CN109828895A - A kind of standardized test method and device of related data difference secret protection performance - Google Patents
A kind of standardized test method and device of related data difference secret protection performance Download PDFInfo
- Publication number
- CN109828895A CN109828895A CN201910033584.5A CN201910033584A CN109828895A CN 109828895 A CN109828895 A CN 109828895A CN 201910033584 A CN201910033584 A CN 201910033584A CN 109828895 A CN109828895 A CN 109828895A
- Authority
- CN
- China
- Prior art keywords
- data
- correlation function
- observation window
- secret protection
- auto
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The invention discloses the standardized test methods and device of a kind of related data difference secret protection performance; the observation window that this method passes through setting specific length; scrambled data to be tested is segmented; then the auto-correlation function and cross-correlation function of scrambled data section are calculated; and the noise in optimal filter filtering scrambled data section is designed, secret protection intensity finally is counted according to the definition of difference privacy to the data of output end.Standardized test method and device in the present invention based on filtering solve the problems, such as that the secret protection performance of existing related data difference method for secret protection can not be measured and be compared, are of great significance for the difference secret protection of related data.
Description
Technical field
The present invention relates to a kind of standardization of information security technology more particularly to related data difference secret protection performance surveys
Method for testing and device.
Background technique
In current method for secret protection, can preferably equilibrium data safety and availability be Dwork in
The difference secret protection frame proposed in 2006, it is substantially a kind of random perturbation strategy, has stringent maxim
Model, and there is no limit can preferably support data point while guaranteeing personal secrets to the background knowledge of attacker
Class application is excavated in analysis, becomes the popular domain of secret protection research in recent years.Difference privacy is initially to solve by mutually solely
The privacy leakage problem for the static data collection that vertical data are constituted and propose, it is independent identically distributed by being superimposed to statistical result
Noise, so that the probability of privacy leakage controls in the strength range of setting.But difference privacy is when handling related data, due to
The noise that data to be protected are related and are added is independent, causes actually active secret protection intensity lower than preset value.
In view of unique advantage of the difference privacy in big data secret protection, there are Many researchers to be expanded to dependency number
According to protection.Current researchers mainly convert two angles from correlation modeling and data and improve.Correlation modeling
Aspect proposes the correlation that data are indicated using Markov model, Bayesian model, correlation matrix model etc., by mould
Global sensitivity function of the parameter as weight calculation difference privacy in type, and according to the hidden of global sensitivity function and setting
Private protection Intensity Design noise;In mode based on data transformation, mainly related data is transformed at independent data
Noise is added in transformation coefficient and carries out inverse transformation, finally issues the sequence after inverse transformation for reason.
Existing method attempts to use modeling or the mode of transformation solves the difference Privacy Protection of related data, but shows
Have method be added noise be still it is independent identically distributed, mainly has the following two problems: 1. existing method is all from protection
Angle propose corresponding solution, due to data to be protected are related and noise is independent, in theory, attacker is still
Noise filtering can be crossed from the noisy data of publication, causes actually active secret protection intensity lower than setting value, but at present simultaneously
Without attacking the test method with standard accordingly, to test the actually active secret protection intensity of existing method;2. due to
The secret protection performance of ununified standardized privacy strength test method, existing method can not carry out lateral comparison and degree
Amount.
Summary of the invention
The technical problem to be solved in the present invention is that for the defects in the prior art, it is hidden to provide a kind of related data difference
The standardized test method and device of private protective value.
The technical solution adopted by the present invention to solve the technical problems is: a kind of related data difference secret protection performance
Standardized test method, comprising the following steps:
Step S1, data prediction: reading the sequence of scrambled data of publication, will be scrambled using the observation window of setting length
Data sectional, using each segment data subsequence as short-term stationary process processing, including following sub-step:
Step S1-1 reads in the scrambled data collection of publication, is denoted as X={ x1,x2,…,xn, wherein n is the data in X
Number;
Step S1-2 sets observation window, and length l, since i-th of data of publication data set X, reading in length is
The data segment to be processed of l is denoted as W, W={ xi,xi+1,…,xi+l-1, xi∈ X, 1≤i≤n-l;
Step S2, filter design: calculating the auto-correlation function and cross-correlation function of the scrambled data section of publication, according to certainly
Correlation function and cross-correlation function design filter, including following sub-step:
Step S2-1, the auto-correlation function R of calculating observation window internal data field WW(xi,xi+l-1), auto-correlation function RW(xi,
xi+l-1) calculation formula are as follows:
Wherein, E [] indicates expectation, xt TIndicate xtTransposition;
Step S2-2, the cross-correlation function P of calculating observation window internal data field WW(xi,xi+l-1), cross-correlation function PW(xi,
xi+l-1) calculation formula are as follows:
Step S2-3 calculates the impulse response of filter, the meter of impulse response according to auto-correlation function and cross-correlation function
Calculation mode are as follows: h (n)=f (RW(xi,xi+l-1))*K(PW(xi,xi+l-1)),
Wherein, f (RW(xi,xi+l-1)) and K (PW(xi,xi+l-1)) be respectively
RW(xi,xi+l-1) and PW(xi,xi+l-1) function, * indicate convolution algorithm;
The scrambled data collection X of publication is filtered using the filter that step S2 is obtained, obtains filtered data by step S3
Collect X ', including following sub-step:
The scrambled data section W that length is l by step S3-1 is input to the filtering that impulse response is h (n) as input terminal
Device;
Step S3-2 slides observation window, reads in l data from remaining untreated data set;
Step S3-3 repeats step S1-2, S2, S3-1 and S3-2, until scrambled data collection X is disposed, is filtered
Data set X ' afterwards;
Step S4 calculates privacy intensity: calculating separately probability density Pr (X), the Pr (X ') of filtering front and back X and X ', calculates
The calculation formula of effective secret protection intensity ε ', ε ' after filtering are as follows:
According to the above scheme, in the step S2-3, according to the preferred filter of auto-correlation function and cross-correlation function design
Impulse response be h (n)=RW(xi,xi+l-1)-1PW(xi,xi+l-1)。
According to the above scheme, the value range of the length l of observation window is 50 to 100 in the step S1-2.
According to the above scheme, in the step S3-2, the sliding type of observation window includes intersection and non-intersecting two kinds.
According to the above scheme, in the step S3-2, the sliding type of observation window is intersection sliding, and intersection data number is set
It is set to the 1/3 to 1/2 of the length of observation window.
Correspondingly, the present invention also provides a kind of related data difference privacy performance standardized test device based on filtering,
Include:
Data preprocessing module will scramble number using the observation window of setting length for reading in the scrambled data of publication
According to segmentation, using each segment data subsequence as short-term stationary process processing;
The data preprocessing module includes following submodule,
Data read in submodule, for reading in the scrambled data collection of publication, are denoted as X={ x1,x2,…,xn, wherein n X
In data amount check;
Sliding window submodule, the observation window for passing through a length as l is by the initial data set segmentation of reading, often
A segment data subsequence is as short-term stationary process processing, and since i-th of data of scrambled data collection X, reading length is l
Scrambled data section, be denoted as W, W={ xi,xi+1,…,xi+l-1, xi∈ X, 1≤i≤n-l;
Filter designs module, filters for being calculated according to the auto-correlation function and cross-correlation function of segment data subsequence
The impulse response of device, the calculation of impulse response are as follows: h (n)=f (RW(xi,xi+l-1))*K(PW(xi,xi+l-1)),
Wherein, f (RW(xi,xi+l-1)) and K (PW(xi,xi+l-1)) respectively
It is RW(xi,xi+l-1) and PW(xi,xi+l-1) function, * indicate convolution algorithm;The filter designs module
Auto-correlation function computational submodule and cross-correlation function computational submodule;
Auto-correlation function computational submodule, the auto-correlation function R of calculating observation window internal data field WW(xi,xi+l-1), from
Correlation function RW(xi,xi+l-1) calculation formula are as follows:
Wherein, E [] indicates expectation, xt TIndicate xtTransposition;
Cross-correlation function computational submodule, the cross-correlation function P for calculating observation window internal data field WW(xi,
xi+l-1), cross-correlation function PW(xi,xi+l-1) calculation formula are as follows:
Filter module is filtered for all segment data subsequence data filterings in the scrambled data collection X to publication
Data set X ' after wave includes following submodule,
Submodule is filtered, is l, auto-correlation function R by length for being filtered to every section of segment data subsequenceW(xi,
xi+l) scrambled data section W as input terminal be input to impulse response be h (n)=RW(xi,xi+l-1)-1PW(xi,xi+l-1) filter
Wave device;
Iterative processing module, for sliding observation window, successively by the sub- sequence of segment data untreated in scrambled data collection X
Column are sent into filter and design module;
Privacy intensity statistics module, for the secret protection intensity value before and after statistical filtering;Calculate separately filtering front and back X and
Probability density Pr (X), the Pr (X ') of X ', the calculation formula of effective secret protection intensity ε ', ε ' after statistical filtering are as follows:
According to the above scheme, in the filter design module, optimal filter is calculated according to auto-correlation function and cross-correlation function
The impulse response of wave device is h (n)=RW(xi,xi+l-1)-1PW(xi,xi+l-1)。
According to the above scheme, in the sliding window submodule, the value range of the length l of observation window is 50 to 100.
According to the above scheme, in the iterative processing module, the sliding type of observation window includes intersection and non-intersecting two
Kind.
According to the above scheme, in the iterative processing module, the sliding type of observation window is intersection sliding, intersection data
Number is set as the 1/3 to 1/2 of the length of observation window.
The beneficial effect comprise that:
(1) invention tests existing related data difference method for secret protection secret protection intensity by design optimal filter
Actually active value, compensate for the blank of related data difference secret protection strength metric method;
(2) test method that invention provides is also used as a kind of Standardized test instruments, to the hidden of current guard method
Private protection intensity validity carries out lateral comparison and measurement;
(3) implementation process and step of the invention, utilization, auto-correlation function and cross-correlation function including sliding window
Generating mode etc., reduces computation complexity, is convenient for algorithm efficient implementation.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples, in attached drawing:
Fig. 1 is delivery system general illustration provided in an embodiment of the present invention.
Fig. 2 is that the principle of related data difference secret protection performance standardized test method provided in an embodiment of the present invention is shown
It is intended to.
Fig. 3 is dissemination method flow chart provided in an embodiment of the present invention.
Fig. 4 is the structural schematic diagram of related data difference secret protection performance standardized test system of the embodiment of the present invention.
Fig. 5 is the simulation result schematic diagram of the embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that described herein, specific examples are only used to explain the present invention, is not used to limit
The fixed present invention.
As shown in Figure 1, coming below by taking the data set that the nearest 10 years daily transaction journal amount of money of some company is constituted as an example
Illustrate specific implementation step of the invention, it is known that be relevant, to be tested data between the daily transaction journal amount of money be existing
There is related data difference privacy mechanism that the scrambled data collection after noise is added, target is using provided by the invention based on filtering
Standardized test method tests the actually active secret protection intensity of existing method.Note scrambled data integrates as X, it is known that X includes
3650 discrete sampling points.
Technical solution of the present invention institute providing method can be used computer software technology and realize automatic running process, Fig. 1 and Fig. 2
It is the group method flow chart and schematic illustration of the embodiment of the present invention respectively, referring to Fig. 1 and Fig. 2, in conjunction with of the invention real in Fig. 3
The specific steps flow chart of example is applied, the present invention gives the standardized test method and system of related data difference secret protection performance
Embodiment specific steps include:
Step S1: data prediction.The scrambled data for reading in publication, using the observation window of specific length by scrambled data
Segmentation, each segmentation subsequence is as short-term stationary process processing.
In embodiment, the scrambled data collection of publication is the number of nearest 10 years daily transaction journal amount of money compositions of some company
It include 3650 discrete sampling points, the length of observation window can be by those skilled in the art according to the overall length of time series according to collection X
It spends sets itself (such as embodiment is set as 50), is implemented as follows,
Step S1-1 reads in the data in the scrambled data collection X of publication, total length n;
In embodiment, the total length n=3650 of the scrambled data collection X of publication.
Step S1-2, sets an observation window, and size l is read since i-th of moment of scrambled data collection of publication
Enter the data segment to be processed that length is l, is denoted as W, W={ xi,xi+1,…,xi+l-1, xi ∈ X, 1≤i≤n-l;
In embodiment, observation window size l=50 reads 50 data since the 1st moment of sequence X to be released,
It is denoted as W, W={ x1,x2,…,x50}。
Step S2, filter design.The auto-correlation function and cross-correlation function of the scrambled data section of publication are calculated, phase is designed
The optimal filter answered.
In embodiment, the auto-correlation function R of the scrambled data section W of publication is calculatedW(x1,x50) and cross-correlation function PW(x1,
x50), the impulse response of optimal filter is solved, is implemented as follows,
Step S2-1, the auto-correlation function R of calculating observation window internal data field WW(xi,xi+l-1), auto-correlation function RW(xi,
xi+l-1) calculation formula are as follows:
Wherein, E [] indicates expectation, xt TIndicate xtTransposition;
In embodiment, the auto-correlation function calculation of observation window internal data field W are as follows:
Step S2-2, the cross-correlation function P of calculating observation window internal data field WW(xi,xi+l-1), cross-correlation function PW(xi,
xi+l-1) calculation formula are as follows:
In embodiment, the cross-correlation function calculation of observation window internal data field W are as follows:
Step S2-3 calculates impulse response h (n)=R of optimal filterW(xi,xi+l-1)-1PW(xi,xi+l-1);
In embodiment, the calculation of the impulse response of optimal filter are as follows:
H (n)=RW(x1,x50)-1PW(x1,x50)
Step S3, filtering.The scrambled data collection X of publication is filtered, filtered data set X ' is obtained.
In embodiment, the scrambled data collection X of publication is filtered, filtered data set X ' is obtained, is implemented as follows,
It is h (n)=R that the scrambled data section W that length is l by step S3-1, which is input to impulse response as input terminal,W(xi,
xi+l-1)-1PW(xi,xi+l-1) filter;
In embodiment, it is h (n)=R that the scrambled data section W for being l using length, which is input to impulse response as input terminal,W(x1,
x50)-1PW(x1,x50) filter;
Step S3-2 slides observation window, reads in l data from remaining untreated data set;
In embodiment, the sliding type of window can have intersection and non-intersecting two kinds, and when implementation can be by those skilled in the art
Member voluntarily selects sliding type according to the correlation properties of data, in the present embodiment, using the sliding type of intersection.In addition, if
Using the sliding type of intersection, the data amount check of intersection can be by those skilled in the art's rule of thumb sets itself, this reality
It applies in example, the data amount check of intersection is set as 25.
Step S3-3 repeats step S1, S2 and S3-1, until scrambled data collection X is disposed, obtains filtered data
Collect X '.
In embodiment, if needed using the sliding window mode of intersection to the data segment in new observation window
Step S1, S2 and S3-1 are repeated, until scrambled data collection X is disposed.
Step S4 counts privacy intensity.Calculate separately probability density Pr (X), the Pr (X ') of filtering front and back X and X ', statistics
The calculation formula of effective secret protection intensity ε ', ε ' after filtering are as follows:
In embodiment, probability density Pr (X), the Pr (X ') of filtering front and back X and X ' are calculated separately, it is effective after statistical filtering
The calculation formula of secret protection intensity ε ', ε ' are as follows:
About emulation experiment
Experimental situation is Windows7 system, and processor is Intel Xeon E3-1240v5 3.50GHz, memory
16.0GB, programmed environment are MATLAB R2014b.It is had chosen herein from four fields for being related to traffic, network, medical treatment and finance
The different time series data collection of Trajectory, Netrace, Flu and Unemployment4 correlation degree of strength carries out
Experiment, each experiment are run 1000 times, assess the validity of test model proposed in this paper and existing guard method, experiment pair
Ratio method includes raw differential privacy mechanism (Baseline), DFT, DWT, CIM, Markov and Bayesian.
It is real at identical secret protection intensity settings ε due to being influenced by random noise and data set correlation
The secret protection intensity and setting value measured on the data set of border will be different.Experiment is to 4 after ε-difference secret protection
Data sequence is inquired, and practical privacy intensity value ε ' of the existing method on 4 data sets is counted.Experimental data and to score
Analysis figure is respectively as shown in table 1 and Fig. 5.
The practical secret protection intensity experiment result of 4 kinds of data sets of table 3-2
From figure 5 it can be seen that the practical secret protection intensity of each method respectively has not when protecting the same data set
Same: for Trajectory, as ε=0.5, the practical secret protection intensity ε ' of MCMC methodology is 0.573, and DWT
ε ' is 0.952.Other data sets also have similar trend, for example, for Unemployment, when ε=0.5,
The ε ' of Bayesian is 0.579, and the ε ' of CIM is 0.621.From the experimental data of table 1 it is found that same guarantor on different data collection
The privacy intensity of maintaining method is also different: as ε=0.1, for Trajectory, the ε ' of Bayesian is 0.165,
It and is 0.135 to the ε ' of Unemployment.
Furthermore, it is possible to observe, when protecting the same data sequence, the ε ' of MCMC, Bayesian and CIM will be lower than
DWT and FPA.This shows that the privacy intensity of the method (MCMC, Bayesian and CIM) of modeling is higher than the method (DWT of transformation
And FPA).From experimental result as can be seen that existing method is after filtering operation, the ε ' value counted is higher than setting value ε,
Illustrate that practical secret protection intensity decreases compared to setting value, illustrates that the Filtering Attacks model of this patent is effective.
It should be understood that for those of ordinary skills, it can be modified or changed according to the above description,
And all these modifications and variations should all belong to the protection domain of appended claims of the present invention.
Claims (10)
1. a kind of standardized test method of related data difference secret protection performance, which comprises the following steps:
Step S1, data prediction: reading the sequence of scrambled data of publication, using the observation window of setting length by scrambled data
Segmentation, using each segment data subsequence as short-term stationary process processing, including following sub-step:
Step S1-1 reads in the scrambled data collection of publication, is denoted as X={ x1,x2,…,xn, wherein n is the data amount check in X;
Step S1-2 sets observation window, and length l, since i-th of data of publication data set X, reading in length is l's
Data segment to be processed is denoted as W, W={ xi,xi+1,…,xi+l-1, xi∈ X, 1≤i≤n-l;
Filter design: step S2 calculates the auto-correlation function and cross-correlation function of the scrambled data section of publication, according to auto-correlation
Function and cross-correlation function design filter, including following sub-step:
Step S2-1, the auto-correlation function R of calculating observation window internal data field WW(xi,xi+l-1), auto-correlation function RW(xi,
xi+l-1) calculation formula are as follows:
Wherein, E [] indicates expectation, xt TIndicate xtTransposition;
Step S2-2, the cross-correlation function P of calculating observation window internal data field WW(xi,xi+l-1), cross-correlation function PW(xi,
xi+l-1) calculation formula are as follows:
Step S2-3 calculates the impulse response of filter, the calculating side of impulse response according to auto-correlation function and cross-correlation function
Formula are as follows: h (n)=f (RW(xi,xi+l-1))*K(PW(xi,xi+l-1)),
Wherein, f (RW(xi,xi+l-1)) and K (PW(xi,xi+l-1)) be respectively
RW(xi,xi+l-1) and PW(xi,xi+l-1) function, * indicate convolution algorithm;
The scrambled data collection X of publication is filtered using the filter that step S2 is obtained, obtains filtered data set by step S3
X ', including following sub-step:
The scrambled data section W that length is l by step S3-1 is input to the filter that impulse response is h (n) as input terminal;
Step S3-2 slides observation window, reads in l data from remaining untreated data set;
Step S3-3 repeats step S1-2, S2, S3-1 and S3-2, until scrambled data collection X is disposed, obtains filtered
Data set X ';
Step S4 calculates privacy intensity: calculating separately probability density Pr (X), the Pr (X ') of filtering front and back X and X ', calculates filtering
The calculation formula of effective secret protection intensity ε ', ε ' afterwards are as follows:
2. the standardized test method of related data difference secret protection performance according to claim 1, which is characterized in that
In the step S2-3, the impulse response of the preferred filter designed according to auto-correlation function and cross-correlation function is h (n)=RW
(xi,xi+l-1)-1PW(xi,xi+l-1)。
3. the standardized test method of related data difference secret protection performance according to claim 1, which is characterized in that
The value range of the length l of observation window is 50 to 100 in the step S1-2.
4. the standardized test method of related data difference secret protection performance according to claim 1, which is characterized in that
In the step S3-2, the sliding type of observation window includes intersection and non-intersecting two kinds.
5. the standardized test method of related data difference secret protection performance according to claim 1, which is characterized in that
In the step S3-2, the sliding type of observation window is intersection sliding, and intersection data number is set as the length of observation window
1/3 to 1/2.
6. a kind of related data difference privacy performance standardized test device based on filtering characterized by comprising
Data preprocessing module, for reading in the scrambled data of publication, the observation window using setting length divides scrambled data
Section, using each segment data subsequence as short-term stationary process processing;
The data preprocessing module includes following submodule,
Data read in submodule, for reading in the scrambled data collection of publication, are denoted as X={ x1,x2,…,xn, wherein n is in X
Data amount check;
Sliding window submodule, for by observation window that length is l by the initial data set segmentation of reading, Mei Gefen
Segment data subsequence is as short-term stationary process processing, and since i-th of data of scrambled data collection X, reading in length is adding for l
Data segment is disturbed, W, W={ x are denoted asi,xi+1,…,xi+l-1, xi∈ X, 1≤i≤n-l;
Filter designs module, for calculating filter according to the auto-correlation function and cross-correlation function of segment data subsequence
Impulse response, the calculation of impulse response are as follows: h (n)=f (RW(xi,xi+l-1))*K(PW(xi,xi+l-1)),
Wherein, f (RW(xi,xi+l-1)) and K (PW(xi,xi+l-1)) respectively
It is RW(xi,xi+l-1) and PW(xi,xi+l-1) function, * indicate convolution algorithm;The filter design module includes from phase
Close function computational submodule and cross-correlation function computational submodule;
Auto-correlation function computational submodule, the auto-correlation function R of calculating observation window internal data field WW(xi,xi+l-1), auto-correlation letter
Number RW(xi,xi+l-1) calculation formula are as follows:
Wherein, E [] indicates expectation, xt TIndicate xtTransposition;
Cross-correlation function computational submodule, the cross-correlation function P for calculating observation window internal data field WW(xi,xi+l-1), mutually
Close function PW(xi,xi+l-1) calculation formula are as follows:
Filter module, for all segment data subsequence data filterings in the scrambled data collection X to publication, after obtaining filtering
Data set X ', include following submodule,
Submodule is filtered, is l, auto-correlation function R by length for being filtered to every section of segment data subsequenceW(xi,xi+l)
It is h (n)=R that scrambled data section W, which is input to impulse response as input terminal,W(xi,xi+l-1)-1PW(xi,xi+l-1) filter;
Iterative processing module successively send segment data subsequence untreated in scrambled data collection X for sliding observation window
Enter filter design module;
Privacy intensity statistics module, for the secret protection intensity value before and after statistical filtering;Calculate separately filtering front and back X's and X '
Probability density Pr (X), Pr (X '), the calculation formula of effective secret protection intensity ε ', ε ' after statistical filtering are as follows:
7. the standardized test device of related data difference secret protection performance according to claim 6, which is characterized in that
It is h (n) according to the impulse response that auto-correlation function and cross-correlation function calculate optimal filter in the filter design module
=RW(xi,xi+l-1)-1PW(xi,xi+l-1)。
8. the standardized test device of related data difference secret protection performance according to claim 6, which is characterized in that
In the sliding window submodule, the value range of the length l of observation window is 50 to 100.
9. the standardized test device of related data difference secret protection performance according to claim 6, which is characterized in that
In the iterative processing module, the sliding type of observation window includes intersection and non-intersecting two kinds.
10. the standardized test device of related data difference secret protection performance according to claim 6, feature exist
In in the iterative processing module, the sliding type of observation window is intersection sliding, and intersection data number is set as observation window
Length 1/3 to 1/2.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910033584.5A CN109828895A (en) | 2019-01-14 | 2019-01-14 | A kind of standardized test method and device of related data difference secret protection performance |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910033584.5A CN109828895A (en) | 2019-01-14 | 2019-01-14 | A kind of standardized test method and device of related data difference secret protection performance |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109828895A true CN109828895A (en) | 2019-05-31 |
Family
ID=66861035
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910033584.5A Pending CN109828895A (en) | 2019-01-14 | 2019-01-14 | A kind of standardized test method and device of related data difference secret protection performance |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109828895A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111797428A (en) * | 2020-06-08 | 2020-10-20 | 武汉大学 | Differential privacy publishing method for medical self-correlation time sequence data |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110282865A1 (en) * | 2010-05-17 | 2011-11-17 | Microsoft Corporation | Geometric mechanism for privacy-preserving answers |
CN105608388A (en) * | 2015-09-24 | 2016-05-25 | 武汉大学 | Differential privacy data publishing method and system based on dependency removal |
CN106227953A (en) * | 2016-07-26 | 2016-12-14 | 武汉大学 | A kind of Laplacian noise sequence generating method with given dependency and maker |
CN106407841A (en) * | 2016-09-28 | 2017-02-15 | 武汉大学 | Correlation time series issuing method and system based on differential privacy |
CN108763956A (en) * | 2018-05-24 | 2018-11-06 | 辽宁工业大学 | A kind of stream data difference secret protection dissemination method based on fractal dimension |
CN108776763A (en) * | 2018-06-08 | 2018-11-09 | 哈尔滨工程大学 | One kind being based on the relevant difference method for secret protection of attribute |
-
2019
- 2019-01-14 CN CN201910033584.5A patent/CN109828895A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110282865A1 (en) * | 2010-05-17 | 2011-11-17 | Microsoft Corporation | Geometric mechanism for privacy-preserving answers |
CN105608388A (en) * | 2015-09-24 | 2016-05-25 | 武汉大学 | Differential privacy data publishing method and system based on dependency removal |
CN106227953A (en) * | 2016-07-26 | 2016-12-14 | 武汉大学 | A kind of Laplacian noise sequence generating method with given dependency and maker |
CN106407841A (en) * | 2016-09-28 | 2017-02-15 | 武汉大学 | Correlation time series issuing method and system based on differential privacy |
CN108763956A (en) * | 2018-05-24 | 2018-11-06 | 辽宁工业大学 | A kind of stream data difference secret protection dissemination method based on fractal dimension |
CN108776763A (en) * | 2018-06-08 | 2018-11-09 | 哈尔滨工程大学 | One kind being based on the relevant difference method for secret protection of attribute |
Non-Patent Citations (2)
Title |
---|
王豪等: "CLM:面向轨迹发布的差分隐私保护方法┺狣?", 《通信学报》 * |
王豪等: "面向轨迹聚类的差分隐私保护方法?", 《华中科技大学学报(自然科学版)》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111797428A (en) * | 2020-06-08 | 2020-10-20 | 武汉大学 | Differential privacy publishing method for medical self-correlation time sequence data |
CN111797428B (en) * | 2020-06-08 | 2024-02-27 | 武汉大学 | Medical self-correlation time sequence data differential privacy release method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Rao et al. | Global filter networks for image classification | |
Wadhwa et al. | Motion microscopy for visualizing and quantifying small motions | |
CN106407841B (en) | Correlation time sequence dissemination method and system based on difference privacy | |
Vollmerhausen et al. | New metric for predicting target acquisition performance | |
Cornish et al. | Bayeswave: Bayesian inference for gravitational wave bursts and instrument glitches | |
Zhou et al. | Blind source separation based vibration mode identification | |
Kaloop et al. | De-noising of GPS structural monitoring observation error using wavelet analysis | |
US20080147763A1 (en) | Method and apparatus for using state space differential geometry to perform nonlinear blind source separation | |
Neri | Estimation of nonlinear psychophysical kernels | |
Sheluhin et al. | Network traffic anomalies detection using a fixing method of multifractal dimension jumps in a real-time mode | |
Morosin et al. | Spatio-temporal analysis of chromospheric heating in a plage region | |
Heidari et al. | Investigating the effect of soil type on non-linear response spectrum using wavelet theory | |
Pont et al. | Singularity analysis of digital signals through the evaluation of their unpredictable point manifold | |
CN109828895A (en) | A kind of standardized test method and device of related data difference secret protection performance | |
Melnik et al. | Local activity indicators for hard-switching adaptive filtering of images with mixed noise | |
Katsevich | Local tomography for the generalized Radon transform | |
Girault et al. | Local stationarity of graph signals: insights and experiments | |
Levant et al. | Wrinkling instability of vesicles in steady linear flow | |
Paul et al. | Defending medical image diagnostics against privacy attacks using generative methods: Application to retinal diagnostics | |
CN104063864A (en) | Image fuzziness assessment method based on quaternary phase congruency model | |
Liu et al. | Despeckling PolSAR images with an adaptive bilateral filter | |
Chen et al. | New image quality assessment method using wavelet leader pyramids | |
Bisht | Methods for structural health monitoring and damage detection of civil and mechanical systems | |
Wang et al. | A modified frequency estimation equating method for the common-item nonequivalent groups design | |
Ouyang et al. | Segmented gray-code kernels for fast pattern matching |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
AD01 | Patent right deemed abandoned | ||
AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20221206 |