CN-121997376-A - Differential privacy protection method based on significant point privacy budget reallocation

CN121997376ACN 121997376 ACN121997376 ACN 121997376ACN-121997376-A

Abstract

The invention relates to the technical field of differential privacy, in particular to a differential privacy protection method, device and equipment based on the redistribution of significant point privacy budgets, which comprises the steps of obtaining time sequence data of a user and presetting the size of a sliding time window and window-level privacy budgets; dividing time series data into a plurality of sliding time windows, uniformly distributing window-level privacy budgets to each time point in the sliding time windows, identifying salient points and non-salient points in each sliding time window, recovering initial privacy budgets corresponding to the non-salient points, distributing the recovered initial privacy budgets to the nearest salient point after the time point where the non-salient points are located, updating the privacy budgets of the salient points, generating disturbance salient point data by utilizing the updated privacy budgets of the salient points, and reconstructing complete time series data by utilizing a least square method according to the disturbance salient point data. The invention can ensure that the reconstructed complete time sequence data meets the differential privacy protection requirement.

Inventors

HU CHUNQIANG
DAI JING
CAI BIN
CHEN JIAJUN
SANG CHUNYAN
QIU BIN

Assignees

重庆大学

Dates

Publication Date: 20260508
Application Date: 20260202

Claims (9)

1. A differential privacy protection method based on significant point privacy budget reallocation, the method comprising: acquiring time sequence data of a user, and presetting a sliding time window size and a window-level privacy budget corresponding to the sliding time window size, wherein the time sequence data is personal health monitoring data; Dividing the time series data into a plurality of sliding time windows according to the sliding time window size; uniformly distributing window-level privacy budgets to each time point in a sliding time window to obtain initial privacy budgets of each time point; Within each sliding time window, identifying salient points and non-salient points according to the data change trend of adjacent time points; recovering the initial privacy budget corresponding to the non-salient point, distributing the recovered initial privacy budget to the nearest salient point after the time point of the non-salient point, and updating the privacy budget of the salient point by utilizing the recovered initial privacy budget; performing local differential privacy disturbance on the salient point data in the sliding time window by using the updated privacy budget of the salient point to generate disturbance salient point data; and reconstructing complete time sequence data by using a least square method according to the disturbance salient point data.
2. The differential privacy preserving method based on significant point privacy budget reallocation as claimed in claim 1, wherein the identifying significant points and non-significant points according to data change trends of adjacent time points comprises: calculating a first-order difference of adjacent time point data: , wherein, For the point in time Is used for the data value of (a), For the point in time Is a function of the data value of the previous time point, For the point in time Is the first order difference of (a); If the condition is satisfied Determining the time point Is a significant point, wherein, For the point in time First order difference of the previous time point; if the condition is not satisfied Determining the time point Is a non-salient point.
3. The differential privacy preserving method based on salient point privacy budget reallocation of claim 1, wherein the performing local differential privacy perturbation on salient point data within a sliding time window comprises: The following Laplace mechanism is adopted to execute local differential privacy disturbance on the salient point data: ; Wherein, the As disturbance salient point data after local differential privacy disturbance, As the data of the salient points, For the global sensitivity to be a function of the global sensitivity, Is the first Privacy budgets for salient points updated with a sliding time window.
4. The differential privacy preserving method based on significant point privacy budget reallocation as set forth in claim 1, wherein the reconstructing complete time series data using a least square method from the perturbed significant point data comprises: Reconstructing the complete time series data using the following least squares objective function: ; Wherein, the For regularized least squares estimation functions, For the complete time series data after reconstruction, In order to smooth the weight parameter(s), For the salient point time index, For the saliency point estimate value, As a disturbance observation of a salient point, As the total number of time points in the time series data, For the point in time Is used for the estimation of the (c), For the point in time Is used for the estimation of the estimated value of (a).
5. The differential privacy preserving method based on significant point privacy budget reallocation as claimed in claim 1, wherein the uniformly allocating the window level privacy budget to each time point within the sliding time window, obtaining an initial privacy budget for each time point, comprises: Distributing fixed window-level privacy budgets to the sliding time window, and equally dividing the window-level privacy budgets to each time point in the sliding time window by utilizing a preset privacy budget distribution formula to obtain initial privacy budgets of each time point, wherein the preset privacy budget distribution formula adopts the following formula: , wherein, Is the first The initial privacy budget of a sliding time window, For a window-level privacy budget, The number of time points for the sliding time window.
6. The differential privacy preserving method based on significant point privacy budget reallocation as set forth in claim 1, wherein the updating the privacy budget of the significant point with the reclaimed initial privacy budget comprises: Updating the privacy budget for the salient point using the following formula: ; Wherein, the Is the first The privacy budget of the updated salient points in the sliding time window, Is the first In a sliding time window The initial privacy budget that was reclaimed for each time step, Is the first An initial privacy budget for points of prominence that are not updated in the sliding time window.
7. Differential privacy preserving apparatus based on significant point privacy budget reallocation, characterized in that it is adapted to implement a differential privacy preserving method based on significant point privacy budget reallocation according to any of the claims 1 to 6, the apparatus comprising: The system comprises a time sequence data dividing module, a sliding time window size dividing module and a window level privacy budget dividing module, wherein the time sequence data dividing module is used for acquiring time sequence data of a user and presetting a sliding time window size and a window level privacy budget corresponding to the sliding time window size; The privacy budget allocation module is used for uniformly allocating window-level privacy budgets to each time point in the sliding time window to obtain initial privacy budgets of each time point, identifying salient points and non-salient points according to the data change trend of adjacent time points in each sliding time window, recovering the initial privacy budgets corresponding to the non-salient points, allocating the recovered initial privacy budgets to the nearest salient point after the time point where the non-salient points are located, and updating the privacy budgets of the salient points by utilizing the recovered initial privacy budgets; The time sequence data reconstruction module is used for executing local differential privacy disturbance on the salient point data in the sliding time window by utilizing the updated privacy budget of the salient points to generate disturbance salient point data, and reconstructing complete time sequence data by utilizing a least square method according to the disturbance salient point data.
8. An electronic device, the electronic device comprising: At least one processor, and A memory communicatively coupled to the at least one processor, wherein, The memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the differential privacy preserving method based on salient point privacy budget reallocation as defined in any one of claims 1 to 6.
9. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements a differential privacy preserving method based on salient point privacy budget reallocation as defined in any of claims 1 to 6.

Description

Differential privacy protection method based on significant point privacy budget reallocation Technical Field The invention belongs to the technical field of differential privacy, and particularly relates to a differential privacy protection method based on significant point privacy budget reallocation. Background With the rapid development of big data technology, public and business decisions are gradually backtracking from static to historical data, turning to dynamic insights and real-time responses to continuous data streams. In the key field of personal health monitoring, real-time analysis of data streams and rapid action take place, the method has become a core capability for improving social management efficiency and preventing systematic risks. However, this real-time data flow-based decision paradigm, while releasing great potential, also places individuals in unprecedented persistent privacy exposure due to their continuous, uninterrupted data acquisition and processing characteristics. In the context of increasingly stringent global privacy regulations (e.g., GDPR, CCPA) and increasingly heightened public privacy awareness, how to improve data availability while protecting user privacy has become a key challenge in mining data value. Differential privacy (DIFFERENTIAL PRIVACY, DP) has become a theoretical foundation for coping with dynamic data environment privacy protection by virtue of strict mathematical definition and quantifiable protection guarantee. The core advantage is that it does not depend on the background knowledge of the attacker and can provide accurate accounting of cumulative privacy losses, which is critical for streaming data processing scenarios where intermediate results need to be continuously published. Conventional differential privacy mechanisms typically ensure that the exposure of individual data points is tightly controlled within a mathematically defined budget whenever they participate in a release by introducing carefully calibrated random noise. Although differential privacy provides strong theoretical support, the value and sensitivity of information often exhibit significant non-uniform distribution characteristics in real streaming data application scenarios. Specifically, a large amount of continuously collected data often reflects only the normal running state of the system, but shows the characteristics of small change amplitude and limited information increment, while data corresponding to a few key time points often reflects behavior change, abnormal state or important events, and has higher information density and potential privacy sensitivity. For example, in a wearable device data stream, most of the time data is in a plateau with low entropy, only a small number of segments contain abnormal events (e.g. arrhythmia, blood oxygen dip) with high clinical diagnostic value. However, existing differential privacy methods for streaming data mostly employ static or uniform privacy budget allocation policies, or budget allocation based on a fixed decay model. This prior art approach suffers mainly from the drawback that existing methods tend to ignore the variability in data importance over time, treat the privacy budget as a "disposable consumable", and evenly distribute over a fixed time window. This coarse-grained allocation pattern results in a large amount of precious privacy budget being consumed on common data points of limited information value, resulting in a significant waste of resources. Furthermore, because the budgets are evenly apportioned, when truly critical data points (e.g., abnormal medical events) occur, often due to insufficient local available budgets, the system must add excessive noise to meet privacy constraints, which can mask salient point information with noise, potentially resulting in less availability of time series data and less accuracy in subsequent data analysis. Disclosure of Invention The invention provides a differential privacy protection method based on significant point privacy budget redistribution, which can ensure that reconstructed complete time sequence data meets the differential privacy protection requirement. In order to achieve the above object, the present invention provides a differential privacy protection method based on significant point privacy budget reallocation, comprising: acquiring time sequence data of a user, and presetting a sliding time window size and a window-level privacy budget corresponding to the sliding time window size, wherein the time sequence data is personal health monitoring data; Dividing the time series data into a plurality of sliding time windows according to the sliding time window size; uniformly distributing window-level privacy budgets to each time point in a sliding time window to obtain initial privacy budgets of each time point; Within each sliding time window, identifying salient points and non-salient points according to the data change trend of adjacent time points; re