Power data desensitization method based on multi-granularity dynamic sensitivity grading, terminal equipment and storage medium

Abstract

The invention discloses a power data desensitization method based on multi-granularity dynamic sensitivity grading, terminal equipment and a storage medium. The power data desensitization method comprises the following steps: dividing power data into four sensitive grades of identity information, transaction records, power consumption behaviors and equipment data; full life cycle protection of data acquisition, storage and transmission is realized through a collaborative desensitization mechanism of field data national secret SM4 encryption / dynamic mask and recorded data Laplacian noise injection / behavior mode generalization; according to the system, Haar wavelet decomposition is adopted to separate user power consumption time sequence data into a low-frequency approximation coefficient and a high-frequency detail coefficient, Paillier homomorphic encryption aggregation is carried out on a low-frequency component, Gaussian noise is added to a high-frequency component, then equalization processing is carried out, and an aggregation power consumption sequence is reconstructed through wavelet inverse transformation.

Description

Technical Field

[0001] This invention relates to the fields of data privacy protection and power data analysis technology, and in particular to a power data desensitization method, terminal equipment and storage medium based on multi-granularity dynamic sensitivity classification. Background Technology

[0002] Traditional sensitivity grading employs fixed strategies, failing to dynamically adjust the desensitization strength based on data usage scenarios, role permissions, and compliance requirements. This leads to compromised data usability in high-security scenarios or over-desensitization in low-risk scenarios. Existing desensitization technologies are mostly limited to simple field-level masking or full-field encryption, lacking multi-granular control over record-level behavioral patterns and dataset-level statistical characteristics, making them vulnerable to privacy-inferring attacks based on data correlation analysis. Data aggregation often employs plaintext transmission or a single encryption mechanism, posing a risk of leaking individual electricity consumption patterns. Furthermore, traditional homomorphic encryption's high computational complexity makes it unsuitable for real-time analysis. These technical deficiencies severely restrict the secure utilization of power data, urgently requiring breakthroughs through dynamic grading, multi-granular protection, and intelligent detection technologies. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a power data anonymization method, terminal equipment and storage medium based on multi-granularity dynamic sensitivity classification, which addresses the shortcomings of the existing technology and achieves a strong level of privacy protection for anonymity while maintaining data availability.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a power data desensitization method based on multi-granularity dynamic sensitivity classification, comprising the following steps:

[0005] S1. Obtain user data provided by the power company to a third party, and classify the user data into dynamic sensitivity levels; the user data includes identity information data, electricity consumption behavior data, transaction record data, and equipment data.

[0006] S2. Based on the dynamic sensitivity level, perform multi-level collaborative desensitization on the user data to obtain desensitized user data;

[0007] S3. Perform full lifecycle protection on the desensitized user data to obtain protected electricity consumption behavior data;

[0008] S4. Perform privacy-preserving aggregation preprocessing on the protected electricity consumption behavior data to obtain encrypted data.

[0009] In step S1, the dynamic sensitivity levels include high sensitivity, medium sensitivity, medium-high sensitivity, and medium-low sensitivity; user identity information data is high sensitivity, electricity consumption behavior data is medium sensitivity, transaction record data is medium-high sensitivity, and device data is medium-low sensitivity.

[0010] In step S2, the specific implementation process of multi-level collaborative de-identification of the user data includes:

[0011] Highly sensitive fields are encrypted using the national standard SM4 encryption; moderately sensitive fields are masked using dynamic masking.

[0012] Multi-granularity anonymization is performed on electricity consumption behavior data and transaction record data. Laplace noise is injected into statistical queries, with the noise level controlled by a privacy budget ε; where ε = K / S -final K is the system-level privacy protection strength constant; S _final =α·S _base +β·R _role +γ·C _scenario S _final For dynamic sensitivity levels, S _base Based on the basic sensitivity level, R _role C represents the role's permission coefficient. _scenario α represents the scenario risk coefficient, and β and γ are configurable weights.

[0013] The specific implementation process of step S3 includes:

[0014] During the data collection phase, user identity information data IDs are removed in real time, and anonymous identifiers are generated.

[0015] During the data storage phase, highly sensitive data is encrypted using AES-256 double encryption, while medium- and low-sensitivity data is encrypted using transparent data encryption.

[0016] During the data transmission phase, the TLS 1.3 algorithm is used to encrypt the transmitted data.

[0017] For electricity consumption data within the electricity consumption behavior category, the following steps are also performed:

[0018] Electricity consumption data is collected at set intervals to obtain an electricity consumption sequence;

[0019] Perform wavelet transform on the electricity consumption sequence to obtain low-frequency approximation coefficients cA and high-frequency detail coefficients cD;

[0020] Using the low-frequency approximation coefficient cA, the Paillier homomorphic encryption algorithm is used to encrypt the ciphertext E(cA); a Gaussian distribution N(0, σ) is added to the high-frequency coefficient cD. 2 (noise).

[0021] The condition satisfies a Gaussian distribution N(0,σ). 2 The standard deviation σ of the noise is determined by the privacy budget ∈ and the query sensitivity Δ: Where σ is the standard deviation of the noise, and the noise follows a Gaussian distribution N(0,σ). 2 ), where ∈ represents the privacy budget, δ represents the differential privacy parameter, and Δ represents the query sensitivity.

[0022] Also includes:

[0023] The electricity consumption sequence can be recovered using the following formula: Where x j It is the j-th electricity consumption sequence after reconstruction. cA avg =Decrypt(C agg ) / N, C agg It is aggregated ciphertext, CD i The high-frequency detail coefficients are obtained by performing a wavelet transform on the i-th electricity consumption sequence. i Let N represent the i-th noise element that follows a Gaussian distribution, and let N represent the length of the electricity consumption sequence. μ is the modular inverse, satisfying μ=(L(g λ mod n 2 )) -1 mod n, where p and q are two distinct prime numbers, and n = p * q.

[0024] As an inventive concept, the present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory; the processor executes the computer program to implement the steps of the above method.

[0025] As an inventive concept, the present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon; when the computer program / instructions are executed by a processor, they implement the steps of the above-described method.

[0026] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention employs a multi-granularity dynamic desensitization mechanism and frequency domain separation for privacy protection, achieving anonymity protection; it uses hierarchical desensitization and wavelet aggregation to maintain data availability. Compared with traditional plaintext aggregation or fully encrypted schemes, this invention achieves the privacy protection goal of "data available but not visible" while ensuring data availability, and supports real-time aggregation response. Attached Figure Description

[0027] Figure 1 This is a dynamic desensitization process throughout the entire lifecycle of an embodiment of the present invention;

[0028] Figure 2 This is the wavelet-homogeneous encryption aggregation system architecture of an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Example 1

[0031] The implementation process of this invention follows a technical path of "data classification → dynamic desensitization → full-cycle protection → encryption preprocessing → cloud aggregation". The specific implementation steps are as follows:

[0032] Step S1: First, obtain user data provided by the power company to third parties, covering multiple types such as identity information, electricity consumption behavior, transaction records, equipment data, and electricity consumption, and build a scenario-driven dynamic sensitivity classification model.

[0033] Step S2: Using the dynamic sensitivity level established in Step S1 as a benchmark, implement a multi-level collaborative desensitization strategy. Output the desensitized power data as the input for the full lifecycle protection in Step S3.

[0034] Step S3: To ensure the security of user data from source to end, implement full lifecycle protection for the de-identified data output in step S2, covering every link from data collection, transmission, storage to final secure destruction, and strictly prevent data leakage at any stage.

[0035] Step S4: Receive the electricity consumption behavior data transmitted and protected in Step S3, perform privacy-preserving aggregation preprocessing, and perform Haar wavelet transform to decompose the data into low-frequency coefficients reflecting the overall trend and high-frequency coefficients reflecting detailed fluctuations. Encrypt the low-frequency coefficients and add Gaussian-distributed random noise to the high-frequency coefficients.

[0036] Step S5: Upload the data processed and encrypted in step S4 (i.e., encrypted low-frequency coefficients and noisy high-frequency coefficients) to a cloud environment with a high level of security protection, and realize privacy protection aggregation and data restoration in the secure cloud environment.

[0037] The data anonymization method and privacy-preserving power data aggregation system provided by this invention fully consider the difficulties and practical scenarios in various environments regarding data anonymization and data privacy protection. It solves the problem by innovatively integrating three major technical modules: dynamic sensitivity grading, multi-granularity anonymization, and efficient privacy aggregation.

[0038] Furthermore, step S1 includes:

[0039] Step S1-1: Divide the power user data into four sensitivity levels: identity information (high sensitivity), electricity consumption behavior (medium sensitivity), transaction records (medium-high sensitivity), and equipment data (medium-low sensitivity), as shown in Table 1.

[0040]

[0041]

[0042] Step S1-2: Perform dynamic sensitivity grading configuration, which includes scenario weight factor configuration, role permission mapping, and regulatory compliance engine. Sensitivity dynamic quantification formula:

[0043] S _final =α·S _base +β·R _role +γ·C _scenario

[0044] Where S _base Based on the basic sensitivity level, R _role C represents the role's permission coefficient. _scenario α represents the scenario risk coefficient, and β and γ are configurable weights.

[0045] The scenario weighting factor configuration dynamically adjusts the desensitization strength based on the data usage scenario. The sensitivity coefficient for third-party shared scenarios is set to 1.2, and for internal operation and maintenance scenarios it is set to 1.0. Role permission mapping is divided into third-party analyst permissions and internal administrator permissions. Third-party analyst permissions use full masking of ID card numbers (e.g., "110101****1234"), and the address is displayed down to the district level (e.g., "Beijing City Road"). Internal administrator permissions use a mobile phone number mask of "138****5678", and the address can be viewed with complete information. The regulatory compliance engine integrates GDPR, Personal Information Protection Law, and other regulatory requirements, and automatically adjusts the desensitization strategy. It automatically detects whether the mobile phone number mask meets the format of "first 3 digits + 4 masked digits + last 4 digits". If it does not meet the format, it triggers a policy update.

[0046] Step S1 constructs a multi-dimensional sensitivity grading model to achieve refined classification and dynamic sensitivity labeling of power data. Compared to traditional static grading methods, this solution introduces scenario weighting factors, role-based access control, and a regulatory compliance engine, enabling dynamic adjustment of data sensitivity levels based on actual usage scenarios, differences in user roles and permissions, and real-time regulatory requirements. This mechanism effectively solves the problems of "over-sensitization" or "insufficient protection" caused by fixed grading strategies, ensuring maximum data availability while meeting privacy compliance requirements. For example, in third-party sharing scenarios, by increasing the sensitivity coefficient to 1.2, the desensitization strength of identity information is enhanced, avoiding the risk of sensitive information leakage.

[0047] Furthermore, step S2 includes:

[0048] Step S2-1: Perform multi-granularity desensitization on the field data processed in Step 1. Highly sensitive fields (ID card number, mobile phone number) are encrypted using the national cryptographic standard SM4, and the key is stored in fragments through the hardware security module (HSM). Mediumly sensitive fields (user address) are masked using dynamic masking rules. For example, the address field is masked in the format of "province + city + district + ** road + number".

[0049] Step S2-2: Perform multi-granularity desensitization on the recorded data (electricity consumption, electricity timestamp) processed in Step 1. Inject Laplace noise into the statistical query; the noise level is controlled by the privacy budget ε, satisfying (ε,δ)-differential privacy. Generalize the electricity consumption period from precise timestamps to "peak / valley / flat periods" to obscure user behavior patterns. Different privacy budgets ε should correspond to different sensitivity levels for electricity consumption behavior data and transaction record data. Implement differentiated desensitization, and determine the basic privacy budget ε_base based on the sensitivity level.

[0050] ε -base =K / S -final

[0051] K is a system-level privacy protection strength constant.

[0052] Step S2 employs a collaborative desensitization technique for field data and record data, forming a multi-granularity privacy protection system. Highly sensitive fields are encrypted using the national cryptographic standard SM4, combined with the key sharding storage mechanism of the hardware security module to ensure key security. Moderately sensitive fields are subject to dynamic masking rules, with the masking granularity dynamically adjusted based on role permissions. In record data desensitization, differential privacy protection is achieved through Laplace noise injection, and electricity usage periods are generalized to "peak / valley / flat periods," preventing attackers from inferring user behavior patterns through timestamp associations. Compared to traditional single-field desensitization, this invention significantly enhances defense against association inference attacks through a multi-layered technology combination, while preserving the data value of statistical queries.

[0053] Furthermore, step S3 includes:

[0054] Step S3-1: During the data acquisition phase, a lightweight desensitization module is embedded in the smart meter terminal to remove the user ID in real time and generate an anonymous identifier, thus fulfilling the identity information desensitization requirement of S2.

[0055] Step 3-2: During the data storage phase, highly sensitive data is encrypted twice using AES-256, and the key is stored in fragments on a remote key management server (KMS); medium and low sensitive data is encrypted using transparent data encryption (TDE), and the key is bound to the database instance.

[0056] Step S3-3: During the data transmission phase, the TLS1.3 algorithm is used to encrypt the transmitted data, and a dynamic de-identification middleware is embedded in the API gateway layer. The mobile phone number in the response data is masked in real time according to the "first 3 digits + **** + last 4 digits" to protect the security of the S2 de-identified data during transmission.

[0057] Step S3 covers a dynamic desensitization mechanism throughout the entire data lifecycle, from acquisition and storage to transmission. During the acquisition phase, a lightweight desensitization module is embedded in the smart meter terminal to generate anonymous device fingerprints in real time, preventing the leakage of the original user ID. In the storage phase, a layered encryption strategy is employed: highly sensitive data is encrypted using AES-256 and the key is stored in segments; medium- and low-sensitivity data is encrypted using transparent data encryption to reduce computational overhead. In the transmission phase, TLS 1.3 encryption and dynamic masking via the API gateway are combined to ensure real-time masking of sensitive fields during transmission. This end-to-end protection mechanism solves the problem of fragmented desensitization processes in traditional solutions. Through end-to-end dynamic adaptation, it effectively defends against data leakage risks at each stage of the lifecycle, while simultaneously considering system performance and security requirements.

[0058] Furthermore, step S4 includes:

[0059] Step S4-1: Collect residents' electricity consumption data every 15 minutes to generate a 24-hour sequence.

[0060] Step S4-2: Perform a single Haar wavelet transform on the electricity consumption sequence generated in step S4-1, separating it into low-frequency approximation coefficients cA and high-frequency detail coefficients cD. For an electricity consumption sequence x = [x1, x2, ..., x...] of length N (even number), ... N The formula for a single-layer Haar wavelet transform is:

[0061] Low-frequency approximation coefficient (cA):

[0062]

[0063] High-frequency detail factor (cD):

[0064]

[0065] Step S4-3: Based on the low-frequency coefficient cA obtained in step S4-2, encrypt using the Paillier homomorphic encryption algorithm to generate ciphertext E(cA). The Paillier public-key encryption function is:

[0066] E(m)=g m ·r n modn 2

[0067] Where g is the generator, r is the random number, n is the modulus, and E(m) represents the encryption result of the plaintext m. The encrypted data is then uploaded to the cloud.

[0068] Step S4-4: Add Gaussian distribution N(0,σ) to the high-frequency coefficients cD obtained in step S4-2. 2 The noise level, σ, is determined by the privacy budget ∈ and the query sensitivity Δ.

[0069]

[0070] Where σ is the standard deviation of the noise, and the noise follows a Gaussian distribution N(0,σ). 2 ), where ∈ represents the privacy budget (a privacy parameter; the smaller the value, the stronger the privacy protection), δ represents the differential privacy parameter, and Δ represents the query sensitivity.

[0071] Step S4 decomposes the electricity consumption time-series data into low-frequency approximation coefficients and high-frequency detail coefficients using Haar wavelet transform, and then employs differentiated encryption and noise addition strategies respectively. Paillier homomorphic encryption is used for the low-frequency components, supporting cloud-based encrypted domain aggregation operations; Gaussian noise is superimposed on the high-frequency components to obfuscate individual electricity consumption characteristics. Compared to traditional full-data encryption aggregation schemes, this method utilizes the frequency domain separation characteristics of wavelet transform to significantly reduce the computational load of homomorphic encryption, while simultaneously protecting individual behavioral patterns in high-frequency details through noise injection, achieving a balance between aggregation efficiency and privacy protection.

[0072] Furthermore, step S5 includes:

[0073] Step S5-1: For the low-frequency component after wavelet decomposition in step S4-2, and the data that has already been encrypted, perform homomorphic addition on the user's encrypted low-frequency coefficients in the cloud:

[0074]

[0075] The formula for decrypting the aggregation result is:

[0076]

[0077] Where L(x) = (x-1) / n, λ = lcm(p-1, q-1), and μ is the modular inverse, satisfying μ = (L(g) / n) λ mod n 2 )) - 1 modn.

[0078] Among them, E(cA) i ) is the encryption low-frequency coefficient of the i-th user (from step S4-3), C agg It is an aggregated ciphertext; the aggregated result is obtained by decrypting it with the private key. The low-frequency mean is cA avg =Decrypt(C agg ) / N.

[0079] For the noisy high-frequency coefficients after wavelet decomposition in step S4-2, the cloud performs an average operation on this part of the data from all users.

[0080] Step S5-2: Reconstruct the data from Step S5-1 by recovering the aggregated power consumption sequence using Haar wavelet inverse transform.

[0081]

[0082] Where cA j These are the low-frequency approximation coefficients (from step S4-2). cD j These are high-frequency detail coefficients (from step S4-3). x j It is the original power consumption sequence after reconstruction.

[0083] Step S5 performs homomorphic addition and aggregation of noisy high-frequency components in the cloud, and reconstructs the final result through inverse wavelet transform. Paillier homomorphic encryption allows direct addition operations on ciphertext, ensuring that the original data does not need to be decrypted during the aggregation process, avoiding privacy leaks in the cloud or on the transmission path; noise averaging of high-frequency components further eliminates residual individual characteristics, meeting the accuracy requirements of power companies for total electricity consumption statistics. Compared with traditional plaintext aggregation or fully encrypted schemes, this scheme achieves the privacy protection goal of "data available but not visible" while ensuring data availability, and supports real-time aggregation response.

[0084] Example 2

[0085] Embodiment 2 of the present invention provides a terminal device corresponding to Embodiment 1 above. The terminal device can be a processing device for a client, such as a mobile phone, a laptop, a tablet computer, a desktop computer, etc., to execute the method of the above embodiments.

[0086] The terminal device in this embodiment includes a memory, a processor, and a computer program stored in the memory; the processor executes the computer program in the memory to implement the steps of the method in Embodiment 1 described above.

[0087] In some implementations, the memory may be high-speed random access memory (RAM), and may also include non-volatile memory, such as at least one disk storage device.

[0088] In other implementations, the processor can be any type of general-purpose processor, such as a central processing unit (CPU) or a digital signal processor (DSP), and there is no limitation here.

[0089] Example 3

[0090] Embodiment 3 of the present invention provides a computer-readable storage medium corresponding to Embodiment 1 above, on which a computer program / instructions are stored. When the computer program / instructions are executed by a processor, they implement the steps of the method of Embodiment 1 above.

[0091] A computer-readable storage medium can be a tangible device that holds and stores instructions for use by an instruction execution device. A computer-readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof.

[0092] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0093] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0094] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the function specified in one or more boxes.

[0095] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0096] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for desensitizing power data based on multi-granularity dynamic sensitivity classification, characterized in that, Includes the following steps: S1. Obtain user data provided by the power company to a third party, and classify the user data into dynamic sensitivity levels; the user data includes identity information data, electricity consumption behavior data, transaction record data, and equipment data. S2. Based on the dynamic sensitivity level, perform multi-level collaborative desensitization on the user data to obtain desensitized user data; S3. Perform full lifecycle protection on the desensitized user data to obtain protected electricity consumption behavior data; S4. Perform privacy-preserving aggregation preprocessing on the protected electricity consumption behavior data to obtain encrypted data.

2. The power data desensitization method based on multi-granularity dynamic sensitivity classification according to claim 1, characterized in that, In step S1, the dynamic sensitivity levels include high sensitivity, medium sensitivity, medium-high sensitivity, and medium-low sensitivity; user identity information data is high sensitivity, electricity consumption behavior data is medium sensitivity, transaction record data is medium-high sensitivity, and device data is medium-low sensitivity.

3. The power data desensitization method based on multi-granularity dynamic sensitivity classification according to claim 2, characterized in that, In step S2, the specific implementation process of multi-level collaborative de-identification of the user data includes: Highly sensitive fields are encrypted using the national standard SM4 encryption; moderately sensitive fields are masked using dynamic masking. Multi-granularity anonymization is performed on electricity consumption behavior data and transaction record data. Laplace noise is injected into statistical queries, with the noise level controlled by a privacy budget ε; where ε = K / S -final K is the system-level privacy protection strength constant; S _final =α·S _base +β·R _role +γ·C _scenario S _final For dynamic sensitivity levels, S _base Based on the basic sensitivity level, R _role C represents the role's permission coefficient. _scenario α represents the scenario risk coefficient, and β and γ are configurable weights.

4. The power data desensitization method based on multi-granularity dynamic sensitivity classification according to claim 2, characterized in that, The specific implementation process of step S3 includes: During the data acquisition phase, user identity information data IDs are removed in real time, and anonymous identifiers are generated; during the data storage phase, highly sensitive data is encrypted using AES-256 double encryption, and medium-to-low sensitive data is encrypted using transparent data encryption. During the data transmission phase, the TLS 1.3 algorithm is used to encrypt the transmitted data.

5. The power data desensitization method based on multi-granularity dynamic sensitivity classification according to claim 2, characterized in that, For electricity consumption data within the electricity consumption behavior category, the following steps are also performed: Electricity consumption data is collected at set intervals to obtain an electricity consumption sequence; Perform wavelet transform on the electricity consumption sequence to obtain low-frequency approximation coefficients cA and high-frequency detail coefficients cD; Using the low-frequency approximation coefficient cA, the Paillier homomorphic encryption algorithm is used to encrypt the ciphertext E(cA); a Gaussian distribution N(0, σ) is added to the high-frequency coefficient cD. 2 The noise is σ, where σ is the standard deviation of the noise.

6. The power data desensitization method based on multi-granularity dynamic sensitivity classification according to claim 5, characterized in that, The condition satisfies a Gaussian distribution N(0,σ). 2 The standard deviation σ of the noise is determined by the privacy budget ∈ and the query sensitivity Δ: Where ∈ represents the privacy budget, δ represents the differential privacy parameter, and Δ represents the query sensitivity.

7. The power data desensitization method based on multi-granularity dynamic sensitivity classification according to claim 5, characterized in that, Also includes: The electricity consumption sequence can be recovered using the following formula: Where x j It is the j-th electricity consumption sequence after reconstruction. cA avg =Decrypt(C agg ) / N, C agg It is a gathering Combined ciphertext, cD i The high-frequency detail coefficients are obtained by performing a wavelet transform on the i-th electricity consumption sequence. i Let N represent the i-th noise element that follows a Gaussian distribution, and let N represent the length of the electricity consumption sequence. λ = lcm(p-1, q-1), μ is the modular inverse, satisfying μ = (L(g λ modn 2 )) -1 mod n, p and q are two distinct prime numbers, n = p * q.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory; characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instructions stored thereon; characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 7.

Images