A privacy-protected data transmission method, device, electronic device, and medium
By using compression sensing algorithm on the sending device to desensitize the target data and using the block sparse Bayesian learning algorithm to perform sparse reconstruction on the receiving device, the risk of privacy data leakage during user information transmission is solved. , realizes privacy protection for data transmission.
Patent Information
- Application Number
- CN202211732520.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-12-30
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2042-12-30
AI Technical Summary
During the transmission of user information, there is a risk of privacy data leakage, and a data transmission solution that can provide privacy protection is needed.
By applying a compression sensing algorithm on the transmitting device, the target data is desensitized and compressed, and the desensitization signal is obtained and sent to the receiver device. The receiving device uses the block sparse Bayesian learning algorithm to sparsely reconstruct the desensitized signal, and obtains the reconstructed signal for data processing.
It realizes effective protection of private data during transmission to prevent data leakage. Even if the desensitized signal is intercepted during transmission, it will be difficult for ordinary interceptors to reconstruct, thereby ensuring data privacy.
Smart Images

Figure CN116015942B_ABST
Abstract
Description
Technical Field
[0001] This document relates to the field of signal processing technology, and in particular to a privacy-protected data transmission method, device, electronic device and medium. Background Art
[0002] With the development of computer information technology, big data applications based on user information are widely used. However, user information is sensitive data. In many scenarios, the user terminal needs to send the user information to the server for data processing, which poses a risk of privacy leakage during the transmission process.
[0003] In view of this, there is an urgent need for a data transmission solution that can provide privacy protection. Summary of the invention
[0004] The embodiments of this specification provide a privacy-protected data transmission method, device, electronic device, and medium, which can provide a privacy-protected data transmission service.
[0005] To solve the above technical problems, the embodiments of this specification are implemented as follows:
[0006] In the first aspect, a privacy-preserving data transmission method is proposed, comprising:
[0007] The sending end device desensitizes and compresses the original signal after the target data is divided into blocks based on the compressed sensing algorithm to obtain a desensitized signal corresponding to the target data;
[0008] The sending end device sends the desensitized signal to the receiving end device;
[0009] The receiving end device performs sparse reconstruction of compression and restoration on the desensitized signal based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, wherein the reconstructed signal is used for data processing.
[0010] In the second aspect, a privacy-preserving data transmission method is proposed, including:
[0011] Based on the compressed sensing algorithm, the original signal after the target data is divided into blocks is desensitized and compressed to obtain a desensitized signal corresponding to the target data;
[0012] The desensitized signal is sent to a receiving device; wherein the receiving device performs sparse reconstruction of the desensitized signal by compression and restoration based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, and the reconstructed signal is used for data processing.
[0013] In the third aspect, a privacy-preserving data transmission method is proposed, including:
[0014] Receive a desensitized signal corresponding to the target data sent by the sending end device, where the desensitized signal is obtained by desensitizing and compressing the original signal after the target data is divided into blocks based on a compressed sensing algorithm;
[0015] The desensitized signal is compressed and restored by sparse reconstruction based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, wherein the reconstructed signal is used for data processing.
[0016] In a fourth aspect, a privacy-preserving data transmission device is proposed, comprising:
[0017] The compression module desensitizes and compresses the original signal after the target data is divided into blocks based on the compressed sensing algorithm to obtain a desensitized signal corresponding to the target data;
[0018] A sending module sends the desensitized signal to a receiving device; wherein the receiving device performs sparse reconstruction of the desensitized signal by compression and restoration based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, and the reconstructed signal is used for data processing.
[0019] In a fifth aspect, an electronic device is provided, comprising: a processor; and a memory arranged to store computer executable instructions, wherein when the executable instructions are executed, the processor performs the following operations:
[0020] Based on the compressed sensing algorithm, the original signal after the target data is divided into blocks is desensitized and compressed to obtain a desensitized signal corresponding to the target data;
[0021] The desensitized signal is sent to a receiving device; wherein the receiving device performs sparse reconstruction of the desensitized signal by compression and restoration based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, and the reconstructed signal is used for data processing.
[0022] In a sixth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores one or more programs, and when the one or more programs are executed by an electronic device including a plurality of application programs, the electronic device performs the following operations:
[0023] Based on the compressed sensing algorithm, the original signal after the target data is divided into blocks is desensitized and compressed to obtain a desensitized signal corresponding to the target data;
[0024] The desensitized signal is sent to a receiving device; wherein the receiving device performs sparse reconstruction of the desensitized signal by compression and restoration based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, and the reconstructed signal is used for data processing.
[0025] In a seventh aspect, a privacy-preserving data transmission device is proposed, comprising:
[0026] A receiving module receives a desensitized signal corresponding to the target data sent by a sending end device, wherein the desensitized signal is obtained by desensitizing and compressing the original signal after the target data is divided into blocks by the sending end device based on a compressed sensing algorithm;
[0027] The reconstruction module performs sparse reconstruction of the desensitized signal by compression and restoration based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, wherein the reconstructed signal is used for data processing.
[0028] In an eighth aspect, an electronic device is provided, comprising: a processor; and a memory arranged to store computer executable instructions, wherein when the executable instructions are executed, the processor performs the following operations:
[0029] Receive a desensitized signal corresponding to the target data sent by the sending end device, where the desensitized signal is obtained by desensitizing and compressing the original signal after the target data is divided into blocks based on a compressed sensing algorithm;
[0030] The desensitized signal is compressed and restored by sparse reconstruction based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, wherein the reconstructed signal is used for data processing.
[0031] In a ninth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores one or more programs, and when the one or more programs are executed by an electronic device including a plurality of application programs, the electronic device performs the following operations:
[0032] Receive a desensitized signal corresponding to the target data sent by the sending end device, where the desensitized signal is obtained by desensitizing and compressing the original signal after the target data is divided into blocks based on a compressed sensing algorithm;
[0033] The desensitized signal is compressed and restored by sparse reconstruction based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, wherein the reconstructed signal is used for data processing.
[0034] Based on the scheme of the embodiments of this specification, the sending end device compresses the target data using a compressed sensing algorithm with low computing power requirements, so that under limited resource conditions, the high-dimensional original signal of the target data can be quickly compressed into a low-dimensional signal to achieve desensitization processing; after that, the sending end device sends the compressed desensitized signal of the target data to the receiving end device, and the receiving end device continuously optimizes and derives the non-sparse position of the desensitized signal through a block sparse Bayesian learning algorithm with high computing power requirements, thereby performing sparse reconstruction to try to restore the reconstructed signal that does not affect the business, and finally performing business-related data processing based on the reconstructed signal. Since in the scheme, except for the sending end device, other devices cannot know the non-sparse position of the desensitized signal, even if the desensitized signal is intercepted during transmission, it is difficult for ordinary interceptors to provide equipment that supports the operation of the sparse Bayesian learning algorithm to attempt to reconstruct, thereby achieving the role of privacy protection. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] The drawings described herein are used to provide a further understanding of this specification and constitute a part of this specification. The illustrative embodiments and descriptions of this specification are used to explain this specification and do not constitute an improper limitation on this specification. In the drawings:
[0036] Figure 1 This is a schematic diagram of a first flow chart of a privacy-protected data transmission method according to an embodiment of this specification.
[0037] Figure 2 This is a second flow chart of the privacy protection data transmission method according to an embodiment of this specification.
[0038] Figure 3 This is a schematic diagram of the first structure of a privacy-protected data transmission device according to an embodiment of this specification.
[0039] Figure 4 This is a second flow chart of the privacy protection data transmission device according to an embodiment of this specification.
[0040] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this specification. DETAILED DESCRIPTION
[0041] In order to make the purpose, technical solutions and advantages of this document clearer, the technical solutions of this specification will be clearly and completely described in combination with the specific embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this document, not all of them. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this document.
[0042] As mentioned above, with the development of computer information technology, big data applications based on user information are widely used. User information is sensitive data. In many scenarios, the user terminal needs to send the user information to the server for data processing, which poses a risk of privacy leakage during the transmission process.
[0043] Taking face recognition as an example, when the user terminal needs to initiate face verification on the server side, it needs to send the user's face image collected locally to the server side for identity recognition. Once the facial features in the face image are leaked during transmission, they may be maliciously stolen.
[0044] In view of this, this specification aims to provide a data transmission solution that can provide privacy protection. Among them, the data transmission scenarios of this specification include: a sending device and a receiving device. The sending device refers to a device that sends private data (hereinafter referred to as target data), which may include a source device that generates target data, such as the user device in the above-mentioned face recognition scenario, and may also include an intermediate node device that forwards the target data, such as a router, a gateway, etc. The receiving device refers to a device that processes the target data, such as a server in the above-mentioned face recognition scenario.
[0045] Here, considering that the sending device usually does not have sufficient computing power and time resources to desensitize the target data, in order to achieve a more ideal privacy protection effect, encryption and ordering should be left to the sending device only, and decryption with higher difficulty should be performed on the receiving device side. Based on this, in the data transmission scheme of this specification, the sending device uses a compressed sensing algorithm with low computing power requirements to compress the target data, so that under limited resource conditions, the high-dimensional original signal of the target data can be quickly compressed into a low-dimensional signal to achieve desensitization processing; after that, the sending device sends the compressed desensitized signal of the target data to the receiving device, and the receiving device continuously optimizes and derives the non-sparse position of the desensitized signal through a block sparse Bayesian learning algorithm with high computing power requirements, so as to perform sparse reconstruction to try to restore the reconstructed signal that does not affect the business, and finally perform business-related data processing based on the reconstructed signal.
[0046] Specifically, based on the above principle, an embodiment of this specification provides a privacy-protected data transmission method. Figure 1 It is a flow chart of the data transmission method, which specifically includes the following steps:
[0047] S102, the sending end device desensitizes and compresses the original signal after the target data is divided into blocks based on a compressed sensing algorithm to obtain a desensitized signal corresponding to the target data.
[0048] The basic idea of compressed sensing is that as long as the high-dimensional signal is sparse or compressible, a measurement matrix can be used to represent the high-dimensional signal in a low-dimensional space. In this step, the transmitting device can project the original signal after the target data is block-sparsely represented by the structure dictionary into a low-dimensional space based on the observation matrix, so as to perform desensitization processing by compressing the low-dimensional space, thereby obtaining a desensitized signal.
[0049] Specifically, the model formula of the compressed sensing algorithm is: y=ΦΨs+n=Φx+n.
[0050] In the above formula: y is the desensitized signal, also known as the observation vector. Φ is the observation matrix. Ψ is the structural dictionary used to make changes to other domains, such as the discrete pre-transform DCT dictionary, the Gabor transform gabor dictionary, etc.; it should be understood that the original signal of the image class is not sparse in the time domain, but it will show sparsity after DCT transform or gabor transform to other domains. s is the sparsity with block structure in the transform domain corresponding to the structural dictionary. n is Gaussian noise. The embodiment of this specification introduces Gaussian noise to ensure that the output corresponding to the same input is different each time, thereby preventing black market cracking. x is the original signal, x=Ψs.
[0051] In the embodiment of this specification, the target data is subdivided into multiple blocks, and the corresponding original signal x can be expressed as follows according to the blocks: x=x 11 +x 12 , x 1L , ..., x G1 +x G2 , x GL ; Where G is the number of blocks of target data, L represents the number of elements (signal points) in each block, and x G Represents the block signal of the g-th block of the original signal x, x GL represents the lth element in the gth block of the original signal x.
[0052] S104, the sending end device sends the desensitized signal to the receiving end device.
[0053] It should be understood that, except for the transmitting device, other devices cannot know the non-sparse positions in all block signals, which requires continuous optimization and derivation of non-sparse positions based on the Bayesian learning algorithm, so as to reconstruct the desensitized signal. The Bayesian learning algorithm has high requirements for computing power. Even if the desensitized signal is intercepted during transmission, it is difficult for ordinary interceptors to provide equipment that supports the operation of the sparse Bayesian learning algorithm, thereby achieving the role of privacy protection.
[0054] S106, the receiving end device performs sparse reconstruction of compression and restoration on the desensitized signal based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, wherein the reconstructed signal is used for data processing.
[0055] It should be understood that sparse reconstruction refers to the process of reconstructing sparse signals from observation vectors, and sparse Bayesian learning (SBL) is one of the methods for sparse signal reconstruction. The model of sparse Bayesian learning is y=Φx+n mentioned above, and sparse reconstruction can be understood as estimating x based on known y and Φ.
[0056] In the compressed sensing problem under the sparse Bayesian framework, introducing different prior distributions for the estimated x has a great impact on the entire optimization problem and its structure, that is, different prior distributions correspond to different optimization problems.
[0057] Here, the embodiment of this specification adopts Spike and Slab as the prior distribution of x.
[0058] Among them, the blocks of x are divided into sparse blocks and non-sparse blocks according to sparsity. Sparse blocks refer to blocks where all elements are sparse elements; non-sparse blocks refer to blocks where non-sparse elements exist. The Spike function formula is y=N(x|0,1), and the Slab function formula is y=I(x=0); under the prior distribution of Spike and Slab, there are the following constraints:
[0059] x g ~(1-γ g )Ⅱ(x g =0)γ g Laplace(0,2σ 2 λ -1 I L ), where σ 2 is the variance of Gaussian noise; λ is the coefficient; γ g is the sparsity indicator variable γ of the g-th block g =0 or 1; g follows the Bernoulli distribution, that is:
[0060] Among them, κ is a hyperparameter; in the g-th group of sparse signals, all elements share a γ g .
[0061] For the embodiments of this specification, the receiving device uses a preset prior distribution as a constraint on the original signal of the target data to determine the maximum a posteriori probability (MAP) derivation of the original signal x of each block corresponding to the target data and the sparsity indicator variable γ of each block.
[0062] Here, assuming x * and γ * is the optimized value, then the maximum posterior probability is derived as:
[0063]
[0064] in,
[0065] The above MAP problem contains l1 regularization, which can induce signal sparsity better than l2 regularization. If ρ = [ρ1, ρ2, ... ρ g , ... ρ G ] are equal, then the last term in the formula becomes the l0-norm constraint of the parameter γ. Therefore, the optimization problem corresponding to the formula is non-convex and cannot be solved using conventional optimization algorithms. Therefore, it is necessary to convert the maximum a posteriori probability derivation into a cost function based on the lasso algorithm, in which the variables of the cost function are associated with the non-sparse block division results of the target data; then, the cost function is optimized and solved for multiple rounds of iterations to determine the non-sparse block division results of the target data corresponding to the cost function after the iteration. It should be understood that the non-sparse block division results of the labeled data can represent the non-sparse position in the desensitized signal. Based on the non-sparse block division results, the receiving end device can perform sparse reconstruction of the compressed and restored desensitized signal to obtain a reconstructed signal.
[0066] The iterative process of the cost function is introduced in detail below.
[0067] Exemplarily, the embodiments of this specification use a greedy algorithm to quickly perform multiple rounds of iterative optimization solutions on the cost function.
[0068] First, the receiving device initializes the non-sparse block partitioning result of the target data.
[0069] Afterwards, the receiving end device performs multiple rounds of iterations according to the principle of the greedy algorithm: selecting a target block corresponding to the target data that has the greatest impact on the cost function, and changing the sparsity indicator variable of the target block, wherein if the sparsity indicator variable of the target block before the change indicates that the target block is a sparse block, then after the change it indicates a non-sparse block; if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then after the change it indicates a sparse block; after the sparsity indicator variable of the target block is changed, if the cost function produces an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated.
[0070] For ease of understanding, we can define the collection here g∈{1,2,...G}. Set That is, the non-sparse block partitioning result of the target data.
[0071] In a known set Under the premise of which non-sparse blocks are included in , it is equivalent to removing the last item in the optimization problem. The optimization problem can be further simplified. Therefore, a greedy algorithm can be proposed to obtain the optimal set And in the collection Let’s solve the MAP problem.
[0072] The specific solution idea is By adding a new block of target data, or from Remove a block from the solution to make it a sparse block solution; after obtaining a new After the collection, solve the problem with l1 norm, and then iterate and loop to form a greedy algorithm architecture.
[0073] When the positions of non-sparse blocks are known, it is assumed that the optimization problem is simplified to the lasso problem, that is:
[0074]
[0075] Here is defined And set the cost function
[0076] Whether it is to combine Add a block to the collection or Remove a block from the target, and the selected block must be able to make the objective function Therefore, the iterative process is as follows Figure 2 As shown, including:
[0077] S21, Initialize the collection
[0078] For example, assuming that the target data corresponds to 5 blocks, two blocks can be randomly selected, such as blocks 3 and 4, as non-sparse blocks. The corresponding set It is {block 3, block 4}.
[0079] S22, traverse all blocks of the target data, solve the lasso problem, and find the The target block with the greatest impact.
[0080] Assuming that the target data corresponds to 5 blocks, the initial set For {block 3, block 4}, in this step, select one block at a time to change its sparsity and try to calculate the change The target block with the largest value.
[0081] for example:
[0082] First, select block 1. Block 1 was originally a sparse block, so change block 1 to a non-sparse block. At this time, the non-sparse blocks include block 1, block 3, and block 4, and recalculate Confirm recalculation Compared with the case where only blocks 3 and 4 are non-sparse blocks The amount of change.
[0083] After that, select block 2. Block 2 was originally a sparse block, so change block 2 to a non-sparse block. At this time, the non-sparse block includes block 2, block 3 and block 4, and recalculate Confirm recalculation Compared with the case where only blocks 3 and 4 are non-sparse blocks The amount of change.
[0084] Then, select block 3. Block 3 was originally a non-sparse block, so change block 3 to a sparse block. At this time, the non-sparse block includes sub-block 4, and recalculate Confirm recalculation Compared with the case where only blocks 3 and 4 are non-sparse blocks The amount of change.
[0085] …
[0086] Finally, select block 5. Block 5 was originally a sparse block, so change block 5 to a non-sparse block. At this time, the non-sparse block includes block 5, block 3, and block 4, and recalculate Confirm recalculation Compared with the case where only blocks 3 and 4 are non-sparse blocks The amount of change.
[0087] Among them, The block with the largest change is the target block.
[0088] S23, changing the sparsity indicator variable γ of the target block.
[0089] That is, if the γ before the target block is changed indicates that the target block is a sparse block, then after the change it indicates that it is a non-sparse block. If the γ before the target block is changed indicates that the target block is a non-sparse block, then after the change it indicates that it is a sparse block.
[0090] S24, determine the corresponding target block γ value after transformation Is it smaller than the previous
[0091] If it is less than, the iteration has not ended yet, and S25 is executed; otherwise, the iteration ends (it can also be actively ended after reaching a certain number of rounds).
[0092] S25, Update collection
[0093] For example, the original collection is {block 3, block 4}. If the target block in this round is block 3, then the updated set is {block 4}; for example, the original set is {block 3, block 4}. If the target block in this round is block 2, then the updated set It is {block 2, block 3, block 4}.
[0094] It should be understood that the above S21 to S25 are based on the greedy algorithm In the process of finding the optimal solution, there is no need to exhaustively enumerate all possible sets. Therefore, it has high efficiency and can shorten the time taken by the receiving device to perform sparse reconstruction to a certain extent.
[0095] In summary, based on the method of the embodiment of this specification, the sending end device uses a compressed sensing algorithm with low computing power requirements to compress the target data, so that under limited resource conditions, the high-dimensional original signal of the target data can be quickly compressed into a low-dimensional signal to achieve desensitization processing; after that, the sending end device sends the compressed desensitized signal of the target data to the receiving end device, and the receiving end device uses a block sparse Bayesian learning algorithm with high computing power requirements to continuously optimize and derive the non-sparse position of the desensitized signal, thereby performing sparse reconstruction to try to restore the reconstructed signal that does not affect the business, and finally perform business-related data processing based on the reconstructed signal. Since in the scheme, except for the sending end device, other devices cannot know the non-sparse position of the desensitized signal, even if the desensitized signal is intercepted during transmission, it is difficult for ordinary interceptors to provide equipment that supports the operation of the sparse Bayesian learning algorithm to attempt to reconstruct, thereby achieving the role of privacy protection.
[0096] The above is an exemplary introduction to the data transmission method of the embodiment of this specification. It should be understood that the method of the embodiment of this specification can be applied to any application scenario that requires sending private data. For example, a scenario initiated by a user terminal that requires a cloud server to perform face verification. For the user terminal as a sending device, a program configured with a compressed sensing algorithm can be called in a local trusted execution environment (Tee) to desensitize and compress the collected user face image; after that, the user terminal sends the desensitized signal of the compressed user face image to the server as the receiving device through the mobile network, and the server calls its own computing resources to run a block sparse Bayesian learning algorithm to try to sparsely reconstruct the desensitized signal to obtain a reconstructed signal that does not affect face verification, thereby identifying the identity of the user of the user terminal based on the reconstructed signal.
[0097] In addition, Figure 1 Corresponding to the method shown, the embodiment of this specification also provides a data transmission device with privacy protection. Figure 3 : is a schematic diagram of the structure of a privacy protection data transmission device 300 provided in an embodiment of this specification, including:
[0098] The compression module 310 desensitizes and compresses the original signal after the target data is divided into blocks based on the compressed sensing algorithm to obtain a desensitized signal corresponding to the target data.
[0099] The sending module 320 sends the desensitized signal to a receiving device; wherein the receiving device performs sparse reconstruction of compression and restoration on the desensitized signal based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, and the reconstructed signal is used for data processing.
[0100] Optionally, the compression module 310 performs low-dimensional space projection on the original signal after the target data is block-sparsely represented by the structure dictionary based on the observation matrix to obtain a desensitized signal.
[0101] It should be understood that the data transmission device in the embodiment of this specification is Figure 1 The execution subject of the steps corresponding to the sending end device side in the method shown can thus be realized Figure 1 The steps and functions of the method shown.
[0102] In addition, Figure 1 Corresponding to the method shown, the embodiment of this specification also provides a data transmission device with privacy protection. Figure 4 4 is a schematic diagram of a privacy-protected data transmission device 400 provided in one embodiment of the present specification, including:
[0103] The receiving module 410 receives a desensitized signal corresponding to the target data sent by the sending end device, wherein the desensitized signal is obtained by desensitizing and compressing the original signal after the target data is divided into blocks by the sending end device based on a compressed sensing algorithm;
[0104] The reconstruction module 420 performs sparse reconstruction of the desensitized signal by compression and restoration based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, wherein the reconstructed signal is used for data processing. Optionally,
[0105] Optionally, the reconstruction module 420 performs sparse reconstruction of compression and restoration on the desensitized signal based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, including: using a preset prior distribution as a constraint on the original signal of the target data, determining the maximized a posteriori probability derivation of the original signal of each block corresponding to the target data and the sparsity indicator variable of each block; converting the maximized a posteriori probability derivation into a cost function based on the lasso algorithm, wherein the variables of the cost function are associated with the non-sparse block division result of the target data; performing multiple rounds of iterative optimization and solution on the cost function to determine that the non-sparse block division result of the target data corresponds to the cost function after the iteration, wherein the non-sparse block refers to a block in which non-sparse elements exist in the target data; based on the determined non-sparse block division result of the target data, performing sparse reconstruction of compression and restoration on the desensitized signal to obtain a reconstructed signal.
[0106] Optionally, the receiving device of the reconstruction module 420 performs multiple rounds of iterative optimization solution on the cost function, including: initializing the non-sparse block division result of the target data; multiple rounds of iterations to execute: selecting a target block corresponding to the target data that has the greatest impact on the cost function, and changing the sparsity indicator variable of the target block, wherein if the sparsity indicator variable of the target block before the change indicates that the target block is a sparse block, then after the change it indicates a non-sparse block, and if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then after the change it indicates a sparse block; after the sparsity indicator variable of the target block is changed, if the cost function produces an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated, wherein a sparse block refers to a block in which all elements in the target data are sparse elements.
[0107] Optionally, the preset prior distribution is a Spike and Slab prior distribution.
[0108] Optionally, the target data includes facial image data, and the data processing includes facial recognition processing.
[0109] It should be understood that the data transmission device in the embodiment of this specification is Figure 1 The execution subject of the steps corresponding to the receiving device side in the method shown can thus be realized Figure 1 The steps and functions of the method shown.
[0110] Figure 5 This is a schematic diagram of the structure of an electronic device provided by an embodiment of this specification. Figure 5 At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. Of course, the electronic device may also include hardware required for other services.
[0111] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.
[0112] The memory is used to store the program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.
[0113] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming the above-mentioned Figure 3 The data transmission device shown. The processor executes the program stored in the memory and is specifically used to perform the following operations:
[0114] Based on the compressed sensing algorithm, the original signal after the target data is divided into blocks is desensitized and compressed to obtain a desensitized signal corresponding to the target data.
[0115] The desensitized signal is sent to a receiving device; wherein the receiving device performs sparse reconstruction of the desensitized signal by compression and restoration based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, and the reconstructed signal is used for data processing.
[0116] Alternatively, the processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming the above-mentioned Figure 4 The data transmission device shown. The processor executes the program stored in the memory and is specifically used to perform the following operations:
[0117] A desensitized signal corresponding to target data sent by a sending end device is received, wherein the desensitized signal is obtained by desensitizing and compressing an original signal after the target data is divided into blocks based on a compressed sensing algorithm.
[0118] The desensitized signal is compressed and restored by sparse reconstruction based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, wherein the reconstructed signal is used for data processing.
[0119] The above is as in this manual Figure 1The method disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in one or more embodiments of this specification can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in one or more embodiments of this specification can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.
[0120] The electronic device may also perform Figure 1 The method will not be described in detail in this manual.
[0121] Of course, in addition to software implementation, the electronic device of this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.
[0122] The embodiments of this specification also provide a computer-readable storage medium, which stores one or more programs.
[0123] The one or more programs include instructions that, when executed by a portable electronic device including a plurality of application programs, enable the portable electronic device to execute Figure 1 The method on the transmitting end device side of the embodiment shown is specifically used to perform the following operations:
[0124] Based on the compressed sensing algorithm, the original signal after the target data is divided into blocks is desensitized and compressed to obtain a desensitized signal corresponding to the target data;
[0125] The desensitized signal is sent to a receiving device; wherein the receiving device performs sparse reconstruction of the desensitized signal by compression and restoration based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, and the reconstructed signal is used for data processing.
[0126] Alternatively, the one or more programs include instructions that, when executed by a portable electronic device including a plurality of application programs, enable the portable electronic device to execute Figure 1 The method on the receiving end device side of the illustrated embodiment is specifically used to perform the following operations:
[0127] Receive a desensitized signal corresponding to the target data sent by the sending end device, where the desensitized signal is obtained by desensitizing and compressing the original signal after the target data is divided into blocks based on a compressed sensing algorithm;
[0128] The desensitized signal is compressed and restored by sparse reconstruction based on a block sparse Bayesian learning algorithm to obtain a reconstructed signal corresponding to the target data, wherein the reconstructed signal is used for data processing.
[0129] In short, the above description is only a preferred embodiment of this specification and is not intended to limit the protection scope of this specification. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of this specification shall be included in the protection scope of one or more embodiments of this specification.
[0130] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
[0131] 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.
[0132] 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.
[0133] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
Claims
1. A privacy-preserving data transmission method, comprising: The sending end device desensitizes and compresses the original signal after the target data is divided into blocks based on the compressed sensing algorithm to obtain a desensitized signal corresponding to the target data; The sending end device sends the desensitized signal to the receiving end device; The receiving end device uses a preset prior distribution as a constraint for the original signal of the target data to determine the maximum a posteriori probability derivation of the original signal of each block and the sparsity indicator variable of each block corresponding to the target data; the receiving end device converts the maximum a posteriori probability derivation into a cost function, and the variables of the cost function are associated with the non-sparse block division result of the target data; the receiving end device performs multiple rounds of iterative optimization and solution on the cost function to determine the non-sparse block division result of the target data corresponding to the cost function after the iteration; the receiving end device performs sparse reconstruction of the desensitized signal by compression and restoration based on the determined non-sparse block division result of the target data to obtain a reconstructed signal, and the reconstructed signal is used for data processing; the multiple rounds of iterative optimization and solution include the receiving end device initializing the non-sparse block division result of the target data; the receiving end device performs multiple rounds of iterations to execute: selecting a target block corresponding to the target data that has the greatest impact on the cost function, and changing the sparsity indicator variable of the target block, wherein if the target block is changed If the sparsity indicator variable before indicates that the target block is a sparse block, then it indicates a non-sparse block after the change; if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then it indicates a sparse block after the change; after the sparsity indicator variable of the target block is changed, if the cost function produces an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated, and a sparse block refers to a block in which all elements in the target data are sparse elements.
2. The method according to claim 1, The cost function adopts the lasso algorithm.
3. The method according to claim 2, The preset prior distribution is the Spike and Slab prior distribution.
4. The method according to claim 1, The sending end device desensitizes and compresses the original signal after the target data is divided into blocks based on the compressed sensing algorithm to obtain a desensitized signal corresponding to the target data, including: Based on the observation matrix, the transmitting device performs low-dimensional space projection on the original signal after the target data is block-sparsely represented by the structured dictionary to obtain a desensitized signal.
5. The method according to any one of claims 1 to 4, The target data includes face image data, and the data processing includes face recognition processing.
6. A privacy-preserving data transmission method, applied to a sending device, comprising: Based on the compressed sensing algorithm, the original signal after the target data is divided into blocks is desensitized and compressed to obtain a desensitized signal corresponding to the target data; The desensitized signal is sent to a receiving device; wherein the receiving device uses a preset prior distribution as a constraint on the original signal of the target data to determine the maximum a posteriori probability derivation of the original signal of each block and the sparsity indicator variable of each block corresponding to the target data; the receiving device converts the maximum a posteriori probability derivation into a cost function, and the variables of the cost function are associated with the non-sparse block division result of the target data; the receiving device performs multiple rounds of iterative optimization and solution on the cost function to determine the non-sparse block division result of the target data corresponding to the cost function after the iteration ends; the receiving device performs sparse reconstruction of the desensitized signal by compression and restoration based on the determined non-sparse block division result of the target data to obtain a reconstructed signal, and the reconstructed signal is used for data processing; the multiple rounds of iterative optimization and solution include the receiving device initializing the non-sparse block division result of the target data; the receiving device performs multiple rounds of iterations to execute: selecting a target block corresponding to the target data that has the greatest impact on the cost function, and changing the sparsity indicator variable of the target block, wherein if the target block is changed before If the sparsity indicator variable indicates that the target block is a sparse block, then it indicates a non-sparse block after the change; if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then it indicates a sparse block after the change; after the sparsity indicator variable of the target block is changed, if the cost function produces an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated, wherein a sparse block refers to a block in which all elements in the target data are sparse elements.
7. A privacy-preserving data transmission method, applied to a receiving device, comprising: Receive a desensitized signal corresponding to the target data sent by the sending end device, where the desensitized signal is obtained by desensitizing and compressing the original signal after the target data is divided into blocks based on a compressed sensing algorithm; The original signal of the target data is constrained by a preset prior distribution, and the original signal of each block corresponding to the target data and the maximum a posteriori probability derivation of the sparsity indicator variable of each block are determined; the maximum a posteriori probability derivation is converted into a cost function, and the variables of the cost function are associated with the non-sparse block division result of the target data; the cost function is optimized and solved by multiple rounds of iterations, and it is determined that the cost function corresponds to the non-sparse block division result of the target data after the iteration ends; Based on the determined non-sparse block division result of the target data, the desensitized signal is compressed and restored to obtain a sparse reconstruction signal, and the reconstructed signal is used for data processing; The optimization solution of multiple rounds of iterations includes initializing the non-sparse block partitioning result of the target data; Multiple rounds of iterations are performed to execute: a target block corresponding to the target data that has the greatest impact on the cost function is selected, and the sparsity indicator variable of the target block is changed, wherein if the sparsity indicator variable of the target block before the change indicates that the target block is a sparse block, then it is indicated as a non-sparse block after the change, and if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then it is indicated as a sparse block after the change; after the sparsity indicator variable of the target block is changed, if the cost function generates an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated, and a sparse block refers to a block in which all elements in the target data are sparse elements.
8. A privacy-protected data transmission device, comprising: The compression module desensitizes and compresses the original signal after the target data is divided into blocks based on the compressed sensing algorithm to obtain a desensitized signal corresponding to the target data; A sending module sends the desensitized signal to a receiving device; wherein the receiving device uses a preset prior distribution as a constraint for the original signal of the target data to determine the maximum a posteriori probability derivation of the original signal of each block and the sparsity indicator variable of each block corresponding to the target data; the receiving device converts the maximum a posteriori probability derivation into a cost function, and the variables of the cost function are associated with the non-sparse block division result of the target data; the receiving device performs multiple rounds of iterative optimization and solution on the cost function to determine the non-sparse block division result of the target data corresponding to the cost function after the iteration; the receiving device performs sparse reconstruction of the desensitized signal by compression and restoration based on the determined non-sparse block division result of the target data to obtain a reconstructed signal, and the reconstructed signal is used for data processing; the multiple rounds of iterative optimization and solution include the receiving device initializing the non-sparse block division result of the target data; the receiving device performs multiple rounds of iterations to execute: selecting a target block corresponding to the target data that has the greatest impact on the cost function, and changing the sparsity indicator variable of the target block, wherein if the target block is changed If the sparsity indicator variable before indicates that the target block is a sparse block, then it indicates a non-sparse block after the change; if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then it indicates a sparse block after the change; after the sparsity indicator variable of the target block is changed, if the cost function produces an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated, and a sparse block refers to a block in which all elements in the target data are sparse elements.
9. An electronic device, comprising: processor; and a memory arranged to store computer executable instructions which, when executed, cause the processor to: Based on the compressed sensing algorithm, the original signal after the target data is divided into blocks is desensitized and compressed to obtain a desensitized signal corresponding to the target data; The desensitized signal is sent to a receiving device; wherein the receiving device uses a preset prior distribution as a constraint for the original signal of the target data to determine the maximum a posteriori probability derivation of the original signal of each block and the sparsity indicator variable of each block corresponding to the target data; the receiving device converts the maximum a posteriori probability derivation into a cost function, and the variables of the cost function are associated with the non-sparse block division result of the target data; the receiving device performs multiple rounds of iterative optimization and solution on the cost function to determine the non-sparse block division result of the target data corresponding to the cost function after the iteration; the receiving device performs sparse reconstruction of the desensitized signal by compression and restoration based on the determined non-sparse block division result of the target data to obtain a reconstructed signal, and the reconstructed signal is used for data processing; the multiple rounds of iterative optimization and solution include the receiving device initializing the non-sparse block division result of the target data; the receiving device performs multiple rounds of iterations to execute: selecting a target block corresponding to the target data that has the greatest impact on the cost function, and changing the sparsity indicator variable of the target block, wherein if the target block is changed If the sparsity indicator variable before indicates that the target block is a sparse block, then it indicates a non-sparse block after the change; if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then it indicates a sparse block after the change; after the sparsity indicator variable of the target block is changed, if the cost function produces an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated, and a sparse block refers to a block in which all elements in the target data are sparse elements.
10. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of application programs, causes the electronic device to perform the following operations: Based on the compressed sensing algorithm, the original signal after the target data is divided into blocks is desensitized and compressed to obtain a desensitized signal corresponding to the target data; The desensitized signal is sent to a receiving device; wherein, The receiving end device uses a preset prior distribution as a constraint for the original signal of the target data to determine the maximum a posteriori probability derivation of the original signal of each block and the sparsity indicator variable of each block corresponding to the target data; the receiving end device converts the maximum a posteriori probability derivation into a cost function, and the variables of the cost function are associated with the non-sparse block division result of the target data; the receiving end device performs multiple rounds of iterative optimization and solution on the cost function to determine the non-sparse block division result of the target data corresponding to the cost function after the iteration; the receiving end device performs sparse reconstruction of the desensitized signal by compression and restoration based on the determined non-sparse block division result of the target data to obtain a reconstructed signal, and the reconstructed signal is used for data processing; the multiple rounds of iterative optimization and solution include the receiving end device initializing the non-sparse block division result of the target data; the receiving end device performs multiple rounds of iterations to execute: selecting a target block corresponding to the target data that has the greatest impact on the cost function, and changing the sparsity indicator variable of the target block, wherein if the target block is changed If the sparsity indicator variable before indicates that the target block is a sparse block, then it indicates a non-sparse block after the change; if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then it indicates a sparse block after the change; after the sparsity indicator variable of the target block is changed, if the cost function produces an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated, and a sparse block refers to a block in which all elements in the target data are sparse elements.
11. A privacy-protected data transmission device, comprising: A receiving module receives a desensitized signal corresponding to the target data sent by a sending end device, wherein the desensitized signal is obtained by desensitizing and compressing the original signal after the target data is divided into blocks by the sending end device based on a compressed sensing algorithm; A reconstruction module, using a preset prior distribution as a constraint for the original signal of the target data, determines the maximized a posteriori probability derivation of the original signal of each block corresponding to the target data and the sparsity indicator variable of each block; converts the maximized a posteriori probability derivation into a cost function, wherein the variables of the cost function are associated with the non-sparse block partitioning result of the target data; performs multiple rounds of iterative optimization and solution on the cost function, and determines that after the iteration, the cost function corresponds to the non-sparse block partitioning result of the target data; Based on the determined non-sparse block division result of the target data, the desensitized signal is compressed and restored to obtain a sparse reconstruction signal, and the reconstructed signal is used for data processing; The optimization solution of multiple rounds of iterations includes initializing the non-sparse block partitioning result of the target data; Multiple rounds of iterations are performed to execute: a target block corresponding to the target data that has the greatest impact on the cost function is selected, and the sparsity indicator variable of the target block is changed, wherein if the sparsity indicator variable of the target block before the change indicates that the target block is a sparse block, then it is indicated as a non-sparse block after the change, and if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then it is indicated as a sparse block after the change; after the sparsity indicator variable of the target block is changed, if the cost function generates an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated, and a sparse block refers to a block in which all elements in the target data are sparse elements.
12. An electronic device comprising: processor; and a memory arranged to store computer executable instructions which, when executed, cause the processor to: Receive a desensitized signal corresponding to the target data sent by the sending end device, where the desensitized signal is obtained by desensitizing and compressing the original signal after the target data is divided into blocks based on a compressed sensing algorithm; The original signal of the target data is constrained by a preset prior distribution, and the original signal of each block corresponding to the target data and the maximum a posteriori probability derivation of the sparsity indicator variable of each block are determined; the maximum a posteriori probability derivation is converted into a cost function, and the variables of the cost function are associated with the non-sparse block division result of the target data; the cost function is optimized and solved by multiple rounds of iterations, and it is determined that the cost function corresponds to the non-sparse block division result of the target data after the iteration ends; Based on the determined non-sparse block division result of the target data, the desensitized signal is compressed and restored to obtain a sparse reconstruction signal, and the reconstructed signal is used for data processing; The optimization solution of multiple rounds of iterations includes initializing the non-sparse block partitioning result of the target data; Multiple rounds of iterations are performed to execute: a target block corresponding to the target data that has the greatest impact on the cost function is selected, and the sparsity indicator variable of the target block is changed, wherein if the sparsity indicator variable of the target block before the change indicates that the target block is a sparse block, then it is indicated as a non-sparse block after the change, and if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then it is indicated as a sparse block after the change; after the sparsity indicator variable of the target block is changed, if the cost function generates an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated, and a sparse block refers to a block in which all elements in the target data are sparse elements.
13. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of application programs, cause the electronic device to perform the following operations: Receive a desensitized signal corresponding to the target data sent by the sending end device, where the desensitized signal is obtained by desensitizing and compressing the original signal after the target data is divided into blocks based on a compressed sensing algorithm; The original signal of the target data is constrained by a preset prior distribution, and the original signal of each block corresponding to the target data and the maximum a posteriori probability derivation of the sparsity indicator variable of each block are determined; the maximum a posteriori probability derivation is converted into a cost function, and the variables of the cost function are associated with the non-sparse block division result of the target data; the cost function is optimized and solved by multiple rounds of iterations, and it is determined that the cost function corresponds to the non-sparse block division result of the target data after the iteration ends; Based on the determined non-sparse block division result of the target data, the desensitized signal is compressed and restored to obtain a sparse reconstruction signal, and the reconstructed signal is used for data processing; The optimization solution of multiple rounds of iterations includes initializing the non-sparse block partitioning result of the target data; Multiple rounds of iterations are performed to execute: a target block corresponding to the target data that has the greatest impact on the cost function is selected, and the sparsity indicator variable of the target block is changed, wherein if the sparsity indicator variable of the target block before the change indicates that the target block is a sparse block, then it is indicated as a non-sparse block after the change, and if the sparsity indicator variable of the target block before the change indicates that the target block is a non-sparse block, then it is indicated as a sparse block after the change; after the sparsity indicator variable of the target block is changed, if the cost function generates an iterative optimization solution, then based on the change of the sparsity indicator variable of the target block, the non-sparse block division result of the target data is updated, and a sparse block refers to a block in which all elements in the target data are sparse elements.
Citation Information
Patent Citations
Physical sign data monitoring system
CN111669409A
