Time series data publishing method based on sampling filtering differential privacy protection model

Abstract

The application discloses a time series data publishing method and device based on a sampling filtering differential privacy protection model, electronic equipment and a readable storage medium, comprising: sampling the original time series data according to a plurality of different preset sampling intervals to obtain sampling data and non-sampling data; adding noise to the sampling data using a relevant Laplace mechanism to obtain noise-added data that satisfies the differential privacy protection requirement after noise addition; performing smoothing and correction processing on the noise-added data using a VMD-Wiener filtering technology to obtain corrected data; combining the corrected data corresponding to the same preset sampling interval and the non-sampling data to obtain to-be-published data corresponding to each preset sampling interval; evaluating each to-be-published data through a dynamic time warping evaluation mechanism to determine an optimal sampling interval and optimal to-be-published data corresponding to the optimal sampling interval; and publishing the optimal to-be-published data, thereby improving privacy and security.

Description

Technical Field

[0001] This application relates to the field of information technology security, and in particular to a method, apparatus, electronic device, and computer-readable storage medium for publishing time-series data based on a sampling filtering differential privacy protection model. Background Technology

[0002] With the rapid development and widespread application of big data, cloud computing, and mobile internet, various types of information data are constantly expanding and extending. The application of information technology is ubiquitous in today's society. While these technologies bring personalized services and improve user convenience, they also bring risks to the use of user data. Data breaches are increasing, and incidents of data attacks, theft, and hijacking are occurring one after another.

[0003] The formally implemented Data Security Law of the People's Republic of China will further ensure that data is effectively protected and legally utilized, better protecting the legitimate rights and interests of individuals and organizations, and safeguarding national sovereignty, security, and development interests. Therefore, a two-pronged approach combining technology and management is needed, proposing systematic countermeasures and solutions, and formulating relevant standards and implementation methods.

[0004] As a crucial component of the new power system, the electricity market, within its management information system, accumulates a vast amount of dynamic electricity data, including financial, marketing, human resources, market, and production management information. This data, originating from various users, contains a wealth of sensitive information. To ensure the safe and efficient operation of the electricity market, effective data exchange is essential, protecting both the security and usability of user data. With the advancement of the new power system's construction, the flow and collaborative application of energy and electricity market data will face increasingly diverse and complex scenarios in the future. Therefore, the proper protection of data security during the operation of the electricity market is a critical issue at present.

[0005] Differential privacy, a privacy protection concept that has emerged in recent years, makes no assumptions about the attacker's background knowledge. It aims to maximize the accuracy of data queries when accessing statistical databases while minimizing the chance of identifying records. Differential privacy provides a quantitative representation and proof of the privacy leakage risk and privacy protection effect, possessing high mathematical logic and allowing for a scientifically sound characterization of the achieved privacy protection effect.

[0006] However, in the era of big data in the electricity market, there are still many technical problems to be solved in order to fully realize the provision of relatively accurate time-series data privacy protection under the constraints of differential privacy. Summary of the Invention

[0007] The purpose of this application is to provide a method, apparatus, electronic device, and computer-readable storage medium for publishing time-series data based on a sampling filtering differential privacy protection model.

[0008] To achieve the above objectives, this application provides a time-series data publishing method based on a sampling-filtered differential privacy protection model in its first aspect. The method includes: sampling the original time-series data at multiple different preset sampling intervals to obtain sampled data and non-sampled data; adding noise to the sampled data using a correlated Laplace mechanism to obtain noise-added data that meets the differential privacy protection requirements; applying VMD-Wiener filtering technology to the noise-added data for stabilization and correction to obtain corrected data; combining the corrected data and non-sampled data corresponding to the same preset sampling interval to obtain data to be published corresponding to each preset sampling interval; evaluating each data to be published using a dynamic time warping evaluation mechanism to determine the optimal sampling interval and the optimal data to be published corresponding to the optimal sampling interval; and publishing the optimal data to be published.

[0009] To achieve the above objectives, this application provides a time-series data publishing device based on a sampling-filter differential privacy protection model in a second aspect. The device includes: a custom interval sampling unit configured to sample the original time-series data at multiple different preset sampling intervals to obtain sampled data and non-sampled data; a differential privacy protection processing unit configured to add noise to the sampled data using a correlated Laplace mechanism to obtain noise-added data that meets the differential privacy protection requirements; a stabilization and re-correction unit configured to perform stabilization and re-correction processing on the noise-added data using VMD-Wiener filtering technology to obtain corrected data; a data combination unit configured to combine the corrected data and non-sampled data corresponding to the same preset sampling interval to obtain data to be published corresponding to each preset sampling interval; an evaluation unit configured to evaluate each data to be published using a dynamic time warping evaluation mechanism to determine the optimal sampling interval and the optimal data to be published corresponding to the optimal sampling interval; and a publishing unit configured to publish the optimal data to be published.

[0010] To achieve the above objectives, this application provides an electronic device in a third aspect, the electronic device comprising:

[0011] Memory, used to store computer programs;

[0012] A processor, configured to implement, when executing a computer program stored in memory, the steps of the time-series data publishing method based on a sampling-filtered differential privacy protection model as described in the first aspect above.

[0013] To achieve the above objectives, this application provides a computer-readable storage medium in a fourth aspect, on which a computer program is stored, which, when executed by a processor, implements the steps of time-series data publication based on the sampling filter differential privacy protection model as described in the first aspect above.

[0014] Compared to existing technologies, the time-series data publishing scheme based on a sampling-filtered differential privacy protection model provided in this application firstly utilizes fixed sampling techniques to add noise and filter the data at sampling points, reducing the overall system runtime and improving system efficiency. Secondly, it utilizes a correlated Laplace mechanism to publish correlated time-series data. Because this technique leverages the indistinguishability of sequences, it generates correlated Laplace noise sequences by using four Gaussian white noise sequences interacting through a linear system, rather than generating independent noise sequences. This ensures high usability of the data through the concatenation of multiple different Gaussian white noise sequences and avoids the problem of lower-than-expected privacy due to the independent and identically distributed nature of ordinary Laplace noise. Next, it employs VMD-Wiener filtering technology to stabilize and correct the noise-added data, making the published data closer to the original values. This improves the usability of the published data while maintaining basic data privacy. Finally, it applies the concept of dynamic time warping to differential privacy, using dynamic time warping to select the optimal sampling interval, obtaining sampling points that achieve a better balance between privacy and usability in time-series data publishing.

[0015] This application also provides a time-series data publishing device, electronic device, and computer-readable storage medium based on a sampling filtering differential privacy protection model, which have the above-mentioned beneficial effects, and will not be elaborated here. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a time-series data publishing method based on a sampling filtering differential privacy protection model, provided for embodiments of this application;

[0018] Figure 2 A method provided for embodiments of this application and Figure 1 Corresponding flowchart;

[0019] Figure 3A schematic diagram illustrating a process for generating noise using the correlated Laplace mechanism, provided in an embodiment of this application;

[0020] Figure 4 This is a schematic diagram illustrating a VMD-Wiener filtering process provided in an embodiment of this application;

[0021] Figure 5 This is a structural block diagram of a time-series data publishing device based on a sampling filtering differential privacy protection model, provided in an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] By understanding electricity-related time-series data in the era of big data in the electricity market, the applicant discovered that to fully achieve relatively accurate time-series data privacy protection under the constraints of differential privacy, the following four key issues need to be addressed:

[0024] (1) How to reduce the amount of computation and improve system efficiency when processing constantly changing dynamic time-series data;

[0025] (2) How to reduce the correlation between query sets, reduce differential privacy sensitivity and the amount of noise added;

[0026] (3) How to avoid noise accumulation caused by the continuous updating of dynamic time series data, which reduces the usability of the data;

[0027] (4) How to determine the optimal sampling point.

[0028] Since the fundamental Laplace mechanism in differential privacy involves adding independent and identically distributed Laplace noise, attackers can use refinement methods (such as filtering) to remove this independent noise from related time series, thereby reducing the intended privacy protection. Especially when data is correlated, simply adding independent and identically distributed noise may not provide sufficient security and could jeopardize personal privacy.

[0029] Furthermore, time-series data is dynamic, constantly updating over time. Consequently, the published updated data inevitably contains noise from previous releases. This noise accumulates over time, causing significant differences between the final published data and the original data, thus substantially reducing the data's usability.

[0030] Therefore, it is urgent to address the problems of high data correlation, noise accumulation, and optimal sampling point in the protection application scenarios of existing differential privacy technologies for dynamic time-series data publishing.

[0031] To solve the aforementioned technical problems, this embodiment provides... Figure 1 A method for publishing time-series data based on a sampling filtering differential privacy protection model is provided, which includes the following steps:

[0032] Step 101: Sample the original time series data according to multiple different preset sampling intervals to obtain sampled data and non-sampled data.

[0033] This step aims to have an execution entity (such as a local server, cloud server, or secure computing server for data processing and analysis) suitable for implementing the time-series data publishing method based on the sampling filtering differential privacy protection model provided in this application perform multiple sampling processes on the original time-series data at multiple different preset sampling intervals, so as to obtain a pair of sampled data and non-sampled data for each preset sampling interval.

[0034] Sampling data refers to a portion of time-series data extracted according to a preset sampling interval, while non-sampling data refers to the remaining portion of time-series data after a portion of the original time-series data has been extracted. The original time-series data can be at least one of the following obtained from the power market of the power system: financial data, marketing data, human resources data, market information, and production management information.

[0035] It should be understood that, considering that time series data usually organizes time series with temporal correlations into a sequence, the term "data" used in this application actually refers to the corresponding sequence.

[0036] Step 102: Add noise to the sampled data using the relevant Laplace mechanism to obtain noise-added data that meets the differential privacy protection requirements after noise addition;

[0037] Building upon step 101, this step aims to have the aforementioned executing entity add noise to the sampled dynamic time-series data using a relevant Laplace mechanism: ^

[0038] f(centroid(A)) = f(centroid(A)) + cLap(Δf / ε) to satisfy differential privacy protection.

[0039] In the above formula Let f(·) represent the query set after adding noise, f(·) represent the query function, centroid() represent the centroid, cLap() represent the relevance Laplace function, Δf represent the sensitivity, and ε represent the privacy budget.

[0040] Step 103: Apply VMD-Wiener filtering technology to stabilize and correct the data after noise addition to obtain the corrected data;

[0041] Based on step 102, this step aims to have the aforementioned execution entity optimize the number of VMD modes based on the cross-correlation between the residual sequence and other modes of the time series data, then use an improved VMD algorithm to stabilize the noisy dynamic time series, and finally use a Wiener filter mode sequence for correction.

[0042] Variational Mode Decomposition (VMD) is a signal decomposition technique that can decompose a non-stationary signal into multiple Intrinsic Mode Functions (IMFs) and a residual term. IMFs are local oscillation modes whose frequencies vary with time and whose frequency bandwidth is adaptive, effectively describing the time-frequency characteristics of non-stationary signals.

[0043] Wiener filtering is a signal processing technique used to remove noise components from signals to improve signal quality. Based on the statistical properties of signals and noise, Wiener filtering separates the signal and noise through a filter, thereby removing the noise component.

[0044] In other words, the VMD-Wiener filtering technique applies Wiener filtering to the IMFs and residual terms obtained from VMD decomposition to remove noise components from the signal, resulting in a clearer signal. This application uses the VMD-Wiener filtering technique to optimize a noisy dynamic time series.

[0045] Step 104: Combine the corrected data and non-sampling data corresponding to the same preset sampling interval to obtain the data to be published corresponding to each preset sampling interval.

[0046] Based on step 103, this step aims to combine the corrected data and non-sampling data corresponding to the same preset sampling interval by the aforementioned executing entity to obtain the data to be published corresponding to each preset sampling interval.

[0047] Step 105: Evaluate each data to be published through a dynamic time warping evaluation mechanism to determine the optimal sampling interval and the optimal data to be published corresponding to the optimal sampling interval;

[0048] Based on step 104, this step aims to have the aforementioned implementing entity evaluate the data at different sampling intervals through a dynamic time warping evaluation mechanism, compare the relationship between the original time series data and the data to be published, and determine the optimal sampling interval and the corresponding optimal data to be published.

[0049] Step 106: Publish the best data to be published.

[0050] Compared to existing technologies, the time-series data publishing method based on a sampling-filtered differential privacy protection model provided in this embodiment firstly uses fixed sampling techniques to add noise and filter the data at the sampling points, reducing the overall system runtime and improving system efficiency. Secondly, it utilizes a correlated Laplace mechanism to publish the relevant time-series data. Because this technique leverages the indistinguishability of sequences, it generates a correlated Laplace noise sequence by using four Gaussian white noise sequences interacting through a linear system, rather than generating independent noise sequences. This ensures high usability of the data through the concatenation of multiple different Gaussian white noises and avoids the problem of lower-than-expected privacy due to the independent and identically distributed characteristics of ordinary Laplace noise. Next, it employs VMD-Wiener filtering technology to stabilize and correct the noise-added data, making the published data closer to the original values. This improves the usability of the published data while maintaining basic data privacy. Finally, it applies the concept of dynamic time warping to differential privacy, using dynamic time warping to select the optimal sampling interval, obtaining sampling points that achieve a better balance between privacy and usability in time-series data publishing.

[0051] To further understand this plan, you may also refer to... Figure 2 The process diagram shown is as follows. Figure 2 This is an abstraction and visualization of the steps shown in the figure.

[0052] In some other embodiments of this application, the method for generating noise using the correlated Laplace mechanism can be as follows: leveraging the indistinguishability of sequences, a correlated Laplace noise sequence is generated using four Gaussian white noise sequences passing through a linear system, and then the noise is generated based on this correlated Laplace noise sequence. Specifically, this process can be implemented through the following steps:

[0053] Through formula Four independent and identically distributed Gaussian white noise sequences K1, K2, K3, and K4 are generated, with parameters determined by the sensitivity function and the protection strength.

[0054] In the above formula, K i Let represent a Gaussian white noise sequence, N be a Gaussian distribution, Δf represent sensitivity, and ε represent the privacy budget.

[0055] Let K1, K2, K3, and K4 be expressed using the impulse response function. The filter yields the relevant Gaussian sequence, generating new autocorrelation noise K1`, K2`, K3` and K4`;

[0056] In the above formula, R XX (τ) represents the autocorrelation function of the time series X, and N0 represents the power spectral density of the time series after adding noise.

[0057] Based on the principle of generating Laplace series from Gaussian series, correlated Laplace noise is generated: cLap=K1' 2 +K2' 2 -K3' 2 -K4' 2 .

[0058] The above process can be found in [reference]. Figure 3 The process diagram shown is shown below.

[0059] In some other embodiments of this application, the VMD-Wiener filtering technique is used to perform stabilization and correction processing on the data after noise addition, which can be achieved through the following steps:

[0060] The formula for calculating the cross-correlation coefficient is used. Determine the optimal number of modes for VMD decomposition; where, N is the number of sequence modes, m is the time interval, and R cc (0) and R ff (0) represents the autocorrelation function of the two modal sequences, respectively;

[0061] VMD is used to decompose the data after noise is added, and to obtain a stable sequence that eliminates high-frequency residuals.

[0062] Formula for Wiener filtering Each optimal mode sequence is filtered separately; where δ 2 Let φ(z) be the autocorrelation sequence value of white noise, φ(z) be the Z-transform of the autocorrelation sequence of the time series, and B(z) be the Z-domain solution of Wiener filtering.

[0063] The optimal mode sequences after filtering are reconstructed to obtain the corrected data.

[0064] The above process can be found in [reference]. Figure 4 The process diagram shown is shown below.

[0065] In some other embodiments of this application, the process of evaluating each of the data to be published through a dynamic time warping evaluation mechanism to determine the optimal sampling interval and the optimal data to be published corresponding to the optimal sampling interval can be specifically described in the following steps:

[0066] Define the initial number of samples n, define different sampling intervals, and perform n samplings at a fixed sampling rate at different time intervals.

[0067] After adding noise to the sampling sequences at different time intervals, the mean sequence of the sampling sequences after n noise additions is obtained using the average value method.

[0068] Calculate the dynamic time warp value between the n sampling sequences with different time intervals after adding noise and the mean sequence, and select the sampling sequence and its sampling point that are closest to the mean sequence as the sampling point with better privacy and practicality.

[0069] The calculation method for the dynamic time warp value is as follows:

[0070] Through formula Perform calculations;

[0071] Where Q and C are the two time series of lengths m and n to be calculated, respectively, w k The optimal path is defined as W = w1, w2, ..., w k ,…,w K Where max(m,n)≤K<m+n-1; the basic conditions that the optimal path of dynamic time warping must satisfy include: boundary conditions, that is, the selected path must start from the lower left corner of the distance matrix and end at the upper right corner of the matrix; continuity, that is, a point on the path cannot cross a certain point to match, ensuring that every point in the two time series must appear in the distance matrix; monotonicity, that is, the points in the matrix must be monotonically increasing, ensuring that there is no intersection in the path; the distance matrix of the dynamic time warping algorithm is:

[0072] c1 c2 … c n

[0073]

[0074] Where, the time series Q = {q1,q2,…,q} m}, C = {c1, c2, ..., c n The path with the minimum curvature cost after adhering to all conditions is the required dynamic time curvature value.

[0075] Due to the complexity of the situation, it is impossible to list and elaborate on them all. Those skilled in the art should realize that there are many examples based on the basic method principles provided in this application and in combination with actual situations. Without sufficient creative effort, they should all be within the protection scope of this application.

[0076] Please see below. Figure 5 , Figure 5This is a structural block diagram of a time-series data publishing device 500 based on a sampling-filtered differential privacy protection model, provided as an embodiment of this application. This embodiment exists as a device embodiment corresponding to the above-described method embodiment. The time-series data publishing device 500 based on a sampling-filtered differential privacy protection model may include:

[0077] The custom interval sampling unit 501 is configured to sample the original time series data according to multiple different preset sampling intervals to obtain sampled data and non-sampled data.

[0078] The differential privacy protection processing unit 502 is configured to add noise to the sampled data using a correlated Laplace mechanism to obtain noise-added data that meets the differential privacy protection requirements after noise addition.

[0079] The stabilization and re-correction unit 503 is configured to use VMD-Wiener filtering technology to perform stabilization and re-correction processing on the noise-added data to obtain corrected data.

[0080] The data combination unit 504 is configured to combine the corrected data and non-sampled data corresponding to the same preset sampling interval to obtain the data to be published corresponding to each preset sampling interval.

[0081] Evaluation unit 505 is configured to evaluate each of the data to be published through a dynamic time warping evaluation mechanism to determine the optimal sampling interval and the optimal data to be published corresponding to the optimal sampling interval.

[0082] The publishing unit 506 is configured to publish the optimal data to be published.

[0083] In some other implementations of this embodiment, the time-series data publishing device 500 based on the sampling filter differential privacy protection model may further include:

[0084] The raw time-series data acquisition unit is configured to acquire the raw time-series data, which includes at least one of financial data, marketing data, human resources data, market information, and production management information, from the power market of the power system.

[0085] In some other implementations of this embodiment, the method for generating noise using the correlated Laplace mechanism includes:

[0086] Taking advantage of the indistinguishability of sequences, a correlated Laplace noise sequence is generated using four Gaussian white noise sequences passing through a linear system.

[0087] The noise is generated based on the relevant Laplace noise sequence.

[0088] In some other implementations of this embodiment, the differential privacy protection processing unit 502 is further configured to:

[0089] Through formula Four independent and identically distributed Gaussian white noise sequences K1, K2, K3, and K4 are generated, with parameters determined by the sensitivity function and the protection strength.

[0090] Let K1, K2, K3, and K4 be expressed using the impulse response function. The filter yields the relevant Gaussian sequence, generating new autocorrelation noise K1`, K2`, K3` and K4`;

[0091] Correspondingly, generating the noise based on the relevant Laplace noise sequence includes:

[0092] Based on the principle of generating Laplace series from Gaussian series, correlated Laplace noise is generated: cLap=K1' 2 +K2' 2 -K3' 2 -K4' 2 .

[0093] In some other implementations of this embodiment, the stabilization and re-correction unit 503 is further configured to:

[0094] The formula for calculating the cross-correlation coefficient is used. Determine the optimal number of modes for VMD decomposition; where, N is the number of sequence modes, m is the time interval, and R cc (0) and R ff (0) represents the autocorrelation function of the two modal sequences, respectively;

[0095] VMD is used to decompose the noise-added data to obtain a stable and high-frequency residual sequence;

[0096] Formula for Wiener filtering Each optimal mode sequence is filtered separately; where δ 2 Let φ(z) be the autocorrelation sequence value of white noise, φ(z) be the Z-transform of the autocorrelation sequence of the time series, and B(z) be the Z-domain solution of Wiener filtering.

[0097] The filtered optimal mode sequences are reconstructed to obtain the corrected data.

[0098] In some other implementations of this embodiment, the evaluation unit 505 may be further configured as follows:

[0099] Based on the data to be published, the mean data is obtained using the average value method;

[0100] Calculate the dynamic time warp value between each of the data to be published and the mean data, and select the data to be published that is closest to the mean data as the optimal data to be published, and determine the sampling interval of the optimal data to be published as the optimal sampling interval.

[0101] The calculation method for the dynamic time warp value is as follows:

[0102] Through formula Perform calculations;

[0103] Where Q and C are the two time series of lengths m and n to be calculated, respectively, w k The optimal path is defined as W = w1, w2, ..., w k ,…,w K Where max(m,n)≤K<m+n-1; the basic conditions that the optimal path of dynamic time warping must satisfy include: boundary conditions, that is, the selected path must start from the lower left corner of the distance matrix and end at the upper right corner of the matrix; continuity, that is, a point on the path cannot cross a certain point to match, ensuring that every point in the two time series must appear in the distance matrix; monotonicity, that is, the points in the matrix must be monotonically increasing, ensuring that there is no intersection in the path; the distance matrix of the dynamic time warping algorithm is:

[0104] c1 c2 … c n

[0105]

[0106] Where, the time series Q = {q1,q2,…,q} m}, C = {c1, c2, ..., c n The path with the minimum curvature cost after adhering to all conditions is the required dynamic time curvature value.

[0107] This embodiment exists as a device embodiment corresponding to the above method embodiment. Compared to existing technologies, the time-series data publishing device based on a sampling-filtered differential privacy protection model provided in this embodiment firstly uses fixed sampling technology to add noise and filter the data at the sampling points, reducing the overall system runtime and improving system efficiency. Secondly, it uses a correlated Laplace mechanism to publish the relevant time-series data. Because this technology utilizes the indistinguishability of sequences, it generates a correlated Laplace noise sequence by using four Gaussian white noise sequences interacting through a linear system, rather than generating independent noise sequences. This ensures high usability of the data through the concatenation of multiple different Gaussian white noises and avoids the problem of lower-than-expected privacy due to the independent and identically distributed characteristics of ordinary Laplace noise. Next, it uses VMD-Wiener filtering technology to stabilize and correct the noisy data, making the published data closer to the original value. Under the premise of ensuring that the data privacy remains basically unchanged, the usability of the published data is improved. Finally, the concept of dynamic time warping is applied to differential privacy. Dynamic time warping is used to select the optimal sampling interval to obtain sampling points that achieve a better balance between the privacy and usability of time-series data publishing.

[0108] Based on the above embodiments, this application also provides an electronic device, which may include a memory and a processor. The memory stores a computer program, and when the processor calls the computer program in the memory, it can implement the steps provided in the above embodiments. Of course, the electronic device may also include various necessary network interfaces, a power supply, and other components.

[0109] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by an execution terminal or processor, can perform the steps provided in the above embodiments. The storage medium may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0110] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0111] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0112] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this application. For those skilled in the art, various improvements and modifications can be made to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

[0113] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

Claims

1. A method for publishing time-series data based on a sampling filtering differential privacy protection model, characterized in that, include: The original time series data is sampled at multiple different preset sampling intervals to obtain sampled data and non-sampled data. The sampling data is noise-added using a correlated Laplace mechanism to obtain noise-added data that meets the differential privacy protection requirements after noise addition. The noise-added data is stabilized and then corrected using VMD-Wiener filtering technology to obtain the corrected data. The corrected data and non-sampled data corresponding to the same preset sampling interval are combined to obtain the data to be published corresponding to each preset sampling interval. The dynamic time warping evaluation mechanism is used to evaluate each of the data to be published, and to determine the optimal sampling interval and the optimal data to be published corresponding to the optimal sampling interval. Publish the optimal data to be published; The step of applying VMD-Wiener filtering technology to stabilize and correct the noise-added data to obtain corrected data includes: The formula for calculating the cross-correlation coefficient is used. Determine the optimal number of modes for VMD decomposition; where, This represents the cross-relationship value at time interval m. , Two modal sequences and The cross-correlation function value at time interval m, where n is the sequence index and N is the number of sequence modes. and These are two different modal sequences. and These are the autocorrelation functions of the two modal sequences, respectively; VMD is used to decompose the noise-added data to obtain a stable and high-frequency residual sequence; Formula for Wiener filtering Each optimal mode sequence is filtered separately; among them... Let be the transfer function of the optimal Wiener filter. The autocorrelation sequence value is white noise. The Z-transform of the autocorrelation sequence of the time series. Solving for the Wiener filter in the Z-domain. This is the inverse Z-transform form of Wiener filtering; The filtered optimal mode sequences are reconstructed to obtain the corrected data.

2. The method according to claim 1, characterized in that, Also includes: Obtain the raw time-series data from the electricity market of the power system, including at least one of financial data, marketing data, human resources data, market information, and production management information.

3. The method according to claim 1, characterized in that, The method for generating noise using the aforementioned correlated Laplace mechanism includes: Taking advantage of the indistinguishability of sequences, a correlated Laplace noise sequence is generated using four Gaussian white noise sequences passing through a linear system. The noise is generated based on the relevant Laplace noise sequence.

4. The method according to claim 3, characterized in that, The method utilizes the indistinguishability of sequences, using four Gaussian white noise sequences passing through a linear system to generate a correlated Laplace noise sequence, including: Through formula Four independent and identically distributed Gaussian white noise sequences, K1, K2, K3, and K4, are generated, with parameters determined by the sensitivity function and the protection strength; among them, It is a Gaussian white noise sequence. It follows a Gaussian distribution. For sensitivity, Budget for privacy; Let K1, K2, K3, and K4 be expressed using the impulse response function. The filter yields a relevant Gaussian sequence, generating new autocorrelation noise K1', K2', K3', and K4'; where, Let X be the autocorrelation function of the time series X. This is the impulse response of the filter. The power spectral density of the time series after adding noise; Correspondingly, generating the noise based on the relevant Laplace noise sequence includes: Based on the principle of generating Laplace series from Gaussian series, relevant Laplace noise is generated: .

5. The method according to claim 1, characterized in that, The evaluation of each piece of data to be published through a dynamic time warping evaluation mechanism to determine the optimal sampling interval and the optimal data to be published corresponding to the optimal sampling interval includes: Based on the data to be published, the mean data is obtained using the average value method; Calculate the dynamic time warp value between each of the data to be published and the mean data, and select the data to be published that is closest to the mean data as the optimal data to be published, and determine the sampling interval of the optimal data to be published as the optimal sampling interval. The calculation method for the dynamic time warp value is as follows: Through formula Perform calculations; in, Given two time series of lengths m and n, respectively, to be calculated. The optimal path, i.e. ,in The basic conditions that the optimal path of the dynamic time warping algorithm must satisfy include: boundary conditions, i.e., the selected path must start from the lower left corner of the distance matrix and end at the upper right corner; continuity, i.e., points on the path cannot cross any point to match, ensuring that every point in both time series must appear in the distance matrix; and monotonicity, i.e., the points in the matrix must be monotonically increasing, ensuring that there are no intersections in the path. The distance matrix of the dynamic time warping algorithm is: Among them, time series , The path with the minimum curvature cost after adhering to all conditions is the required dynamic time curvature value.

6. A time-series data publishing device based on a sampling filtering differential privacy protection model, characterized in that, include: A custom interval sampling unit is configured to sample the original time series data at multiple different preset sampling intervals to obtain sampled data and non-sampled data. The differential privacy protection processing unit is configured to add noise to the sampled data using a correlated Laplace mechanism to obtain noise-added data that meets the differential privacy protection requirements after noise addition. The stabilization and re-correction unit is configured to use VMD-Wiener filtering technology to perform stabilization and re-correction processing on the noise-added data to obtain corrected data. The data combination unit to be published is configured to combine the corrected data and non-sampled data corresponding to the same preset sampling interval to obtain the data to be published corresponding to each of the preset sampling intervals. The evaluation unit is configured to evaluate each of the data to be published through a dynamic time warping evaluation mechanism to determine the optimal sampling interval and the optimal data to be published corresponding to the optimal sampling interval. The publishing unit is configured to publish the optimal data to be published; The stabilization and re-correction unit is further configured as follows: The formula for calculating the cross-correlation coefficient is used. Determine the optimal number of modes for VMD decomposition; where, This represents the cross-relationship value at time interval m. , Two modal sequences and The cross-correlation function value at time interval m, where n is the sequence index and N is the number of sequence modes. and These are two different modal sequences. and These are the autocorrelation functions of the two modal sequences, respectively; VMD is used to decompose the noise-added data to obtain a stable and high-frequency residual sequence; Formula for Wiener filtering Each optimal mode sequence is filtered separately; among them... Let be the transfer function of the optimal Wiener filter. The autocorrelation sequence value is white noise. The Z-transform of the autocorrelation sequence of the time series. Solving for the Wiener filter in the Z-domain. This is the inverse Z-transform form of Wiener filtering; The filtered optimal mode sequences are reconstructed to obtain the corrected data.

7. An electronic device, characterized in that, include: Memory, used for computer programs; A processor configured to implement, when executing a computer program stored in the memory, the steps of the time-series data publishing method based on a sampling-filtered differential privacy protection model as described in any one of claims 1 to 5.

8. A readable storage medium, characterized in that, The readable storage medium stores a computer program, which, when executed by a processor, can implement the steps of the time-series data publishing method based on the sampling filter differential privacy protection model as described in any one of claims 1 to 5.

Images