakage, encryption algorithms can be used to encrypt and store user information, and keys can be changed regularly. Each participant needs to perform preprocessing operations before uploading data to reduce the risk of sensitive information leakage, which includes de-identification, noise addition or other privacy protection technologies. Participants establish a secure connection by using protocols such as TLS (Transport Layer Security, referred to as TLS protocol) / SSL (Secure Sockets Layer, referred to as SSL protocol), and perform encrypted communication during the transmission of the target transmission data to prevent eavesdropping and tampering. For the distributed structure, the traceability and immutability of the process are ensured, and the gradient descent algorithm is used locally to calculate the loss function and optimize it. The architecture performs end-to-end encryption and distributed computing of data to ensure that data is always encrypted during transmission and storage. At the same time, a secure multi-party computing framework is built, such as secure multi-party computing protocols, obfuscated circuits, zero-knowledge proofs, etc., to ensure that participants do not leak their own data during collaborative computing. This design effectively prevents the risk of data leakage and tampering in the cloud environment, and complies with the data protection regulations of many countries and regions. At the same time, homomorphic encryption technology can be used to allow calculations on encrypted data without decrypting the data, thereby further improving data security. In addition, the intelligent scheduler is used to optimize the utilization and dynamic allocation of computing resources in a multi-cloud environment to ensure the efficient execution of the secure multi-party computing process. The intelligent scheduler can perform intelligent scheduling according to the needs of data processing tasks and the performance characteristics of different cloud service providers, thereby improving computing efficiency and resource utilization.

[0047] In some optional implementations, the above step S201 performs gradient calculation on multiple cloud environment data on the initial computing architecture, which can be implemented by the following steps: Step S2014: By formula The cloud environment data is subjected to gradient calculation, wherein x is an independent variable, y is a dependent variable, f(x, y) represents a functional relationship between x and y, and ▽(f(x, y)) represents a partial derivative of the function f(x, y). The method performs gradient calculation through the above steps, so that the gradient data can be accurately calculated.

[0048] Specifically, assume that the ideal objective function is f(x) = y, there are M participants who perform local training and upload their local gradients to the cloud server for global ideal gradient update, each user holds independent and identically distributed data, and the gradient is calculated by taking partial derivatives of multivariate functions and writing these partial derivatives in the form of vectors ▽(f(x,y)). In the function f(x,y), the equation is defined as: For the function f(x,y) at the point (x0,y0), in the opposite direction of the gradient vector It is the place where the gradient descends fastest. It is easier to find the minimum value of the loss function along the opposite direction of the gradient vector. Based on this, the design of the computing architecture is discussed accordingly. Its main purpose is to seek the reliability of this algorithm under unsafe conditions for some users, so as to ensure the value of the design architecture in the actual use environment.

[0049] In some optional implementations, the above step S202 compares the ideal gradient with the gradient data, and determines the unsafe data according to the comparison result, which can be achieved by the following steps: The ideal gradient is compared with the gradient data, where x a represents the ideal gradient, x b represents the gradient data, sgn(·) represents the symbol extraction function, and I(·) represents the judgment function. The method compares the data through the above steps and determines the comparison result, so that the gradient data can be screened and unsafe data can be eliminated.

[0050] Specifically, in the actual distributed training process, the negative impact of unsafe users on the global ideal gradient needs to be considered. Unsafe users can be divided into malicious users and unreliable users. Malicious users seek to attack the system as a whole by injecting malicious information. Unreliable users are easily confused with them, which refer to users who use low-quality data to complete local training. Malicious users inject malicious information, such as tampering with data and adding noise. In order to identify and eliminate malicious users, the following methods are commonly used: establish a user reputation mechanism, evaluate users based on their historical behavior and contributions, and mark users with low reputation as potential malicious users. Use machine learning algorithms, such as anomaly detection algorithms, to identify abnormal patterns in user behavior, thereby identifying potential malicious users and performing integrity checks on data submitted by users. Unreliable users may submit local gradients that are far away from the model convergence direction due to data quality issues, insufficient computing power, etc., which has a negative impact on the global ideal gradient. In order to identify and eliminate unreliable users, the following methods are commonly used: use reliability indicators, such as gradient consistency and convergence speed, to evaluate the reliability of local gradients submitted by users. The local gradients submitted by users are weighted and aggregated, and higher weights are given to gradients with high reliability, thereby reducing the impact of unreliable gradients on the global ideal gradient. Federated learning technology can be used to train the model locally, and then the training results are uploaded to the cloud server for aggregation, thereby reducing dependence on unreliable users. The local gradients they generate may be far away from the convergence direction of the model, which is manifested as inconsistent signs and numerical differences. Therefore, such information needs to be eliminated. In this regard, the algorithm does not specifically distinguish whether the user has an attack intention, and all of them are subject to data screening. To address this problem, its data processing consists of two parts, namely the irrelevant component exclusion algorithm and the weighted aggregation algorithm. As mentioned earlier, the sign and value of the gradient in a specific dimension determine the direction of the gradient itself, which also forms the main difference between irrelevant components. In order to cope with the distributed training method involving multiple users, the comparison between the ideal gradient and the local gradient submitted by a specific user can be verified by comparing the vector dimensions. The formula is as follows I(x a ,x b ), so as to ensure that irrelevant components can be accurately eliminated before the global ideal gradient is aggregated. Specifically, it uses the dimension of the gradient of the previous round rather than the current round as the comparison object. Although this cannot adapt to the optimization results of the rapid changes in the previous iterations, it can avoid the current comparison waiting for all uploads to complete as much as possible on the basis of a small loss of convergence, which can significantly reduce the algorithm time consumption.

[0051] In some optional implementations, the above step S203 verifies the reliability of the security data, which can be achieved by the following steps: The reliability of the security data is verified, wherein: is the reliability index, x a represents the ideal gradient, x b represents the gradient data, l is the number of dimensions, M is the number of users, and dis(·) is the distance function. The method performs reliability verification through the above steps, so that the reliability of the data can be determined.

[0052] Specifically, in order to obtain the components of each dimension in the ideal gradient, it is necessary to first consider the reliability of the components of the local dimension in each dimension. Common reliability verification methods: Consistency index: compare the signs and values ​​of the local gradients submitted by different users in each dimension. If there are large differences, it means that these gradients may come from unreliable users; Convergence speed index: evaluate the rate of decrease of the local gradient to the model loss function. If the convergence speed is too slow, it means that these gradients may come from unreliable users; Distance index: calculate the distance between the local gradients submitted by different users. If the distance is too large, it means that these gradients may come from unreliable users; establish a game theory model to analyze the behavioral motivations of users in the distributed training process, so as to evaluate the reliability of the local gradients submitted by users; Therefore, the corresponding reliability indicators constructed in this application are

[0053] In some optional implementations, the method further includes the following steps: Step S204: By formula Determine the distance function, where x a =(x1,y1),x b =(x2,y2). This method determines the distance function through the above-mentioned Euclidean distance, so that the distance can be calculated more accurately.

[0054] In the specific implementation process, the distance function uses the Euclidean distance. The calculation formula of the Euclidean distance between two points under the standard two-dimensional space condition is shown in the above formula, which can be further extended to high-dimensional space, then: Therefore, this reliability analysis result allows the ideal gradient to be updated weightedly using this reliability as a weight.

[0055] In some optional implementations, when the verification is passed, the method further includes step S205: using the reliability index of the reliability verification as a weight to update the gradient data to update the initial computing architecture. The method performs gradient update through the above steps, thereby updating the above initial computing architecture to obtain an optimized computing architecture.

[0056] In the specific implementation process, the gradient data submitted by the user is weighted and aggregated, and the gradients with high reliability are given higher weights, thereby reducing the impact of unreliable gradients on the global ideal gradient. Federated learning technology can be used to train the model locally, and then upload the training results to the cloud server for aggregation, thereby reducing dependence on unreliable users. The structure of this application can effectively screen multi-source data sharing across cloud environments, and finally perform necessary post-processing on the output results, such as denoising and inverse transformation, to restore the required business-related information, thereby realizing distributed computing, and being able to implement multi-party computing architecture design under a more secure organizational framework.

[0057] In the actual application process, the performance testing steps of the above architecture are also included.

[0058] First, the test environment is Java 1.7.0 to build an experimental environment, and the Paillier encryption function library is used for data encryption. Based on this, a data sharing multi-party computing model in the cloud environment is built for performance testing. The cloud server settings are shown in Table 1.

[0059] To simulate multi-user participation, 8 cores and 8GB RAM are uniformly used to build user participation of different scales, and the test data uses the public dataset MNIST.

[0060] Then, an accuracy test is conducted. The reason for choosing to evaluate the unreliable user identification performance is that in practical applications, facing large-scale data streams and heterogeneous environments, it is crucial to ensure high-precision unreliable user identification.

[0061] Table 1 Cloud server equipment environment

[0062] equipment Configuration operating system Ubuntu 18.04 RAM 16 GB SSD 256GB CPU 2.10GHz

[0063] This application tests three unreliable user ratios of 30%, 20% and 10%, and conducts experiments under low, medium and high concurrency conditions. Specifically, the three cases of 100 total users (low concurrency), 300 total users (medium concurrency) and 500 total users (high concurrency) are considered to comprehensively evaluate the performance of the proposed secure multi-party computing architecture design under different scales and ratios. For a higher unreliable user ratio, the data is organized as follows Figure 3As shown. When faced with a higher proportion of unreliable users, it can be noticed that the recognition accuracy is relatively low. This may be due to the challenge faced by the system in dealing with a large number of potentially malicious behaviors, resulting in a certain degree of misjudgment. In this case, the model needs more iterations to learn from the noise and improve its prediction ability. Nevertheless, as the number of iterations increases and more data is taken into account, after more than 10 iterations, the accuracy at different concurrency scales has exceeded 20%, and after 20 iterations, it generally exceeds 95%, reflecting a gradual improvement in accuracy. For a medium level of unreliable user ratio, the data is organized as follows Figure 4 As shown. As the proportion of unreliable users drops to 20%, we see an improvement in overall recognition accuracy. This shows that the system begins to distinguish between real and abnormal behaviors more effectively, and the model can converge to a high-precision state more quickly. When the number of iterations exceeds 5, the accuracy under high concurrency conditions has quickly risen to nearly 80%. At this time, the accuracy of data at different concurrency scales also has significant differences, that is, higher concurrency provides a larger scale of training samples for model optimization. After 10 iterations, the medium and high level concurrency data can already achieve a high accuracy of 90% or above, and can also reach more than 90% under low concurrency conditions. After 15 iterations, the difference in first-mover advantages of different concurrency scales has narrowed, and the overall accuracy has approached 100%, reflecting the high accuracy of the overall performance of the model. Therefore, the comprehensive results show that in this case the model has a certain degree of robustness, and can achieve good results with a small amount of data. For a high level of unreliable user ratio, the data is organized as follows Figure 5 As shown. When there are only 10% untrusted users, the system shows the best recognition accuracy. At this time, the model obviously benefits from fewer interference factors and can make correct predictions quickly and stably. At this time, the difference between different concurrency scales is further expanded. After 10 iterations, the accuracy under low concurrency conditions is slightly close to 100%, and when the iterations exceed 20 times, the data under all concurrency conditions has stabilized at 100% accuracy level. Therefore, the experimental results strongly suggest that the secure multi-party computing architecture design has a strong adaptability in dealing with small-scale potential malicious participants.

[0064] Finally, we conducted a complexity analysis and conducted corresponding tests on the changes in system running time and communication overhead under different gradient levels and concurrent user scales to verify the complexity of the algorithm operation itself. Figure 6As shown. As the gradient increases, the system running time gradually increases. This is because more gradients need to be calculated and updated, resulting in more calculations. A slower rise rate was observed at 5000 gradients. This may be because the system has reached a certain saturation state at this time, and the impact of each additional gradient added after that is relatively weakened. As the number of concurrent users increases, the system running time is also extended. When processing more simultaneous requests, the server needs to allocate more resources to process each request. For 5000 concurrent users, under this large-scale load condition, the system may encounter bottlenecks or performance limitations, resulting in a smaller degree of progress rate. The test results of communication overhead are shown in Figure 7 As shown. At different gradient levels, the communication overhead shows a clear linear growth trend. This is because more parameters need to be transmitted and updated, resulting in an increase in the amount of data transmitted. Although a slower rate of increase was observed in the case of 5000 model parameters, there is still a clear linear relationship. This shows that there will be corresponding communication costs even at a larger scale. As more users participate in the operation at the same time, the bandwidth and data transmission capacity requirements will increase accordingly. Each user sends a request to the server and receives a response, which consumes a certain amount of network resources. A higher number of concurrent users is often accompanied by higher communication costs. In actual system design, it is necessary to consider how to effectively manage and optimize these communication overheads to maintain the overall efficiency of the system.

[0065] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the method for determining the computing architecture for cross-cloud environment data sharing of the present application will be described in detail below in combination with specific embodiments.

[0066] This embodiment relates to a specific method for determining a computing architecture for cross-cloud environment data sharing, such as Figure 8 As shown, the following steps are included:

[0067] Step S1: Sharing user information: The local gradient (gradient data) is uploaded to the server for updating the global ideal gradient. Participants (users) can upload the local gradient to the cloud server to update the global ideal gradient.

[0068] Step S2: Elimination of unsafe user information: Malicious or low-quality data may have a negative impact on the global ideal gradient. How to identify and exclude malicious or low-quality data to avoid affecting the global ideal gradient;

[0069] Step S3: Reliability verification: Use the distance function as the weight to perform weighted update on multi-source data to achieve multi-source data sharing and ensure model accuracy;

[0070] Step S4: Distributed computing implementation: The algorithm implements distributed computing, which can perform calculations and information sharing among multiple participants, and realizes multi-party computing architecture design under a more secure organizational framework, thereby improving computing efficiency and model accuracy.

[0071] This embodiment also relates to a computing architecture for sharing data across cloud environments, such as Fig. 9 As shown, the secure multi-party computing architecture has multiple functional branches: a distributed structure uses a gradient descent algorithm locally to calculate the loss function and optimize it; TSL / SSL and other protocols establish a secure connection, encrypt communications during transmission to prevent eavesdropping and tampering, and encrypt communications during transmission; a highly scalable and configurable secure multi-party computing solution that effectively supports data cooperation projects of various sizes and complexities, and requires identity authentication when logging in; an intelligent scheduler that optimizes the utilization and dynamic allocation of computing resources in a multi-cloud environment; supports secure data sharing across cloud environments, securely transmits and processes data between different cloud service providers, and secure data sharing can be performed between different data cooperation projects; the secure multi-party computing architecture implements end-to-end data encryption and distributed computing.

[0072] The embodiment of the present application also provides a computing architecture determination device for cross-cloud environment data sharing. It should be noted that the computing architecture determination device for cross-cloud environment data sharing in the embodiment of the present application can be used to execute the computing architecture determination method for cross-cloud environment data sharing provided in the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and those that have been described will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0073] The following introduces a computing architecture determination device for cross-cloud environment data sharing provided in an embodiment of the present application.

[0074] Fig.10 It is a schematic diagram of a computing architecture determination device for cross-cloud environment data sharing according to an embodiment of the present application.

[0075] like Fig.10 As shown, the device comprises:

[0076] A computing unit 11 is used to establish an initial computing architecture, and perform gradient computing on the cloud environment data on the initial computing architecture to obtain gradient data;

[0077] Specifically, the computing architecture for cross-cloud data sharing aims to solve data privacy protection and security issues, and is particularly suitable for scenarios where sensitive data needs to be shared in multiple cloud environments. The architecture is based on secure multi-party computing technology and combines key functions such as encrypted communication, access control, and intelligent scheduler. In terms of access control, the architecture adopts strict authentication and authorization policies to ensure that only authorized users and services can access and process specific data. Participants establish secure connections by using protocols such as TLS / SSL, and encrypt communications during transmission to prevent eavesdropping and tampering. For distributed structures, the traceability and immutability of the process are ensured, and the gradient descent algorithm is used locally to calculate the loss function and optimize it. The architecture performs end-to-end encryption and distributed computing of data to ensure that data remains encrypted during transmission and storage. The architecture uses an intelligent scheduler to optimize the utilization and dynamic allocation of computing resources in a multi-cloud environment, ensuring the efficient execution of the secure multi-party computing process.

[0078] A first determining unit 12 is used to obtain an ideal gradient, compare the ideal gradient with the gradient data, determine unsafe data, and remove the unsafe data to obtain safe data, wherein the ideal gradient is a gradient corresponding to a case where all the cloud environment data are safe, the unsafe data represents unsafe data existing in the cloud environment data, and the safe data represents safe data existing in the cloud environment data;

[0079] Specifically, after the gradient data is calculated, the unsafe data is eliminated. The unsafe data can be unsafe user information. In order to cope with the distributed training device involving multiple users, the ideal gradient and the local gradient submitted by a specific user can be compared and verified by vector dimension comparison, thereby ensuring that irrelevant components can be accurately eliminated before the aggregation of the global ideal gradient. Specifically, it uses the dimension of the gradient of the previous round rather than the current round as the comparison object. Although this cannot adapt to the optimization results of the rapid changes in the previous iteration, it can avoid the current comparison as much as possible while waiting for all uploads to complete on the basis of a small loss of convergence, which can significantly reduce the algorithm time consumption.

[0080] The second determining unit 13 is used to verify the reliability of the security data, and if the verification is passed, determine the computing architecture for sharing the cross-cloud environment data as the initial computing architecture.

[0081] Specifically, in order to obtain the components of each dimension in the ideal gradient, it is necessary to first consider the reliability of the components of the local dimension in each dimension. Common reliability verification devices include: consistency index: compare the signs and values ​​of the local gradients submitted by different users in each dimension. If there are large differences, it means that these gradients may come from unreliable users; convergence speed index: evaluate the rate of decrease of the local gradient to the model loss function. If the convergence speed is too slow, it means that these gradients may come from unreliable users; distance index: calculate the distance between the local gradients submitted by different users. If the distance is too large, it means that these gradients may come from unreliable users; establish a game theory model to analyze the behavioral motivations of users in the distributed training process, so as to evaluate the reliability of the local gradients submitted by users; this application will construct specific reliability indicators below.

[0082] Through this embodiment, an initial computing architecture is established, and gradient calculations are performed on multiple cloud environment data on the initial computing architecture to obtain gradient data corresponding to each cloud environment data; any two gradient data are compared, unsafe data is determined based on the comparison result, and unsafe data is eliminated to obtain safe data, wherein unsafe data represents unsafe data existing in cloud environment data, and safe data represents safe data existing in cloud environment data; reliability verification is performed on safe data, and when the verification passes, the computing architecture for sharing data across cloud environments is determined to be the initial computing architecture. Compared with the prior art, which is unable to both guarantee the security of data on each platform and coordinate data resources on each platform for big data computing, the present application evaluates and verifies the security data of the initial computing architecture, and further determines the computing architecture for sharing data across cloud environments. Therefore, the security of data during the calculation process of data across cloud environments can be improved.

[0083] In the specific implementation process, the above-mentioned computing unit includes a sending module, a first computing module and an allocation module. The sending module is used to obtain user identity authentication information, verify the user identity authentication information, and send a login permission identification to the user if the verification is passed, so as to establish the access control policy of the initial computing architecture; the first computing module is used to encrypt the target transmission data during the process of data transmission between users, and calculate the encrypted data to establish the encryption communication policy of the initial computing architecture; the allocation module is used to allocate the computing resources of the cloud environment through the intelligent scheduler to establish the intelligent scheduling policy of the initial computing architecture. The device establishes the initial computing structure through the above steps, and performs data calculation and data security policy layout on the initial computing architecture to ensure data security.

[0084] Specifically, in terms of access control, the architecture adopts strict authentication and authorization policies to ensure that only authorized users and services can access and process specific data. Participants need to register and obtain a unique identifier. During the registration process, they will provide user authentication information and generate a login ID for encrypted communication and authentication, which can be a public-private key pair. In order to prevent identity information leakage, encryption algorithms can be used to encrypt and store user information, and keys can be changed regularly. Each participant needs to perform preprocessing operations before uploading data to reduce the risk of sensitive information leakage, which includes de-identification, noise addition or other privacy protection technologies. Participants establish a secure connection by using protocols such as TLS (Transport Layer Security, referred to as TLS protocol) / SSL (Secure Sockets Layer, referred to as SSL protocol), and perform encrypted communication during the transmission of the target transmission data to prevent eavesdropping and tampering. For the distributed structure, the traceability and immutability of the process are ensured, and the gradient descent algorithm is used locally to calculate the loss function and optimize it. The architecture performs end-to-end encryption and distributed computing of data to ensure that data is always encrypted during transmission and storage. At the same time, a secure multi-party computing framework is built, such as secure multi-party computing protocols, obfuscated circuits, zero-knowledge proofs, etc., to ensure that participants do not leak their own data during collaborative computing. This design effectively prevents the risk of data leakage and tampering in the cloud environment, and complies with the data protection regulations of many countries and regions. At the same time, homomorphic encryption technology can be used to allow calculations on encrypted data without decrypting the data, thereby further improving data security. In addition, the intelligent scheduler is used to optimize the utilization and dynamic allocation of computing resources in a multi-cloud environment to ensure the efficient execution of the secure multi-party computing process. The intelligent scheduler can perform intelligent scheduling according to the needs of data processing tasks and the performance characteristics of different cloud service providers, thereby improving computing efficiency and resource utilization.

[0085] In some optional embodiments, the calculation unit further includes a second calculation module for calculating The cloud environment data is subjected to gradient calculation, wherein x is an independent variable, y is a dependent variable, f(x, y) represents a functional relationship between x and y, and ▽(f(x, y)) represents a partial derivative of the function f(x, y). The device performs gradient calculation through the above steps, so that the gradient data can be accurately calculated.

[0086] Specifically, assume that the ideal objective function is f(x) = y, there are M participants who perform local training and upload their local gradients to the cloud server for global ideal gradient update, each user holds independent and identically distributed data, and the gradient calculation device is to find partial derivatives of multivariate functions and write these partial derivatives in the form of vectors ▽(f(x,y)). In the function f(x,y), the equation is defined as: For the function f(x,y) at the point (x0,y0), in the opposite direction of the gradient vector It is the place where the gradient descends fastest. It is easier to find the minimum value of the loss function along the opposite direction of the gradient vector. Based on this, the design of the computing architecture is discussed accordingly. Its main purpose is to seek the reliability of this algorithm under unsafe conditions for some users, so as to ensure the value of the design architecture in the actual use environment.

[0087] In some optional implementations, the first determining unit includes a comparison module for determining The ideal gradient is compared with the gradient data, where x a represents the ideal gradient, x b represents the gradient data, sgn(·) represents the symbol extraction function, and I(·) represents the judgment function. The device compares the data through the above steps and determines the comparison result, so that the gradient data can be screened and unsafe data can be eliminated.

[0088] Specifically, in the actual distributed training process, it is necessary to consider the negative impact of unsafe users on the global ideal gradient. Unsafe users can be divided into malicious users and unreliable users. Malicious users seek to attack the system as a whole by injecting malicious information. Unreliable users are easily confused with them, which refer to those users who use low-quality data to complete local training. Malicious users inject malicious information, such as tampering with data, adding noise, etc. In order to identify and eliminate malicious users, the following devices are commonly used: Establish a user reputation mechanism, evaluate users based on their historical behavior and contributions, and mark users with low reputation as potential malicious users. Use machine learning algorithms, such as anomaly detection algorithms, to identify abnormal patterns in user behavior, thereby identifying potential malicious users and performing integrity checks on data submitted by users. Unreliable users may cause the local gradients they submit to be far away from the model convergence direction due to data quality problems, insufficient computing power, etc., which has a negative impact on the global ideal gradient. In order to identify and eliminate unreliable users, the following devices are commonly used: Use reliability indicators, such as gradient consistency, convergence speed, etc., to evaluate the reliability of local gradients submitted by users. The local gradients submitted by users are weighted and aggregated, and higher weights are given to gradients with high reliability, thereby reducing the impact of unreliable gradients on the global ideal gradient. Federated learning technology can be used to train the model locally, and then the training results are uploaded to the cloud server for aggregation, thereby reducing dependence on unreliable users. The local gradients they generate may be far away from the convergence direction of the model, which is manifested as inconsistent signs and numerical differences. Therefore, such information needs to be eliminated. In this regard, the algorithm does not specifically distinguish whether the user has an attack intention, and all of them are subject to data screening and processing. To address this problem, its data processing consists of two parts, namely the irrelevant component exclusion algorithm and the weighted aggregation algorithm. As mentioned earlier, the sign and value of the gradient in a specific dimension determine the direction of the gradient itself, which also forms the main difference between irrelevant components. In order to cope with distributed training devices involving multiple users, the comparison between the ideal gradient and the local gradient submitted by a specific user can be verified by vector dimension comparison. The formula is as follows: I(x a ,x b ), so as to ensure that irrelevant components can be accurately eliminated before the global ideal gradient is aggregated. Specifically, it uses the dimension of the gradient of the previous round rather than the current round as the comparison object. Although this cannot adapt to the optimization results of the rapid changes in the previous iterations, it can avoid the current comparison waiting for all uploads to complete as much as possible on the basis of a small loss of convergence, which can significantly reduce the algorithm time consumption.

[0089] In some optional implementations, the second determination unit includes a verification module for determining The reliability of the security data is verified, wherein: is the reliability index, xa represents the ideal gradient, x b represents the gradient data, l is the number of dimensions, M is the number of users, and dis(·) is the distance function. The device performs reliability verification through the above steps, so that the reliability of the data can be determined.

[0090] Specifically, in order to obtain the components of each dimension in the ideal gradient, it is necessary to first consider the reliability of the components of the local dimension in each dimension. Common reliability verification devices: Consistency index: compare the signs and values ​​of the local gradients submitted by different users in each dimension. If there are large differences, it means that these gradients may come from unreliable users; Convergence speed index: evaluate the rate of decrease of the local gradient to the model loss function. If the convergence speed is too slow, it means that these gradients may come from unreliable users; Distance index: calculate the distance between the local gradients submitted by different users. If the distance is too large, it means that these gradients may come from unreliable users; establish a game theory model to analyze the behavioral motivations of users in the distributed training process, so as to evaluate the reliability of the local gradients submitted by users; Therefore, the corresponding reliability indicators constructed in this application are

[0091] In some optional implementations, the device further includes a third determining unit configured to determine the value of Determine the distance function, where x a =(x1,y1),x b =(x2, y2). The device determines the distance function through the above-mentioned Euclidean distance, so that the distance can be calculated more accurately.

[0092] In the specific implementation process, the distance function uses the Euclidean distance. The calculation formula of the Euclidean distance between two points under the standard two-dimensional space condition is shown in the above formula, which can be further extended to high-dimensional space, then: Therefore, this reliability analysis result allows the ideal gradient to be updated weightedly using this reliability as a weight.

[0093] In some optional implementations, when the verification is passed, the device further includes an updating unit for updating the gradient data using the reliability index of the reliability verification as a weight to update the initial computing architecture. The device performs gradient update through the above steps, thereby updating the above initial computing architecture to obtain an optimized computing architecture.

[0094] In the specific implementation process, the gradient data submitted by the user is weighted and aggregated, and the gradients with high reliability are given higher weights, thereby reducing the impact of unreliable gradients on the global ideal gradient. Federated learning technology can be used to train the model locally, and then upload the training results to the cloud server for aggregation, thereby reducing dependence on unreliable users. The structure of this application can effectively screen multi-source data sharing across cloud environments, and finally perform necessary post-processing on the output results, such as denoising and inverse transformation, to restore the required business-related information, thereby realizing distributed computing, and being able to implement multi-party computing architecture design under a more secure organizational framework.

[0095] In the actual application process, the performance testing steps of the above architecture are also included.

[0096] First, the test environment is Java 1.7.0 to build an experimental environment, and the Paillier encryption function library is used for data encryption. Based on this, a data sharing multi-party computing model in the cloud environment is built for performance testing. The cloud server settings are shown in Table 1.

[0097] To simulate multi-user participation, 8 cores and 8GB RAM are uniformly used to build user participation of different scales, and the test data uses the public dataset MNIST.

[0098] After that, the accuracy test is carried out. The reason for choosing to evaluate the unreliable user identification performance is that in practical applications, facing large-scale data streams and heterogeneous environments, it is crucial to ensure high-precision unreliable user identification. This application tests three unreliable user ratios of 30%, 20% and 10%, and conducts experiments under low, medium and high concurrency conditions respectively. Specifically, the three cases of 100 total users (low concurrency), 300 total users (medium concurrency) and 500 total users (high concurrency) are considered in order to comprehensively evaluate the performance of the proposed secure multi-party computing architecture design in dealing with different scales and proportions. For a higher proportion of unreliable users, the data is organized as follows Figure 3 As shown. When faced with a higher proportion of unreliable users, it can be noticed that the recognition accuracy is relatively low. This may be due to the challenge faced by the system in dealing with a large number of potentially malicious behaviors, resulting in a certain degree of misjudgment. In this case, the model needs more iterations to learn from the noise and improve its prediction ability. Nevertheless, as the number of iterations increases and more data is taken into account, after more than 10 iterations, the accuracy at different concurrency scales has exceeded 20%, and after 20 iterations, it generally exceeds 95%, reflecting a gradual improvement in accuracy. For a medium level of unreliable user ratio, the data is organized as follows Figure 4As shown. As the proportion of unreliable users drops to 20%, we see an improvement in overall recognition accuracy. This shows that the system begins to distinguish between real and abnormal behaviors more effectively, and the model can converge to a high-precision state more quickly. When the number of iterations exceeds 5, the accuracy under high concurrency conditions has quickly risen to nearly 80%. At this time, the accuracy of data at different concurrency scales also has significant differences, that is, higher concurrency provides a larger scale of training samples for model optimization. After 10 iterations, the medium and high level concurrency data can already achieve a high accuracy of 90% or above, and can also reach more than 90% under low concurrency conditions. After 15 iterations, the difference in first-mover advantages of different concurrency scales has narrowed, and the overall accuracy has approached 100%, reflecting the high accuracy of the overall performance of the model. Therefore, the comprehensive results show that in this case the model has a certain degree of robustness, and can achieve good results with a small amount of data. For a high level of unreliable user ratio, the data is organized as follows Figure 5 As shown. When there are only 10% untrusted users, the system shows the best recognition accuracy. At this time, the model obviously benefits from fewer interference factors and can make correct predictions quickly and stably. At this time, the difference between different concurrency scales is further expanded. After 10 iterations, the accuracy under low concurrency conditions is slightly close to 100%, and when the iterations exceed 20 times, the data under all concurrency conditions has stabilized at 100% accuracy level. Therefore, the experimental results strongly suggest that the secure multi-party computing architecture design has a strong adaptability in dealing with small-scale potential malicious participants.

[0099] Finally, we conducted a complexity analysis and conducted corresponding tests on the changes in system running time and communication overhead under different gradient levels and concurrent user scales to verify the complexity of the algorithm operation itself. Figure 6 As shown. As the gradient increases, the system running time gradually increases. This is because more gradients need to be calculated and updated, resulting in more calculations. A slower rise rate was observed at 5000 gradients. This may be because the system has reached a certain saturation state at this time, and the impact of each additional gradient added after that is relatively weakened. As the number of concurrent users increases, the system running time is also extended. When processing more simultaneous requests, the server needs to allocate more resources to process each request. For 5000 concurrent users, under this large-scale load condition, the system may encounter bottlenecks or performance limitations, resulting in a smaller degree of progress rate. The test results of communication overhead are shown in Figure 7As shown. At different gradient levels, the communication overhead shows a clear linear growth trend. This is because more parameters need to be transmitted and updated, resulting in an increase in the amount of data transmitted. Although a slower rate of increase was observed in the case of 5000 model parameters, there is still a clear linear relationship. This shows that there will be corresponding communication costs even at a larger scale. As more users participate in the operation at the same time, the bandwidth and data transmission capacity requirements will increase accordingly. Each user sends a request to the server and receives a response, which consumes a certain amount of network resources. A higher number of concurrent users is often accompanied by higher communication costs. In actual system design, it is necessary to consider how to effectively manage and optimize these communication overheads to maintain the overall efficiency of the system.

[0100] The computing architecture determination device for cross-cloud environment data sharing includes a processor and a memory, and the above-mentioned computing unit, the first determination unit, the second determination unit, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to implement corresponding functions. The above-mentioned modules are all located in the same processor; or, the above-mentioned modules are located in different processors in any combination.

[0101] The processor contains a kernel, which calls the corresponding program unit from the memory. One or more kernels can be set, and the security of data in the cross-cloud environment data calculation process can be improved by adjusting the kernel parameters.

[0102] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0103] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is run, the device where the computer-readable storage medium is located is controlled to execute the computing architecture determination method for cross-cloud environment data sharing.

[0104] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a program stored in the memory and executable on the processor, and a method for determining a computing architecture for implementing cross-cloud environment data sharing when the processor executes the program.

[0105] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0106] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the computing architecture determination method for cross-cloud environment data sharing described in various embodiments of the present application.

[0107] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0108] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0109] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0112] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0113] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0114] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0115] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0116] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0117] 1) The method for determining the computing architecture for cross-cloud environment data sharing of the present application, establishes an initial computing architecture, performs gradient calculation on multiple cloud environment data on the initial computing architecture, and obtains gradient data corresponding to each cloud environment data; obtains an ideal gradient, compares the ideal gradient with the gradient data, determines unsafe data based on the comparison result, and removes unsafe data to obtain safe data, wherein unsafe data represents unsafe data existing in cloud environment data, and safe data represents safe data existing in cloud environment data; performs reliability verification on safe data, and when the verification passes, determines that the computing architecture for cross-cloud environment data sharing is the initial computing architecture. Compared with the prior art, which cannot both guarantee the security of data on each platform and coordinate the data resources of each platform for big data computing, the present application evaluates and verifies the security data of the initial computing architecture, and further determines the computing architecture for cross-cloud environment data sharing. Therefore, the security of data in the calculation process of cross-cloud environment data can be improved.

[0118] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for determining a computing architecture for cross-cloud environment data sharing, characterized in that: include: Establishing an initial computing architecture, performing gradient calculation on multiple cloud environment data on the initial computing architecture, and obtaining gradient data corresponding to each of the cloud environment data; Obtain an ideal gradient, compare the ideal gradient with the gradient data, determine unsafe data according to the comparison result, and remove the unsafe data to obtain safe data, wherein the ideal gradient is a gradient corresponding to a case where all the cloud environment data are safe, the unsafe data represents unsafe data existing in the cloud environment data, and the safe data represents safe data existing in the cloud environment data; The reliability of the security data is verified, and if the verification passes, the computing architecture for sharing the cross-cloud environment data is determined to be the initial computing architecture.

2. The method for determining a computing architecture for cross-cloud environment data sharing according to claim 1, characterized in that: Establish the initial computing architecture, including: Obtaining user identity authentication information, verifying the user identity authentication information, and sending a login permission identifier to the user if the verification passes, so as to establish an access control policy for the initial computing architecture; During the process of data transmission between users, target transmission data is encrypted and the encrypted data is calculated to establish an encryption communication strategy of the initial calculation architecture; The computing resources of the cloud environment are allocated through an intelligent scheduler to establish an intelligent scheduling strategy for the initial computing architecture.

3. The method for determining a computing architecture for cross-cloud environment data sharing according to claim 1, characterized in that: Performing gradient calculation on multiple cloud environment data on the initial computing architecture includes: By formula Performing gradient calculation on the cloud environment data, wherein x is an independent variable, y is a dependent variable, and f(x, y) represents a functional relationship between x and y. Represents the partial derivative of the function f(x,y).

4. The method for determining a computing architecture for cross-cloud environment data sharing according to claim 1, characterized in that: Comparing the ideal gradient with the gradient data, and determining unsafe data according to the comparison result, comprises: By formula The ideal gradient is compared with the gradient data, where x a represents the ideal gradient, x b represents the gradient data, sgn(·) represents the sign extraction function, and I(·) represents the judgment function.

5. The method for determining a computing architecture for cross-cloud environment data sharing according to claim 1, characterized in that: The reliability of the security data is verified, including: By formula The reliability of the security data is verified, wherein: is the reliability index, x a represents the ideal gradient, x b represents the gradient data, l is the number of dimensions, M is the number of users, and dis(·) is the distance function.

6. The method for determining a computing architecture for cross-cloud environment data sharing according to claim 5, characterized in that: The method further comprises: By formula Determine the distance function, where x a =(x1,y1),x b =(x2,y2)。 7. The method for determining a computing architecture for cross-cloud environment data sharing according to claim 1, characterized in that: If the verification is successful, the method further includes: The reliability index of the reliability verification is used as a weight to update the gradient data so as to update the initial calculation architecture.

8. A computing architecture determination device for cross-cloud environment data sharing, characterized in that: include: A computing unit, used to establish an initial computing architecture, and perform gradient computing on the cloud environment data on the initial computing architecture to obtain gradient data; a first determining unit, configured to obtain an ideal gradient, compare the ideal gradient with the gradient data, determine unsafe data, and remove the unsafe data to obtain safe data, wherein the ideal gradient is a gradient corresponding to a case where all the cloud environment data are safe, the unsafe data represents unsafe data existing in the cloud environment data, and the safe data represents safe data existing in the cloud environment data; The second determination unit is used to verify the reliability of the security data, and if the verification is passed, determine the computing architecture for sharing the cross-cloud environment data as the initial computing architecture.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the computing architecture determination method for cross-cloud environment data sharing as described in any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a computing architecture determination method for executing cross-cloud environment data sharing as described in any one of claims 1 to 7.

Images